Artificial intelligence does wonders in healthcare. The technology helped issue the first COVID-19 warning before the WHO and CDC did so. It can slash the phenomenon of alarm fatigue. IBM’s Watson Health leverages the power of A.I. to bring drugs to the market faster. And it does so while cutting costs by over 50%. Speaking…Artificial Intelligence Discovers Unusual Associations in Medicine — The Medical Futurist
Cloud Security Operations Company DisruptOps Closes $9 Million In Funding
— Read on pulse2.com/disruptops-9-million-funding/
27 Best Chrome OS Tips and Tricks You Should Use
Chrome OS was released in 2011 and since then it has seen a steady improvement across the spectrum. While some are still critical of its viability in today’s market, Chrome OS is doing pretty good in the US and mainly in the education sector. So if you already own one or looking to buy a Chromebook, we have got some great features and recommendations for you. In this article, we bring you the 27 best Chrome OS tips and tricks which are going to help you master this new operating system with ease. We have mentioned unique tricks for both beginners and advanced users so the article has something for every set of users. Now having said all of that, let’s go through the list without further ado.
Best Chrome OS Tips and Tricks in 2020
As mentioned above, the article is categorized into two sections: one for Beginners and another for Advanced users. Based on your level of expertise, you can go through the article and learn more about Chromebooks. Also, let me add that I have included all the new features and hidden settings that are available as on Chrome OS 79 update. So without further delay, let’s start with beginner’s tips and tricks of Chrome OS.
- Chrome OS Tips and Tricks for Beginners
1. Set a Pin for a Seamless Login
As we already know, Chromebook is deeply tied to our Google account. So, before accessing anything, we have to enter our Google account password every single time. I find it very tiring and frankly a chore. Coming from a Windows PC, I want to set a pin for hassle-free login and thankfully, now Google allows you to set a PIN. To create a PIN, open Settings (Cogwheel icon) from Quick Settings menu and open “Screen lock and sign-in” under the “You and Google” section. Here, enter your Google account password for the last time and set a 6-digit Pin. Voila, you are done.
2. Enable Offline Mode
You can enable Offline Mode on Chromebook for Google Docs and Drive. For Google Docs, install this extension first and turn on the checkbox for “Offline” Mode from here. You can also make a particular document offline by going to the Files menu and checking the option for “Make available Offline”. However, make sure to modify all these settings while you are connected to the internet.
3. Master Search on Chromebook
The best part about Chromebook is that Google Search is integrated across the system and the web. So, just press the dedicated search button on your keyboard and start typing and hit enter. No need to open Chrome then open a new Chrome tab and look for things. No matter where you are– under the Settings page or in Chrome itself– the search button always opens the result in a new tab instantly. I would say it’s better than opening a tab through Ctrl + T shortcut. Other than that, you can also search for Chrome apps and settings through the search button.
4. Summon Google Assistant
You can also get your things done through Google Assistant, similar to what you can do on your smartphone. It works quite well and does not have any limitation in comparison to the Assistant on our Android smartphones. Just press “Search” and “A” keys simultaneously on your keyboard and Google Assistant will pop-up, ready to listen to your command. You can also trigger Google Assistant through voice. Open Settings and click on “Search and Assistants” in the left-pane menu. Here, click on “Google Assistant” and enable the toggle for “Ok Google”.
5. Enable Caps Lock
I know the frustration when you find out Chromebooks don’t have a dedicated Caps Lock button. But worry not, you can enable it easily by pressing the “Search” and “Alt” keys at once. A notification will pop-up that Caps Lock has been turned on it will remain there until you turn off Caps Lock using the same shortcut. I know it’s not straightforward, but at least there is a way out.
6. Navigate Chrome OS Through Gestures
Unlike Windows and macOS, you don’t have a range of gestures on Chrome OS. However, you can access the overview window by a simple 3-finger slide up/down gesture. Also, if you are in the Chrome browser, you can use the 3-finger slide in the left and right direction to move between the tabs effortlessly. This is one of the best gestures I have found on Chrome OS and I use it regularly while multi-tasking. Apart from that, you can close a tab by simply tapping 3 fingers at once on a Chrome tab.
7. Preview Files Quickly
Sure, the file manager is not that great on Chrome OS, but I love the fact that it has a quick preview shortcut for any file, similar to what we have on macOS. Just select a file and press the “Space” key. Instantly, the file will be previewed in a large window with all the details (file size, format, etc.) in the right pane. Besides that, you can also switch between files while being in the preview mode and exit it by pressing the “Esc” key. It’s these small things that make Chrome OS a delight to use.
8. Touchpad Scrolling Behavior
Many people who are coming from Windows PC find the default scrolling behavior on Chrome OS the opposite and frankly, it was quite jarring for me too. However, there is a way to change the scrolling behavior from the Settings page. Click on “Device” from the left menu and open “Mouse and touchpad”. Here, scroll down and change the scrolling to “Australian” which is also called natural scrolling on macOS.
9. Virtual Desktop
Users have been asking for a virtual desktop environment on Chrome OS for quite some time. Finally, it’s available in the stable channel starting with Chrome OS 78. For multi-taskers, it’s a huge help as it lets them focus on important things first. So here is how you can access it. Just do a 3-finger slide up gesture and you will find the “New Desk” option on the top-right corner. Alternatively, you can use these shortcuts as well: Shift + Search + = to create new virtual desktop and Shift + Search + – to delete one. You can switch between them using Search + ] and Search + [ shortcuts.
10. Helpful Chrome OS Shortcuts
While there are endless keyboard shortcuts on Chrome OS, here are some crucial ones that will help you navigate through Chrome OS like a pro.
- Search for anything: Press the Search button
- Trigger Google Assistant: Search + A
- Lock your Chromebook: Search + L
- Take a Screenshot: Ctrl + Overview button (you will find it just above the 6 key)
- Delete: Alt + Backspace
- Snap Windows to left/right: Alt + ] and Alt + [
- Keyboard Cheatsheet: Ctrl + Alt + ?
11. Add a Secondary Google Account Inside the Existing Profile
On Windows, you can add multiple Google accounts under a single Chrome profile, but on Chrome OS, Google creates a new profile for every new Google account. While I get the idea behind it, sometimes I just want to check my work email or access Drive files from another Google account and that’s when it becomes frustrating. Nevertheless, now Google has provided an option to add a secondary account in the existing profile itself. Here is how you can access it. Open Settings and click on your name under the “You and Google” section. Here, click on “Add Account” and enter the new Google account credentials. Enjoy!
12. Enable Android Apps
As most of you know, Google has brought Android app support on Chromebooks and it’s a great step towards creating a coherent Google ecosystem. You can install and use millions of Android apps directly from the Google Play Store. In case, it’s not turned on by default, you can enable Play Store from the Settings page. Click on “Apps” on the left pane and open “Google Play Store”. Here, allow various permissions and set up Play Store. That’s it. Enjoy Android apps on your Chromebook.
If Play Store settings are not showing on your Chromebook then it might be in beta and you may have to change your update channel to access Play Store. I have mentioned how to change the update channel in the next section so go through those steps. Also, you can find the details about Play Store support for your Chromebook from here.
13. Enable Floating Keyboard
You can enable the floating keyboard if you are using your Chromebook in a tablet or tent mode. It will help you type using the touch-screen display. So to enable it, open Settings and click on “Advanced” from the left-pane menu. Here, scroll down and open “Accessibility”. Now, just enable the toggle for the “on-screen” keyboard. Now, a keyboard icon will show up on your shelf. Click on it and voila, the keyboard is on your screen.
14. Share your Chromebook
As we have mentioned before the Chromebook is, in some sense, a very private computer. It’s directly linked to your Google account so anyone gaining access to your Chromebook means they can view your photos, check your emails, read your notes among many other things. In a way, it’s not easy to share your Chromebook as it’s with Windows-based computers. So, what if someone asks to use your Chromebook for a while, say your sister or someone from your family? Well, you can create a separate account or better, take advantage of the “Guest Mode”. It lets you access most of the Chrome OS features without adding any account. Open the Quick Settings menu from the bottom-right corner and sign out of your current profile. Now, click on “Guest Mode” at the bottom and you will be in a temporary profile.
15. Reset Your Chromebook
If you are having any problem with your Chrome OS, you can easily factory-reset (called Powerwash on Chrome OS) your machine, similar to Android devices. The great part about Chrome OS is that all your files are synced to Google Drive so you don’t have to worry about data loss. Just open the Settings and click on “Advanced” from the left-pane menu and then move to “Reset Settings”. Here, click on the “Reset” button and the computer will restart to complete the process. After that, sign in with your Google account and everything will fall into its place just like before.
16. Other Chrome OS Tips
If you were reading something on your Android and suddenly moved to a Chromebook then you can continue reading it on your computer easily. Just press the “Search” button once and you will find the link in the first entry. Click on it and there you have it.
Other than that, if you want to quickly call a phone number from a webpage, you can right-click on it and send it to your Android device. Your smartphone will prompt you with a one-tap call notification.
- Chrome OS Tips and Tricks for Advanced Users
17. Add your Android Device
Google has brought support for Android devices on Chromebook which can help you get many things done seamlessly. In case, there is no WiFi available, Google allows the Chromebook to tether your Android device automatically for always-on data connectivity. Apart from that, you can also seamlessly sign into your Chromebook by unlocking your Android device. Basically, both the devices are always in communication so that you get the best of the Google experience. So if you want to add your Android device, open Settings and click on “Connected Devices” in the left pane. Here, set up your Android device and you are good to go.
18. Change Update Channel
If you are someone who wants to taste the new and exciting features of Chrome OS then you will have to jump the ship from Stable to another channel. There are four update channels: Stable, Developer, Beta, and Canary. I would advise you to stick with Developer or Beta channel as these are relatively bug-free than Canary and also allow you to enjoy the upcoming features. So to change the update channel, open Settings -> Click on “About Chrome OS” in the left pane -> Change Channel. Here, choose the update channel of your choice then go back and check for updates. After downloading the update and installation, just restart your Chromebook and you will be in.
19. Create Standalone Apps from Websites
If you want some websites to behave like an app with a separate window and launcher menu then you can create such standalone apps on a Chromebook. However, keep in mind, it’s only possible on those websites which adhere to PWA (Progressive Web App) standard. For example, you can easily create apps from websites of Spotify, Twitter and our own portal, beebom.com. To do so, open the website of your choice and click on the 3-dot menu on the top-right corner. Here, open More Tools -> Create Shortcut -> Check the box for “Open as Window” and click on the “Create” button. Now, you will find the website as a separate app available in your launcher and you can even pin it to your shelf.
20. Run Any Android App
While the Play Store support is great on Chromebook, many users find it inadequate as some of their favorite apps are officially not available on the Play Store. In that case, you will have to download the APK and run it through the ARC Welder. I have explained the steps in great detail in a separate article on how to install the Kodi app on Chromebook. You can follow a similar step for installing other apps as well.
21. Force Android App to be Resizable
If some Android app is not allowing you to resize its window size or does not open in the landscape orientation then you can force enable the flag from the Android Settings page. Open Chrome OS Settings and click on Apps in the left-pane -> Google Play Store -> Manage Android Preferences. Now we are on the Android Settings page so navigate to System -> About Device -> click on the “Build Number” continuously for 7-8 times. Now, go back and open the “Developer Options“. Scroll down to the bottom and enable the toggle for “Force activities to be resizable”. Finally, restart your Chromebook and the Android apps should be resizable now.
22. Use Third-Party DNS
Similar to Android smartphones, we can change the DNS configuration on Chromebooks too. You can use third-party DNS like Cloudfare or OpenDNS to improve your internet speed. So to modify the DNS, open Settings and click on the WiFi network that you are connected to. Here, scroll down and click on “Network” to expand the menu. Finally, select “Custom name servers” and enter your choice of DNS address.
23. Find System Information
Chrome OS is very minimal and consumer-facing so it does not provide details like system memory, CPU usage, background processes and similar metrics for the end-user. However, if you want to access those metrics, there are some hidden commands that let you do it. Just enter
chrome://system/ in the address bar and hit enter. You will get all the information from hardware to software on this page. Other than that, you can also install an extension called Cog (Free) which allows you to monitor many metrics like CPU usage, temperature, etc. in a slick and graphical interface.
24. Share Windows Folders to Chromebook (Network File Share)
One of the best features of Windows OS is that you can seamlessly access files and folders from other Windows computers using a common wireless network. So to make the desktop experience similar on Chrome OS, Google has also brought Network File Share and it’s embedded in the native File Manager. Basically, you can remotely access all your Windows files and folders on your Chromebook without any hassle. I have written a detailed guide on how to use this feature so go through the steps and you will be all set.
25. Enable Linux on Chromebook
Recently, Google added support for Linux on Chrome OS and it’s simply magical to use both the operating systems side by side. While the project is still in beta, the Terminal works quite well with support for all the Linux commands. You can even install Linux apps on Chrome OS, but let me clarify, at this point, the performance is not that great. Anyway, if you want to enable it, open Settings and simply navigate to “Linux (Beta)”. Here, turn on the toggle for Linux and go through the on-screen instructions.
However, do note that Chrome OS and Linux, both have different storage systems so you will have to share your local folders to Linux. You can do so from the native File Manager itself. Just right-click on the folder that you want to share and choose “Select with Linux”. That’s it.
26. Enable ADB on Chrome OS (Android Debugging Bridge)
Earlier, there was an unofficial way to enable ADB on Chrome OS through Linux and platform tools. However, things have changed now since Google has brought native ADB support on Chrome OS. You can find the dedicated page from Settings -> Linux -> Develop Android apps and enable the toggle for “Enable ADB Debugging”. You are good to go.
27. Learning Chrome Shell (Crosh)
Similar to Command Prompt on Windows, Bash in Linux and Terminal on macOS, Chrome OS has its command line called Chrome Shell (Crosh). It allows you to modify many system settings that are otherwise not available in Settings or Flags page. So, if you want to dive deep into Chrome OS, you must learn what Crosh offers and how it can help you do many things. We have already written a tutorial on the best Crosh commands so go through that as a primer.
How To Become A Hacker
Eric Steven Raymond
Copyright © 2001 Eric S. Raymond
|Revision 1.52||03 Jasnuary 2020||esr|
|Go makes a place as a plausible learning language, displacing Java.|
|Revision 1.51||06 October 2017||esr|
|Link to “Things Every Hacker Once Knew.” Mention USB-stick distros. Many updated translation links.|
|Revision 1.50||19 July 2015||esr|
|Added link to “Let’s Go Larval”.|
|Revision 1.49||21 November 2014||esr|
|Added link to “How To Learn Hacking”.|
|Revision 1.48||19 June 2014||esr|
|freshmeat/freecode is dead, alas.|
|Revision 1.47||20 May 2014||esr|
|Fix up various stale links. Join a hackerspace!|
|Revision 1.46||25 Sep 2013||esr|
|Add micropatronage explanation and gittip link. Why you should not ask me for advice on how to get started.|
|Revision 1.45||12 May 2013||esr|
|Open Solaris isn’t, and Unity screwed the pooch.|
|Revision 1.44||20 May 2012||esr|
|Updated the critique of Java.|
|Revision 1.43||07 Feb 2011||esr|
|Python passed Perl in popularity in 2010.|
|Revision 1.42||22 Oct 2010||esr|
|Added “Historical note”.|
|Revision 1.40||3 Nov 2008||esr|
|Revision 1.39||14 Aug 2008||esr|
|Revision 1.38||8 Jan 2008||esr|
|Deprecate Java as a language to learn early.|
|Revision 1.37||4 Oct 2007||esr|
|Recommend Ubuntu as a Unix distro for newbies.|
Table of Contents
Why This Document?
As editor of the Jargon File and author of a few other well-known documents of similar nature, I often get email requests from enthusiastic network newbies asking (in effect) “how can I learn to be a wizardly hacker?”. Back in 1996 I noticed that there didn’t seem to be any other FAQs or web documents that addressed this vital question, so I started this one. A lot of hackers now consider it definitive, and I suppose that means it is. Still, I don’t claim to be the exclusive authority on this topic; if you don’t like what you read here, write your own.
If you are reading a snapshot of this document offline, the current version lives at http://catb.org/~esr/faqs/hacker-howto.html.
Note: there is a list of Frequently Asked Questions at the end of this document. Please read these—twice—before mailing me any questions about this document.
Numerous translations of this document are available: Arabic Belorussian Bulgarian Chinese, Czech. Danish Dutch Estonian French German, Greek Hungarian, Italian Hebrew, Japanese Lithuanian Norwegian, Persian Polish Portuguese (Brazilian), Romanian Spanish, Turkish, and Swedish. Note that since this document changes occasionally, they may be out of date to varying degrees.
The five-dots-in-nine-squares diagram that decorates this document is called a glider. It is a simple pattern with some surprising properties in a mathematical simulation called Life that has fascinated hackers for many years. I think it makes a good visual emblem for what hackers are like — abstract, at first a bit mysterious-seeming, but a gateway to a whole world with an intricate logic of its own. Read more about the glider emblem here.
If you find this document valuable, please support me on Patreon or SubscribeStar. And consider also supporting other hackers who have produced code that you use and value via Loadsharers. Lots of small but continuing donations add up quickly, and can free the people who have given you gifts of their labor to create more value.
What Is a Hacker?
The Jargon File contains a bunch of definitions of the term ‘hacker’, most having to do with technical adeptness and a delight in solving problems and overcoming limits. If you want to know how to become a hacker, though, only two are really relevant.
There is a community, a shared culture, of expert programmers and networking wizards that traces its history back through decades to the first time-sharing minicomputers and the earliest ARPAnet experiments. The members of this culture originated the term ‘hacker’. Hackers built the Internet. Hackers made the Unix operating system what it is today. Hackers make the World Wide Web work. If you are part of this culture, if you have contributed to it and other people in it know who you are and call you a hacker, you’re a hacker.
The hacker mind-set is not confined to this software-hacker culture. There are people who apply the hacker attitude to other things, like electronics or music — actually, you can find it at the highest levels of any science or art. Software hackers recognize these kindred spirits elsewhere and may call them ‘hackers’ too — and some claim that the hacker nature is really independent of the particular medium the hacker works in. But in the rest of this document we will focus on the skills and attitudes of software hackers, and the traditions of the shared culture that originated the term ‘hacker’.
There is another group of people who loudly call themselves hackers, but aren’t. These are people (mainly adolescent males) who get a kick out of breaking into computers and phreaking the phone system. Real hackers call these people ‘crackers’ and want nothing to do with them. Real hackers mostly think crackers are lazy, irresponsible, and not very bright, and object that being able to break security doesn’t make you a hacker any more than being able to hotwire cars makes you an automotive engineer. Unfortunately, many journalists and writers have been fooled into using the word ‘hacker’ to describe crackers; this irritates real hackers no end.
The basic difference is this: hackers build things, crackers break them.
If you want to be a hacker, keep reading. If you want to be a cracker, go read the alt.2600 newsgroup and get ready to do five to ten in the slammer after finding out you aren’t as smart as you think you are. And that’s all I’m going to say about crackers.
The Hacker Attitude
Hackers solve problems and build things, and they believe in freedom and voluntary mutual help. To be accepted as a hacker, you have to behave as though you have this kind of attitude yourself. And to behave as though you have the attitude, you have to really believe the attitude.
But if you think of cultivating hacker attitudes as just a way to gain acceptance in the culture, you’ll miss the point. Becoming the kind of person who believes these things is important for you — for helping you learn and keeping you motivated. As with all creative arts, the most effective way to become a master is to imitate the mind-set of masters — not just intellectually but emotionally as well.
Or, as the following modern Zen poem has it:
To follow the path:
look to the master,
follow the master,
walk with the master,
see through the master,
become the master.
So, if you want to be a hacker, repeat the following things until you believe them:
1. The world is full of fascinating problems waiting to be solved.
Being a hacker is lots of fun, but it’s a kind of fun that takes lots of effort. The effort takes motivation. Successful athletes get their motivation from a kind of physical delight in making their bodies perform, in pushing themselves past their own physical limits. Similarly, to be a hacker you have to get a basic thrill from solving problems, sharpening your skills, and exercising your intelligence.
If you aren’t the kind of person that feels this way naturally, you’ll need to become one in order to make it as a hacker. Otherwise you’ll find your hacking energy is sapped by distractions like sex, money, and social approval.
(You also have to develop a kind of faith in your own learning capacity — a belief that even though you may not know all of what you need to solve a problem, if you tackle just a piece of it and learn from that, you’ll learn enough to solve the next piece — and so on, until you’re done.)
2. No problem should ever have to be solved twice.
Creative brains are a valuable, limited resource. They shouldn’t be wasted on re-inventing the wheel when there are so many fascinating new problems waiting out there.
To behave like a hacker, you have to believe that the thinking time of other hackers is precious — so much so that it’s almost a moral duty for you to share information, solve problems and then give the solutions away just so other hackers can solve new problems instead of having to perpetually re-address old ones.
Note, however, that “No problem should ever have to be solved twice.” does not imply that you have to consider all existing solutions sacred, or that there is only one right solution to any given problem. Often, we learn a lot about the problem that we didn’t know before by studying the first cut at a solution. It’s OK, and often necessary, to decide that we can do better. What’s not OK is artificial technical, legal, or institutional barriers (like closed-source code) that prevent a good solution from being re-used and force people to re-invent wheels.
(You don’t have to believe that you’re obligated to give all your creative product away, though the hackers that do are the ones that get most respect from other hackers. It’s consistent with hacker values to sell enough of it to keep you in food and rent and computers. It’s fine to use your hacking skills to support a family or even get rich, as long as you don’t forget your loyalty to your art and your fellow hackers while doing it.)
3. Boredom and drudgery are evil.
Hackers (and creative people in general) should never be bored or have to drudge at stupid repetitive work, because when this happens it means they aren’t doing what only they can do — solve new problems. This wastefulness hurts everybody. Therefore boredom and drudgery are not just unpleasant but actually evil.
To behave like a hacker, you have to believe this enough to want to automate away the boring bits as much as possible, not just for yourself but for everybody else (especially other hackers).
(There is one apparent exception to this. Hackers will sometimes do things that may seem repetitive or boring to an observer as a mind-clearing exercise, or in order to acquire a skill or have some particular kind of experience you can’t have otherwise. But this is by choice — nobody who can think should ever be forced into a situation that bores them.)
4. Freedom is good.
Hackers are naturally anti-authoritarian. Anyone who can give you orders can stop you from solving whatever problem you’re being fascinated by — and, given the way authoritarian minds work, will generally find some appallingly stupid reason to do so. So the authoritarian attitude has to be fought wherever you find it, lest it smother you and other hackers.
(This isn’t the same as fighting all authority. Children need to be guided and criminals restrained. A hacker may agree to accept some kinds of authority in order to get something he wants more than the time he spends following orders. But that’s a limited, conscious bargain; the kind of personal surrender authoritarians want is not on offer.)
Authoritarians thrive on censorship and secrecy. And they distrust voluntary cooperation and information-sharing — they only like ‘cooperation’ that they control. So to behave like a hacker, you have to develop an instinctive hostility to censorship, secrecy, and the use of force or deception to compel responsible adults. And you have to be willing to act on that belief.
5. Attitude is no substitute for competence.
To be a hacker, you have to develop some of these attitudes. But copping an attitude alone won’t make you a hacker, any more than it will make you a champion athlete or a rock star. Becoming a hacker will take intelligence, practice, dedication, and hard work.
Therefore, you have to learn to distrust attitude and respect competence of every kind. Hackers won’t let posers waste their time, but they worship competence — especially competence at hacking, but competence at anything is valued. Competence at demanding skills that few can master is especially good, and competence at demanding skills that involve mental acuteness, craft, and concentration is best.
If you revere competence, you’ll enjoy developing it in yourself — the hard work and dedication will become a kind of intense play rather than drudgery. That attitude is vital to becoming a hacker.
Basic Hacking Skills
The hacker attitude is vital, but skills are even more vital. Attitude is no substitute for competence, and there’s a certain basic toolkit of skills which you have to have before any hacker will dream of calling you one.
This toolkit changes slowly over time as technology creates new skills and makes old ones obsolete. For example, it used to include programming in machine language, and didn’t until recently involve HTML. But right now it pretty clearly includes the following:
1. Learn how to program.
This, of course, is the fundamental hacking skill. If you don’t know any computer languages, I recommend starting with Python. It is cleanly designed, well documented, and relatively kind to beginners. Despite being a good first language, it is not just a toy; it is very powerful and flexible and well suited for large projects. I have written a more detailed evaluation of Python. Good tutorials are available at the Python web site; there’s an excellent third-party one at Computer Science Circles.
I used to recommend Java as a good language to learn early, but this critique has changed my mind (search for “The Pitfalls of Java as a First Programming Language” within it). A hacker cannot, as they devastatingly put it “approach problem-solving like a plumber in a hardware store”; you have to know what the components actually do. Now I think it is probably best to learn C and Lisp first, then Java.
There is perhaps a more general point here. If a language does too much for you, it may be simultaneously a good tool for production and a bad one for learning. It’s not only languages that have this problem; web application frameworks like RubyOnRails, CakePHP, Django may make it too easy to reach a superficial sort of understanding that will leave you without resources when you have to tackle a hard problem, or even just debug the solution to an easy one.
A better alternative to Java is to learn Go. This relatively new language is pretty easy to move to from Python, and learning it give you a serious leg up on the possible next step, which is learning C. Additionally, one of the unknowns about the next few years is to what extent Go might actually displace C as a systems-programming language. There is a possible future in which that happens over much of C’s traditional range.
If you get into serious programming, you will eventually have to learn C, the core language of Unix. C++ is very closely related to C; if you know one, learning the other will not be difficult. Neither language is a good one to try learning as your first, however. And, actually, the more you can avoid programming in C the more productive you will be.
C is very efficient, and very sparing of your machine’s resources. Unfortunately, C gets that efficiency by requiring you to do a lot of low-level management of resources (like memory) by hand. All that low-level code is complex and bug-prone, and will soak up huge amounts of your time on debugging. With today’s machines as powerful as they are, this is usually a bad tradeoff — it’s smarter to use a language that uses the machine’s time less efficiently, but your time much more efficiently. Thus, Python.
Other languages of particular importance to hackers include Perl and LISP. Perl is worth learning for practical reasons; it’s very widely used for active web pages and system administration, so that even if you never write Perl you should learn to read it. Many people use Perl in the way I suggest you should use Python, to avoid C programming on jobs that don’t require C’s machine efficiency. You will need to be able to understand their code.
LISP is worth learning for a different reason — the profound enlightenment experience you will have when you finally get it. That experience will make you a better programmer for the rest of your days, even if you never actually use LISP itself a lot. (You can get some beginning experience with LISP fairly easily by writing and modifying editing modes for the Emacs text editor, or Script-Fu plugins for the GIMP.)
It’s best, actually, to learn all five of Python, C/C++, Perl, and LISP. Besides being the most important hacking languages, they represent very different approaches to programming, and each will educate you in valuable ways. Go is not quite to the point where it can be included among the most important hacking languages, but it seems headed for that status.
But be aware that you won’t reach the skill level of a hacker or even merely a programmer simply by accumulating languages — you need to learn how to think about programming problems in a general way, independent of any one language. To be a real hacker, you need to get to the point where you can learn a new language in days by relating what’s in the manual to what you already know. This means you should learn several very different languages.
I can’t give complete instructions on how to learn to program here — it’s a complex skill. But I can tell you that books and courses won’t do it — many, maybe most of the best hackers are self-taught. You can learn language features — bits of knowledge — from books, but the mind-set that makes that knowledge into living skill can be learned only by practice and apprenticeship. What will do it is (a) reading code and (b) writing code.
Peter Norvig, who is one of Google’s top hackers and the co-author of the most widely used textbook on AI, has written an excellent essay called Teach Yourself Programming in Ten Years. His “recipe for programming success” is worth careful attention.
Learning to program is like learning to write good natural language. The best way to do it is to read some stuff written by masters of the form, write some things yourself, read a lot more, write a little more, read a lot more, write some more … and repeat until your writing begins to develop the kind of strength and economy you see in your models.
I have had more to say about this learning process in How To Learn Hacking. It’s a simple set of instructions, but not an easy one.
Finding good code to read used to be hard, because there were few large programs available in source for fledgeling hackers to read and tinker with. This has changed dramatically; open-source software, programming tools, and operating systems (all built by hackers) are now widely available. Which brings me neatly to our next topic…
2. Get one of the open-source Unixes and learn to use and run it.
I’ll assume you have a personal computer or can get access to one. (Take a moment to appreciate how much that means. The hacker culture originally evolved back when computers were so expensive that individuals could not own them.) The single most important step any newbie can take toward acquiring hacker skills is to get a copy of Linux or one of the BSD-Unixes, install it on a personal machine, and run it.
Yes, there are other operating systems in the world besides Unix. But they’re distributed in binary — you can’t read the code, and you can’t modify it. Trying to learn to hack on a Microsoft Windows machine or under any other closed-source system is like trying to learn to dance while wearing a body cast.
Under Mac OS X it’s possible, but only part of the system is open source — you’re likely to hit a lot of walls, and you have to be careful not to develop the bad habit of depending on Apple’s proprietary code. If you concentrate on the Unix under the hood you can learn some useful things.
Unix is the operating system of the Internet. While you can learn to use the Internet without knowing Unix, you can’t be an Internet hacker without understanding Unix. For this reason, the hacker culture today is pretty strongly Unix-centered. (This wasn’t always true, and some old-time hackers still aren’t happy about it, but the symbiosis between Unix and the Internet has become strong enough that even Microsoft’s muscle doesn’t seem able to seriously dent it.)
So, bring up a Unix — I like Linux myself but there are other ways (and yes, you can run both Linux and Microsoft Windows on the same machine). Learn it. Run it. Tinker with it. Talk to the Internet with it. Read the code. Modify the code. You’ll get better programming tools (including C, LISP, Python, and Perl) than any Microsoft operating system can dream of hosting, you’ll have fun, and you’ll soak up more knowledge than you realize you’re learning until you look back on it as a master hacker.
To get your hands on a Linux, see the Linux Online! site; you can download from there or (better idea) find a local Linux user group to help you with installation.
During the first ten years of this HOWTO’s life, I reported that from a new user’s point of view, all Linux distributions are almost equivalent. But in 2006-2007, an actual best choice emerged: Ubuntu. While other distros have their own areas of strength, Ubuntu is far and away the most accessible to Linux newbies. Beware, though, of the hideous and nigh-unusable “Unity” desktop interface that Ubuntu introduced as a default a few years later; the Xubuntu or Kubuntu variants are better.
You can find BSD Unix help and resources at www.bsd.org.
A good way to dip your toes in the water is to boot up what Linux fans call a live CD, a distribution that runs entirely off a CD or USB stick without having to modify your hard disk. This may be slow, because CDs are slow, but it’s a way to get a look at the possibilities without having to do anything drastic.
I have written a primer on the basics of Unix and the Internet.
I used to recommend against installing either Linux or BSD as a solo project if you’re a newbie. Nowadays the installers have gotten good enough that doing it entirely on your own is possible, even for a newbie. Nevertheless, I still recommend making contact with your local Linux user’s group and asking for help. It can’t hurt, and may smooth the process.
3. Learn how to use the World Wide Web and write HTML.
Most of the things the hacker culture has built do their work out of sight, helping run factories and offices and universities without any obvious impact on how non-hackers live. The Web is the one big exception, the huge shiny hacker toy that even politicians admit has changed the world. For this reason alone (and a lot of other good ones as well) you need to learn how to work the Web.
This doesn’t just mean learning how to drive a browser (anyone can do that), but learning how to write HTML, the Web’s markup language. If you don’t know how to program, writing HTML will teach you some mental habits that will help you learn. So build a home page.
But just having a home page isn’t anywhere near good enough to make you a hacker. The Web is full of home pages. Most of them are pointless, zero-content sludge — very snazzy-looking sludge, mind you, but sludge all the same (for more on this see The HTML Hell Page).
To be worthwhile, your page must have content — it must be interesting and/or useful to other hackers. And that brings us to the next topic…
4. If you don’t have functional English, learn it.
As an American and native English-speaker myself, I have previously been reluctant to suggest this, lest it be taken as a sort of cultural imperialism. But several native speakers of other languages have urged me to point out that English is the working language of the hacker culture and the Internet, and that you will need to know it to function in the hacker community.
Back around 1991 I learned that many hackers who have English as a second language use it in technical discussions even when they share a birth tongue; it was reported to me at the time that English has a richer technical vocabulary than any other language and is therefore simply a better tool for the job. For similar reasons, translations of technical books written in English are often unsatisfactory (when they get done at all).
Linus Torvalds, a Finn, comments his code in English (it apparently never occurred to him to do otherwise). His fluency in English has been an important factor in his ability to recruit a worldwide community of developers for Linux. It’s an example worth following.
Being a native English-speaker does not guarantee that you have language skills good enough to function as a hacker. If your writing is semi-literate, ungrammatical, and riddled with misspellings, many hackers (including myself) will tend to ignore you. While sloppy writing does not invariably mean sloppy thinking, we’ve generally found the correlation to be strong — and we have no use for sloppy thinkers. If you can’t yet write competently, learn to.
Status in the Hacker Culture
Like most cultures without a money economy, hackerdom runs on reputation. You’re trying to solve interesting problems, but how interesting they are, and whether your solutions are really good, is something that only your technical peers or superiors are normally equipped to judge.
Accordingly, when you play the hacker game, you learn to keep score primarily by what other hackers think of your skill (this is why you aren’t really a hacker until other hackers consistently call you one). This fact is obscured by the image of hacking as solitary work; also by a hacker-cultural taboo (gradually decaying since the late 1990s but still potent) against admitting that ego or external validation are involved in one’s motivation at all.
Specifically, hackerdom is what anthropologists call a gift culture. You gain status and reputation in it not by dominating other people, nor by being beautiful, nor by having things other people want, but rather by giving things away. Specifically, by giving away your time, your creativity, and the results of your skill.
There are basically five kinds of things you can do to be respected by hackers:
1. Write open-source software
The first (the most central and most traditional) is to write programs that other hackers think are fun or useful, and give the program sources away to the whole hacker culture to use.
(We used to call these works “free software”, but this confused too many people who weren’t sure exactly what “free” was supposed to mean. Most of us now prefer the term “open-source” software).
Hackerdom’s most revered demigods are people who have written large, capable programs that met a widespread need and given them away, so that now everyone uses them.
But there’s a bit of a fine historical point here. While hackers have always looked up to the open-source developers among them as our community’s hardest core, before the mid-1990s most hackers most of the time worked on closed source. This was still true when I wrote the first version of this HOWTO in 1996; it took the mainstreaming of open-source software after 1997 to change things. Today, “the hacker community” and “open-source developers” are two descriptions for what is essentially the same culture and population — but it is worth remembering that this was not always so. (For more on this, see the section called “Historical Note: Hacking, Open Source, and Free Software”.)
2. Help test and debug open-source software
They also serve who stand and debug open-source software. In this imperfect world, we will inevitably spend most of our software development time in the debugging phase. That’s why any open-source author who’s thinking will tell you that good beta-testers (who know how to describe symptoms clearly, localize problems well, can tolerate bugs in a quickie release, and are willing to apply a few simple diagnostic routines) are worth their weight in rubies. Even one of these can make the difference between a debugging phase that’s a protracted, exhausting nightmare and one that’s merely a salutary nuisance.
If you’re a newbie, try to find a program under development that you’re interested in and be a good beta-tester. There’s a natural progression from helping test programs to helping debug them to helping modify them. You’ll learn a lot this way, and generate good karma with people who will help you later on.
3. Publish useful information
Another good thing is to collect and filter useful and interesting information into web pages or documents like Frequently Asked Questions (FAQ) lists, and make those generally available.
Maintainers of major technical FAQs get almost as much respect as open-source authors.
4. Help keep the infrastructure working
The hacker culture (and the engineering development of the Internet, for that matter) is run by volunteers. There’s a lot of necessary but unglamorous work that needs done to keep it going — administering mailing lists, moderating newsgroups, maintaining large software archive sites, developing RFCs and other technical standards.
People who do this sort of thing well get a lot of respect, because everybody knows these jobs are huge time sinks and not as much fun as playing with code. Doing them shows dedication.
5. Serve the hacker culture itself
Finally, you can serve and propagate the culture itself (by, for example, writing an accurate primer on how to become a hacker :-)). This is not something you’ll be positioned to do until you’ve been around for while and become well-known for one of the first four things.
The hacker culture doesn’t have leaders, exactly, but it does have culture heroes and tribal elders and historians and spokespeople. When you’ve been in the trenches long enough, you may grow into one of these. Beware: hackers distrust blatant ego in their tribal elders, so visibly reaching for this kind of fame is dangerous. Rather than striving for it, you have to sort of position yourself so it drops in your lap, and then be modest and gracious about your status.
The Hacker/Nerd Connection
Contrary to popular myth, you don’t have to be a nerd to be a hacker. It does help, however, and many hackers are in fact nerds. Being something of a social outcast helps you stay concentrated on the really important things, like thinking and hacking.
For this reason, many hackers have adopted the label ‘geek’ as a badge of pride — it’s a way of declaring their independence from normal social expectations (as well as a fondness for other things like science fiction and strategy games that often go with being a hacker). The term ‘nerd’ used to be used this way back in the 1990s, back when ‘nerd’ was a mild pejorative and ‘geek’ a rather harsher one; sometime after 2000 they switched places, at least in U.S. popular culture, and there is now even a significant geek-pride culture among people who aren’t techies.
If you can manage to concentrate enough on hacking to be good at it and still have a life, that’s fine. This is a lot easier today than it was when I was a newbie in the 1970s; mainstream culture is much friendlier to techno-nerds now. There are even growing numbers of people who realize that hackers are often high-quality lover and spouse material.
If you’re attracted to hacking because you don’t have a life, that’s OK too — at least you won’t have trouble concentrating. Maybe you’ll get a life later on.
Points For Style
Again, to be a hacker, you have to enter the hacker mindset. There are some things you can do when you’re not at a computer that seem to help. They’re not substitutes for hacking (nothing is) but many hackers do them, and feel that they connect in some basic way with the essence of hacking.
- Learn to write your native language well. Though it’s a common stereotype that programmers can’t write, a surprising number of hackers (including all the most accomplished ones I know of) are very able writers.
- Read science fiction. Go to science fiction conventions (a good way to meet hackers and proto-hackers).
- Join a hackerspace and make things (another good way to meet hackers and proto-hackers).
- Train in a martial-arts form. The kind of mental discipline required for martial arts seems to be similar in important ways to what hackers do. The most popular forms among hackers are definitely Asian empty-hand arts such as Tae Kwon Do, various forms of Karate, Kung Fu, Aikido, or Ju Jitsu. Western fencing and Asian sword arts also have visible followings. In places where it’s legal, pistol shooting has been rising in popularity since the late 1990s. The most hackerly martial arts are those which emphasize mental discipline, relaxed awareness, and precise control, rather than raw strength, athleticism, or physical toughness.
- Study an actual meditation discipline. The perennial favorite among hackers is Zen (importantly, it is possible to benefit from Zen without acquiring a religion or discarding one you already have). Other styles may work as well, but be careful to choose one that doesn’t require you to believe crazy things.
- Develop an analytical ear for music. Learn to appreciate peculiar kinds of music. Learn to play some musical instrument well, or how to sing.
- Develop your appreciation of puns and wordplay.
The more of these things you already do, the more likely it is that you are natural hacker material. Why these things in particular is not completely clear, but they’re connected with a mix of left- and right-brain skills that seems to be important; hackers need to be able to both reason logically and step outside the apparent logic of a problem at a moment’s notice.
Work as intensely as you play and play as intensely as you work. For true hackers, the boundaries between “play”, “work”, “science” and “art” all tend to disappear, or to merge into a high-level creative playfulness. Also, don’t be content with a narrow range of skills. Though most hackers self-describe as programmers, they are very likely to be more than competent in several related skills — system administration, web design, and PC hardware troubleshooting are common ones. A hacker who’s a system administrator, on the other hand, is likely to be quite skilled at script programming and web design. Hackers don’t do things by halves; if they invest in a skill at all, they tend to get very good at it.
Finally, a few things not to do.
- Don’t use a silly, grandiose user ID or screen name.
- Don’t get in flame wars on Usenet (or anywhere else).
- Don’t call yourself a ‘cyberpunk’, and don’t waste your time on anybody who does.
- Don’t post or email writing that’s full of spelling errors and bad grammar.
The only reputation you’ll make doing any of these things is as a twit. Hackers have long memories — it could take you years to live your early blunders down enough to be accepted.
The problem with screen names or handles deserves some amplification. Concealing your identity behind a handle is a juvenile and silly behavior characteristic of crackers, warez d00dz, and other lower life forms. Hackers don’t do this; they’re proud of what they do and want it associated with their real names. So if you have a handle, drop it. In the hacker culture it will only mark you as a loser.
Historical Note: Hacking, Open Source, and Free Software
When I originally wrote this how-to in late 1996, some of the conditions around it were very different from the way they look today. A few words about these changes may help clarify matters for people who are confused about the relationship of open source, free software, and Linux to the hacker community. If you are not curious about this, you can skip straight to the FAQ and bibliography from here.
The hacker ethos and community as I have described it here long predates the emergence of Linux after 1990; I first became involved with it around 1976, and, its roots are readily traceable back to the early 1960s. But before Linux, most hacking was done on either proprietary operating systems or a handful of quasi-experimental homegrown systems like MIT’s ITS that were never deployed outside of their original academic niches. While there had been some earlier (pre-Linux) attempts to change this situation, their impact was at best very marginal and confined to communities of dedicated true believers which were tiny minorities even within the hacker community, let alone with respect to the larger world of software in general.
What is now called “open source” goes back as far as the hacker community does, but until 1985 it was an unnamed folk practice rather than a conscious movement with theories and manifestos attached to it. This prehistory ended when, in 1985, arch-hacker Richard Stallman (“RMS”) tried to give it a name — “free software”. But his act of naming was also an act of claiming; he attached ideological baggage to the “free software” label which much of the existing hacker community never accepted. As a result, the “free software” label was loudly rejected by a substantial minority of the hacker community (especially among those associated with BSD Unix), and used with serious but silent reservations by a majority of the remainder (including myself).
Despite these reservations, RMS’s claim to define and lead the hacker community under the “free software” banner broadly held until the mid-1990s. It was seriously challenged only by the rise of Linux. Linux gave open-source development a natural home. Many projects issued under terms we would now call open-source migrated from proprietary Unixes to Linux. The community around Linux grew explosively, becoming far larger and more heterogenous than the pre-Linux hacker culture. RMS determinedly attempted to co-opt all this activity into his “free software” movement, but was thwarted by both the exploding diversity of the Linux community and the public skepticism of its founder, Linus Torvalds. Torvalds continued to use the term “free software” for lack of any alternative, but publicly rejected RMS’s ideological baggage. Many younger hackers followed suit.
In 1996, when I first published this Hacker HOWTO, the hacker community was rapidly reorganizing around Linux and a handful of other open-source operating systems (notably those descended from BSD Unix). Community memory of the fact that most of us had spent decades developing closed-source software on closed-source operating systems had not yet begun to fade, but that fact was already beginning to seem like part of a dead past; hackers were, increasingly, defining themselves as hackers by their attachments to open-source projects such as Linux or Apache.
The term “open source”, however, had not yet emerged; it would not do so until early 1998. When it did, most of the hacker community adopted it within the following six months; the exceptions were a minority ideologically attached to the term “free software”. Since 1998, and especially after about 2003, the identification of ‘hacking’ with ‘open-source (and free software) development’ has become extremely close. Today there is little point in attempting to distinguish between these categories, and it seems unlikely that will change in the future.
It is worth remembering, however, that this was not always so.
Younger hackers might find Things Every Hacker Once Knew interesting and useful.
I have also written A Brief History Of Hackerdom.
I have written a paper, The Cathedral and the Bazaar, which explains a lot about how the Linux and open-source cultures work. I have addressed this topic even more directly in its sequel Homesteading the Noosphere.
Rick Moen has written an excellent document on how to run a Linux user group.
Rick Moen and I have collaborated on another document on How To Ask Smart Questions. This will help you seek assistance in a way that makes it more likely that you will actually get it.
If you need instruction in the basics of how personal computers, Unix, and the Internet work, see The Unix and Internet Fundamentals HOWTO.
When you release software or write patches for software, try to follow the guidelines in the Software Release Practice HOWTO.
If you enjoyed the Zen poem, you might also like Rootless Root: The Unix Koans of Master Foo.
Frequently Asked Questions
Q: How do I tell if I am already a hacker?Q: Will you teach me how to hack?Q: How can I get started, then?Q: When do you have to start? Is it too late for me to learn?Q: How long will it take me to learn to hack?Q: Is Visual Basic a good language to start with?Q: Would you help me to crack a system, or teach me how to crack?Q: How can I get the password for someone else’s account?Q: How can I break into/read/monitor someone else’s email?Q: How can I steal channel op privileges on IRC?Q: I’ve been cracked. Will you help me fend off further attacks?Q: I’m having problems with my Windows software. Will you help me?Q: Where can I find some real hackers to talk with?Q: Can you recommend useful books about hacking-related subjects?Q: Do I need to be good at math to become a hacker?Q: What language should I learn first?Q: What kind of hardware do I need?Q: I want to contribute. Can you help me pick a problem to work on?Q: Do I need to hate and bash Microsoft?Q: But won’t open-source software leave programmers unable to make a living?Q: Where can I get a free Unix?
|Q:||How do I tell if I am already a hacker?|
|A:||Ask yourself the following three questions:Do you speak code, fluently?Do you identify with the goals and values of the hacker community?Has a well-established member of the hacker community ever called you a hacker?If you can answer yes to all three of these questions, you are already a hacker. No two alone are sufficient.The first test is about skills. You probably pass it if you have the minimum technical skills described earlier in this document. You blow right through it if you have had a substantial amount of code accepted by an open-source development project.The second test is about attitude. If the five principles of the hacker mindset seemed obvious to you, more like a description of the way you already live than anything novel, you are already halfway to passing it. That’s the inward half; the other, outward half is the degree to which you identify with the hacker community’s long-term projects.Here is an incomplete but indicative list of some of those projects: Does it matter to you that Linux improve and spread? Are you passionate about software freedom? Hostile to monopolies? Do you act on the belief that computers can be instruments of empowerment that make the world a richer and more humane place?But a note of caution is in order here. The hacker community has some specific, primarily defensive political interests — two of them are defending free-speech rights and fending off “intellectual-property” power grabs that would make open source illegal. Some of those long-term projects are civil-liberties organizations like the Electronic Frontier Foundation, and the outward attitude properly includes support of them. But beyond that, most hackers view attempts to systematize the hacker attitude into an explicit political program with suspicion; we’ve learned, the hard way, that these attempts are divisive and distracting. If someone tries to recruit you to march on your capitol in the name of the hacker attitude, they’ve missed the point. The right response is probably “Shut up and show them the code.”The third test has a tricky element of recursiveness about it. I observed in the section called “What Is a Hacker?” that being a hacker is partly a matter of belonging to a particular subculture or social network with a shared history, an inside and an outside. In the far past, hackers were a much less cohesive and self-aware group than they are today. But the importance of the social-network aspect has increased over the last thirty years as the Internet has made connections with the core of the hacker subculture easier to develop and maintain. One easy behavioral index of the change is that, in this century, we have our own T-shirts.Sociologists, who study networks like those of the hacker culture under the general rubric of “invisible colleges”, have noted that one characteristic of such networks is that they have gatekeepers — core members with the social authority to endorse new members into the network. Because the “invisible college” that is hacker culture is a loose and informal one, the role of gatekeeper is informal too. But one thing that all hackers understand in their bones is that not every hacker is a gatekeeper. Gatekeepers have to have a certain degree of seniority and accomplishment before they can bestow the title. How much is hard to quantify, but every hacker knows it when they see it.|
|Q:||Will you teach me how to hack?|
|A:||Since first publishing this page, I’ve gotten several requests a week (often several a day) from people to “teach me all about hacking”. Unfortunately, I don’t have the time or energy to do this; my own hacking projects, and working as an open-source advocate, take up 110% of my time.Even if I did, hacking is an attitude and skill you basically have to teach yourself. You’ll find that while real hackers want to help you, they won’t respect you if you beg to be spoon-fed everything they know.Learn a few things first. Show that you’re trying, that you’re capable of learning on your own. Then go to the hackers you meet with specific questions.If you do email a hacker asking for advice, here are two things to know up front. First, we’ve found that people who are lazy or careless in their writing are usually too lazy and careless in their thinking to make good hackers — so take care to spell correctly, and use good grammar and punctuation, otherwise you’ll probably be ignored. Secondly, don’t dare ask for a reply to an ISP account that’s different from the account you’re sending from; we find people who do that are usually thieves using stolen accounts, and we have no interest in rewarding or assisting thievery.|
|Q:||How can I get started, then?|
|A:||The best way for you to get started would probably be to go to a LUG (Linux user group) meeting. You can find such groups on the LDP General Linux Information Page; there is probably one near you, possibly associated with a college or university. LUG members will probably give you a Linux if you ask, and will certainly help you install one and get started.Your next step (and your first step if you can’t find a LUG nearby) should be to find an open-source project that interests you. Start reading code and reviewing bugs. Learn to contribute, and work your way in.The only way in is by working to improve your skills. If you ask me personally for advice on how to get started, I will tell you these exact same things, because I don’t have any magic shortcuts for you. I will also mentally write you off as a probable loser – because if you lacked the stamina to read this FAQ and the intelligence to understand from it that the only way in is by working to improve your skills, you’re hopeless.Another interesting possibility is to go visit a hackerspace. There is a burgeoning movement of people creating physical locations – maker’s clubs – where they can hang out to work on hardware and software projects together, or work solo in a cogenial atmosphere. Hackerspaces often collect tools and specialized equipment that would be too expensive or logistically inconvenient for individuals to own. Hackerspaces are easy to find on the Internet; one may be located near you.|
|Q:||When do you have to start? Is it too late for me to learn?|
|A:||Any age at which you are motivated to start is a good age. Most people seem to get interested between ages 15 and 20, but I know of exceptions in both directions.|
|Q:||How long will it take me to learn to hack?|
|A:||That depends on how talented you are and how hard you work at it. Most people who try can acquire a respectable skill set in eighteen months to two years, if they concentrate. Don’t think it ends there, though; in hacking (as in many other fields) it takes about ten years to achieve mastery. And if you are a real hacker, you will spend the rest of your life learning and perfecting your craft.|
|Q:||Is Visual Basic a good language to start with?|
|A:||If you’re asking this question, it almost certainly means you’re thinking about trying to hack under Microsoft Windows. This is a bad idea in itself. When I compared trying to learn to hack under Windows to trying to learn to dance while wearing a body cast, I wasn’t kidding. Don’t go there. It’s ugly, and it never stops being ugly.There is a specific problem with Visual Basic; mainly that it’s not portable. Though there is a prototype open-source implementations of Visual Basic, the applicable ECMA standards don’t cover more than a small set of its programming interfaces. On Windows most of its library support is proprietary to a single vendor (Microsoft); if you aren’t extremely careful about which features you use — more careful than any newbie is really capable of being — you’ll end up locked into only those platforms Microsoft chooses to support. If you’re starting on a Unix, much better languages with better libraries are available. Python, for example.Also, like other Basics, Visual Basic is a poorly-designed language that will teach you bad programming habits. No, don’t ask me to describe them in detail; that explanation would fill a book. Learn a well-designed language instead.One of those bad habits is becoming dependent on a single vendor’s libraries, widgets, and development tools. In general, any language that isn’t fully supported under at least Linux or one of the BSDs, and/or at least three different vendors’ operating systems, is a poor one to learn to hack in.|
|Q:||Would you help me to crack a system, or teach me how to crack?|
|A:||No. Anyone who can still ask such a question after reading this FAQ is too stupid to be educable even if I had the time for tutoring. Any emailed requests of this kind that I get will be ignored or answered with extreme rudeness.|
|Q:||How can I get the password for someone else’s account?|
|A:||This is cracking. Go away, idiot.|
|Q:||How can I break into/read/monitor someone else’s email?|
|A:||This is cracking. Get lost, moron.|
|Q:||How can I steal channel op privileges on IRC?|
|A:||This is cracking. Begone, cretin.|
|Q:||I’ve been cracked. Will you help me fend off further attacks?|
|A:||No. Every time I’ve been asked this question so far, it’s been from some poor sap running Microsoft Windows. It is not possible to effectively secure Windows systems against crack attacks; the code and architecture simply have too many flaws, which makes securing Windows like trying to bail out a boat with a sieve. The only reliable prevention starts with switching to Linux or some other operating system that is designed to at least be capable of security.|
|Q:||I’m having problems with my Windows software. Will you help me?|
|A:||Yes. Go to a DOS prompt and type “format c:”. Any problems you are experiencing will cease within a few minutes.|
|Q:||Where can I find some real hackers to talk with?|
|A:||The best way is to find a Unix or Linux user’s group local to you and go to their meetings (you can find links to several lists of user groups on the LDP site at ibiblio).(I used to say here that you wouldn’t find any real hackers on IRC, but I’m given to understand this is changing. Apparently some real hacker communities, attached to things like GIMP and Perl, have IRC channels now.)|
|Q:||Can you recommend useful books about hacking-related subjects?|
|A:||I maintain a Linux Reading List HOWTO that you may find helpful. The Loginataka may also be interesting.For an introduction to Python, see the tutorial on the Python site.|
|Q:||Do I need to be good at math to become a hacker?|
|A:||No. Hacking uses very little formal mathematics or arithmetic. In particular, you won’t usually need trigonometry, calculus or analysis (there are exceptions to this in a handful of specific application areas like 3-D computer graphics). Knowing some formal logic and Boolean algebra is good. Some grounding in finite mathematics (including finite-set theory, combinatorics, and graph theory) can be helpful.Much more importantly: you need to be able to think logically and follow chains of exact reasoning, the way mathematicians do. While the content of most mathematics won’t help you, you will need the discipline and intelligence to handle mathematics. If you lack the intelligence, there is little hope for you as a hacker; if you lack the discipline, you’d better grow it.I think a good way to find out if you have what it takes is to pick up a copy of Raymond Smullyan’s book What Is The Name Of This Book?. Smullyan’s playful logical conundrums are very much in the hacker spirit. Being able to solve them is a good sign; enjoying solving them is an even better one.|
|Q:||What language should I learn first?|
|A:||HTML if you don’t already know it. There are a lot of glossy, hype-intensive bad HTML books out there, and distressingly few good ones. The one I like best is HTML: The Definitive Guide.But HTML is not a full programming language. When you’re ready to start programming, I would recommend starting with Python. You will hear a lot of people recommending Perl, but it’s harder to learn and (in my opinion) less well designed.C is really important, but it’s also much more difficult than either Python or Perl. Don’t try to learn it first.Windows users, do not settle for Visual Basic. It will teach you bad habits, and it’s not portable off Windows. Avoid.|
|Q:||What kind of hardware do I need?|
|A:||It used to be that personal computers were rather underpowered and memory-poor, enough so that they placed artificial limits on a hacker’s learning process. This stopped being true in the mid-1990s; any machine from an Intel 486DX50 up is more than powerful enough for development work, X, and Internet communications, and the smallest disks you can buy today are plenty big enough.The important thing in choosing a machine on which to learn is whether its hardware is Linux-compatible (or BSD-compatible, should you choose to go that route). Again, this will be true for almost all modern machines. The only really sticky areas are modems and wireless cards; some machines have Windows-specific hardware that won’t work with Linux.There’s a FAQ on hardware compatibility; the latest version is here.|
|Q:||I want to contribute. Can you help me pick a problem to work on?|
|A:||No, because I don’t know your talents or interests. You have to be self-motivated or you won’t stick, which is why having other people choose your direction almost never works.|
|Q:||Do I need to hate and bash Microsoft?|
|A:||No, you don’t. Not that Microsoft isn’t loathsome, but there was a hacker culture long before Microsoft and there will still be one long after Microsoft is history. Any energy you spend hating Microsoft would be better spent on loving your craft. Write good code — that will bash Microsoft quite sufficiently without polluting your karma.|
|Q:||But won’t open-source software leave programmers unable to make a living?|
|A:||This seems unlikely — so far, the open-source software industry seems to be creating jobs rather than taking them away. If having a program written is a net economic gain over not having it written, a programmer will get paid whether or not the program is going to be open-source after it’s done. And, no matter how much “free” software gets written, there always seems to be more demand for new and customized applications. I’ve written more about this at the Open Source pages.|
|Q:||Where can I get a free Unix?|
|A:||If you don’t have a Unix installed on your machine yet, elsewhere on this page I include pointers to where to get the most commonly used free Unix. To be a hacker you need motivation and initiative and the ability to educate yourself. Start now…|
Lược sử Deep Learning
Như đã một lần nhắc đến trong bài đầu tiên của blog, trí tuệ nhân tạo đang len lỏi vào trong cuộc sống và ảnh hưởng sâu rộng tới mỗi chúng ta. Kể từ khi tôi viết bài đầu tiên, tần suất chúng ta nghe thấy các cụm từ ‘artificial intelligence’, ‘machine learning’, ‘deep learning’ cũng ngày một tăng lên. Nguyên nhân chính dẫn đến việc này (và việc ra đời blog này) là sự xuất hiện của deep learning trong 5-6 năm gần đây.
Một lần nữa xin được dùng lại hình vẽ mô tả mối quan hệ giữa artificial intelligence, machine learning, và deep learning:
Mối quan hệ giữa AI, Machine Learning và Deep Learning.
(Nguồn: What’s the Difference Between Artificial Intelligence, Machine Learning, and Deep Learning?)
Trong bài viết này, tôi sẽ trình bày sơ lược về lịch sử deep learning. Trong các bài tiếp theo, tôi có tham vọng viết thật kỹ về các thành phần cơ bản của các hệ thống deep learning. Xa hơn nữa, blog sẽ có thêm các bài hướng dẫn cho nhiều bài toán thực tế.
Blog luôn đón nhận những đóng góp để chất lượng các bài viết được tốt hơn. Nếu bạn có đóng góp nào, vui lòng để lại trong phần comment, tôi sẽ cập nhật bài viết cho phù hợp. Cảm ơn bạn.
Những dấu mốc quan trọng của deep learning
Deep learning được nhắc đến nhiều trong những năm gần đây, nhưng những nền tảng cơ bản đã xuất hiện từ rất lâu …
Chúng ta cùng quan sát hình dưới đây:
Lịch sử deep learning (Nguồn: Hình được lấy từ Deep Learning 101 – Part 1: History and Background . Tác giả bài viết không biết chính xác nguồn gốc của hình.)
Một trong những nền móng đầu tiên của neural network và deep learning là perceptron learning algorithm (hoặc gọn là perceptron). Perceptron là một thuật toán supervised learning giúp giải quyết bài toán phân lớp nhị phân, được khởi nguồn bởi Frank Rosenblatt năm 1957 trong một nghiên cứu được tài trợ bởi Văn phòng nghiên cứu hải quân Hoa Kỳ (U.S Office of Naval Research – từ một cơ quan liên quan đến quân sự). Thuật toán perceptron được chứng minh là hội tụ nếu hai lớp dữ liệu là linearly separable. Với thành công này, năm 1958, trong một hội thảo, Rosenblatt đã có một phát biểu gây tranh cãi. Từ phát biểu này, tờ New York Times đã có một bài báo cho rằng perceptron được Hải quân Hoa Kỳ mong đợi “có thể đi, nói chuyện, nhìn, viết, tự sinh sản, và tự nhận thức được sự tồn tại của mình”. (Chúng ta biết rằng cho tới giờ các hệ thống nâng cao hơn perceptron nhiều lần vẫn chưa thể).
Mặc dù thuật toán này mang lại nhiều kỳ vọng, nó nhanh chóng được chứng minh không thể giải quyết những bài toán đơn giản. Năm 1969, Marvin Minsky và Seymour Papert trong cuốn sách nổi tiếng Perceptrons đã chứng minh rằng không thể ‘học’ được hàm số XOR khi sử dụng perceptron. Phát hiện này làm choáng váng giới khoa học thời gian đó (bây giờ chúng ta thấy việc này khá hiển nhiên). Perceptron được chứng minh rằng chỉ hoạt động nếu dữ liệu là linearly separable.
Phát hiện này khiến cho các nghiên cứu về perceptron bị gián đoạn gần 20 năm. Thời kỳ này còn được gọi là Mùa đông AI thứ nhất (The First AI winter).
Cho tới khi…
MLP và Backpropagation ra đời (80s)
Geoffrey Hinton tốt nghiệp PhD ngành neural networks năm 1978. Năm 1986, ông cùng với hai tác giả khác xuất bản một bài báo khoa học trên Nature với tựa đề “Learning representations by back-propagating errors”. Trong bài báo này, nhóm của ông chứng minh rằng neural nets với nhiều hidden layer (được gọi là multi-layer perceptron hoặc MLP) có thể được huấn luyện một cách hiệu quả dựa trên một quy trình đơn giản được gọi là backpropagation (backpropagation là tên gọi mỹ miều của quy tắc chuỗi – chain rule – trong tính đạo hàm. Việc tính được đạo hàm của hàm số phức tạp mô tả quan hệ giữa đầu vào và đầu ra của một neural net là rất quan trọng vì hầu hết các thuật toán tối ưu đều được thực hiện thông qua việc tính đạo hàm, gradient descent là một ví dụ). Việc này giúp neural nets thoát được những hạn chế của perceptron về việc chỉ biểu diễn được các quan hệ tuyến tính. Để biểu diễn các quan hệ phi tuyến, phía sau mỗi layer là một hàm kích hoạt phi tuyến, ví dụ hàm sigmoid hoặc tanh. (ReLU ra đời năm 2012). Với hidden layers, neural nets được chứng minh rằng có khả năng xấp xỉ hầu hết bất kỳ hàm số nào qua một định lý được gọi là universal approximation theorem. Neurel nets quay trở lại cuộc chơi.
Thuật toán này mang lại một vài thành công ban đầu, nổi trội là convolutional neural nets (convnets hay CNN) (còn được gọi là LeNet) cho bài toán nhận dạng chữ số viết tay được khởi nguồn bởi Yann LeCun tại AT&T Bell Labs (Yann LeCun là sinh viên sau cao học của Hinton tại đại học Toronto năm 1987-1988). Dưới đây là bản demo được lấy từ trang web của LeNet, network là một CNN với 5 layer, còn được gọi là LeNet-5 (1998).
LeNet-5 cho bài toán nhận diện chữ số viết tay. (Nguồn: http://yann.lecun.com)
Mô hình này được sử dụng rộng rãi trong các hệ thống đọc số viết tay trên các check (séc ngân hàng) và mã vùng bưu điện của nước Mỹ.
LeNet là thuật toán tốt nhất thời gian đó cho bài toán nhận dạng ảnh chữ số viết tay. Nó tốt hơn MLP thông thường (với fully connected layer) vì nó có khả năng trích xuất được đặc trưng trong không gian hai chiều của ảnh thông qua các filters (bộ lọc) hai chiều. Hơn nữa, các filter này nhỏ nên việc lưu trữ và tính toán cũng tốt hơn so với MLP thông thường. (Yan LeCun có xuất phát từ Electrical Engineering nên rất quen thuộc với các bộ lọc.)
Mùa đông AI thứ hai (90s – đầu 2000s)
Các mô hình tương tự được kỳ vọng sẽ giải quyết nhiều bài toán image classification khác. Tuy nhiên, không như các chữ số, các loại ảnh khác lại rất hạn chế vì máy ảnh số chưa phổ biến tại thời điểm đó. Ảnh được gán nhãn lại càng hiếm. Trong khi để có thể huấn luyện được mô hình convnets, ta cần rất nhiều dữ liệu huấn luyện. Ngay cả khi dữ liệu có đủ, một vấn đề nan giải khác là khả năng tính toán của các máy tính thời đó còn rất hạn chế.
Một hạn chế khác của các kiến trúc MLP nói chung là hàm mất mát không phải là một hàm lồi. Việc này khiến cho việc tìm nghiệm tối ưu toàn cục cho bài toán tối ưu hàm mất mát trở nên rất khó khăn. Một vấn đề khác liên quan đến giới hạn tính toán của máy tính cũng khiến cho việc huấn luyện MLP không hiệu quả khi số lượng hidden layers lớn lên. Vấn đề này có tên là vanishing gradient.
Nhắc lại rằng hàm kích hoạt được sử dụng thời gian đó là sigmoid hoặc tanh – là các hàm bị chặn trong khoảng (0, 1) hoặc (-1, 1) (Nhắc lại đạo hàm của hàm sigmoid σ(z)σ(z) là σ(z)(1−σ(z))σ(z)(1−σ(z)) là tích của hai số nhỏ hơn 1). Khi sử dụng backpropagation để tính đạo hàm cho các ma trận hệ số ở các lớp đầu tiên, ta cần phải nhân rất nhiều các giá trị nhỏ hơn 1 với nhau. Việc này khiến cho nhiều đạo hàm thành phần bằng 0 do xấp xỉ tính toán. Khi đạo hàm của một thành phần bằng 0, nó sẽ không được cập nhật thông qua gradient descent!
Những hạn chế này khiến cho neural nets một lần nữa rơi vào thời kỳ băng giá. Vào thời điểm những năm 1990 và đầu những năm 2000, neural nets dần được thay thế bởi support vector machines –SVM. SVMs có ưu điểm là bài toán tối ưu để tìm các tham số của nó là một bài toán lồi – có nhiều các thuật toán tối ưu hiệu quả giúp tìm nghiệm của nó. Các kỹ thuật về kernel cũng phát triển giúp SVMs giải quyết được cả các vấn đề về việc dữ liệu không phân biệt tuyến tính.
Nhiều nhà khoa học làm machine learning chuyển sang nghiên cứu SVM trong thời gian đó, trừ một vài nhà khoa học cứng đầu…
Cái tên được làm mới – Deep Learning (2006)
Năm 2006, Hinton một lần nữa cho rằng ông biết bộ não hoạt động như thế nào, và giới thiệu ý tưởng của tiền huấn luyện không giám sát (unsupervised pretraining) thông qua deep belief nets (DBN). DBN có thể được xem như sự xếp chồng các unsupervised networks đơn giản như restricted Boltzman machine hay autoencoders.
Lấy ví dụ với autoencoder. Mỗi autoencoder là một neural net với một hidden layer. Số hidden unit ít hơn số input unit, và số output unit bằng với số input unit. Network này đơn giản được huấn luyện để kết quả ở output layer giống với kết quả ở input layer (và vì vậy được gọi là autoencoder). Quá trình dữ liệu đi từ input layer tới hidden layer có thể coi là mã hoá, quá trình dữ liệu đi từ hidden layer ra output layer có thể được coi là giải mã. Khi output giống với input, ta có thể thấy rằng hidden layer với ít unit hơn có để mã hoá input khá thành công, và có thể được coi mang những tính chất của input. Nếu ta bỏ output layer, cố định (freeze) kết nối giữa input và hidden layer, coi đầu ra của hidden layer là một input mới, sau đó huấn luyện một autoencoder khác, ta được thêm một hidden layer nữa. Quá trình này tiếp tục kéo dài ta sẽ được một network đủ sâu mà output của network lớn này (chính là hidden layer của autoencoder cuối cùng) mang thông tin của input ban đầu. Sau đó ta có thể thêm các layer khác tuỳ thuộc vào bài toán (chẳng hạn thêm softmax layer ở cuối cho bài toán classification). Cả network được huấn luyện thêm một vài epoch nữa. Quá trình này được gọi là tinh chỉnh (fine tuining).
Tại sao quá trình huấn luyện như trên mang lại nhiều lợi ích?
Một trong những hạn chế đã đề cập của MLP là vấn đề vanishing gradient. Những ma trận trọng số ứng với các layer đầu của network rất khó được huấn luyện vì đạo hàm của hàm mất mát theo các ma trận này nhỏ. Với ý tưởng của DBN, các ma trận trọng số ở những hidden layer đầu tiên được tiền huấn luyện (pretrained). Các trọng số được tiền huấn luyện này có thể coi là giá trị khởi tạo tốt cho các hidden layer phía đầu. Việc này giúp phần nào tránh được sự phiền hà của vanishing gradient.
Kể từ đây, neural networks với nhiều hidden layer được đổi tên thành deep learning.
Vấn đề vanishing gradient được giải quyết phần nào (vẫn chưa thực sự triệt để), nhưng vẫn còn những vấn đề khác của deep learning: dữ liệu huấn luyện quá ít, và khả năng tính toán của CPU còn rất hạn chế trong việc huấn luyện các deep networks.
Năm 2010, giáo sư Fei-Fei Li, một giáo sư ngành computer vision đầu ngành tại Stanford, cùng với nhóm của bà tạo ra một cơ sở dữ liệu có tên ImageNet với hàng triệu bức ảnh thuộc 1000 lớp dữ liệu khác nhau đã được gán nhãn. Dự án này được thực hiện nhờ vào sự bùng nổ của internet những năm 2000 và lượng ảnh khổng lồ được upload lên internet thời gian đó. Các bức ảnh này được gán nhãn bởi rất nhiều người (được trả công).
Bộ cơ sở dữ liệu này được cập nhật hàng năm, và kể từ năm 2010, nó được dùng trong một cuộc thi thường niên có tên ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Trong cuộc thi này, dữ liệu huấn luyện được giao cho các đội tham gia. Mỗi đội cần sử dụng dữ liệu này để huấn luyện các mô hình phân lớp, các mô hình này sẽ được áp dụng để dự đoán nhãn của dữ liệu mới (được giữ bởi ban tổ chức). Trong hai năm 2010 và 2011, có rất nhiều đội tham gia. Các mô hình trong hai năm này chủ yếu là sự kết hợp của SVM với các feature được xây dựng bởi các bộ hand-crafted descriptors (SIFT, HoG, v.v.). Mô hình giành chiến thắng có top-5 error rate là 28% (càng nhỏ càng tốt). Mô hình giành chiến thắng năm 2011 có top-5 error rate là 26%. Cải thiện không nhiều!
Ngoài lề: top-5 error rate được tính như sau. Mỗi mô hình dự đoán 5 nhãn của một bức ảnh. Nếu nhãn thật của bức ảnh nằm trong 5 nhãn đó, ta có một điểm được phân lớp chính xác. Ngoài ra, bức ảnh đó được coi là một error. Top-5 error rate là tỉ lệ số bức ảnh error trong toàn bộ số ảnh kiểm thử với error được tính theo cách này. Top-1 error cộng với classification accuracy (phần trăm) chính bằng 100 phần trăm.
Đột phá (2012)
Năm 2012, cũng tại ILSVRC, Alex Krizhevsky, Ilya Sutskever, và Geoffrey Hinton (lại là ông) tham gia và đạt kết quả top-5 error rate 16%. Kết quả này làm sững sờ giới nghiên cứu thời gian đó. Mô hình là một Deep Convolutional Neural Network, sau này được gọi là AlexNet.
Trong bài báo này, rất nhiều các kỹ thuật mới được giới thiệu. Trong đó hai đóng góp nổi bật nhất là hàm ReLU và dropout. Hàm ReLU (ReLU(x)=max(x,0)ReLU(x)=max(x,0)) với cách tính và đạo hàm đơn giản (bằng 1 khi đầu vào không âm, bằng 0 khi ngược lại) giúp tốc độ huấn luyện tăng lên đáng kể. Ngoài ra, việc ReLU không bị chặn trên bởi 1 (như softmax hay tanh) khiến cho vấn đề vanishing gradient cũng được giải quyết phần nào. Dropout cũng là một kỹ thuật đơn giản và cực kỳ hiệu quả. Trong quá trình training, nhiều hidden unit bị tắt ngẫu nhiên và mô hình được huấn luyện trên các bộ tham số còn lại. Trong quá trình test, toàn bộ các unit sẽ được sử dụng. Cách làm này khá là có lý khi đối chiếu với con người. Nếu chỉ dùng một phần năng lực đã đem lại hiệu quả thì dùng toàn bộ năng lực sẽ mang lại hiệu quả cao hơn. Việc này cũng giúp cho mô hình tránh được overfitting và cũng được coi giống với kỹ thuật ensemble trong các hệ thống machine learning khác. Với mỗi cách tắt các unit, ta có một mô hình khác nhau. Với nhiều tổ hợp unit bị tắt khác nhau, ta thu được nhiều mô hình. Việc kết hợp ở cuối cùng được coi như sự kết hợp của nhiều mô hình (và vì vậy, nó giống với ensemble learning).
Một trong những yếu tố quan trọng nhất giúp AlexNet thành công là việc sử dụng GPU (card đồ hoạ) để huấn luyện mô hình. GPU được tạo ra cho game thủ, với khả năng chạy song song nhiều lõi, đã trở thành một công cụ cực kỳ phù hợp với các thuật toán deep learning, giúp tăng tốc thuật toán lên nhiều lần so với CPU.
Sau AlexNet, tất cả các mô hình giành giải cao trong các năm tiếp theo đều là các deep networks (ZFNet 2013, GoogLeNet 2014, VGG 2014, ResNet 2015). Tôi sẽ giành một bài của blog để viết về các kiến trúc quan trọng này. Xu thế chung có thể thấy là các mô hình càng ngày càng deep. Xem hình dưới đây.
Kết quả ILSVRC qua các năm. (Nguồn: CNN Architectures: LeNet, AlexNet, VGG, GoogLeNet, ResNet and more …)
Những công ty công nghệ lớn cũng để ý tới việc phát triển các phòng nghiên cứu deep learning trong thời gian này. Rất nhiều các ứng dụng công nghệ đột phá đã được áp dụng vào cuộc sống hàng ngày. Cũng kể từ năm 2012, số lượng các bài báo khoa học về deep learning tăng lên theo hàm số mũ. Các blog về deep learning cũng tăng lên từng ngày.
Điều gì mang đến sự thành công của deep learning?
Rất nhiều những ý tưởng cơ bản của deep learning được đặt nền móng từ những năm 80-90 của thế kỷ trước, tuy nhiên deep learning chỉ đột phá trong khoảng 5-6 năm nay. Vì sao?
Có nhiều nhân tố dẫn đến sự bùng nổ này:
- Sự ra đời của các bộ dữ liệu lớn được gán nhãn.
- Khả năng tính toán song song tốc độ cao của GPU.
- Sự ra đời của ReLU và các hàm kích hoạt liên quan làm hạn chế vấn đề vanishing gradient.
- Sự cải tiến của các kiến trúc: GoogLeNet, VGG, ResNet, … và các kỹ thuật transfer learning, fine tuning.
- Nhiều kỹ thuật regularization mới: dropout, batch normalization, data augmentation.
- Nhiều thư viện mới hỗ trợ việc huấn luyện deep network với GPU: theano, caffe, mxnet, tensorflow, pytorch, keras, …
- Nhiều kỹ thuật tối ưu mới: Adagrad, RMSProp, Adam, …
Rất nhiều bạn đọc có yêu cầu tôi viết về deep learning từ lâu. Tuy nhiên, trước đó tôi tự nhận rằng mình chưa đủ kiến thức về lĩnh vực này để viết cho độc giả. Chỉ khi có những bài cơ bản về machine learning và bản thân đã tích luỹ được một lượng kiến thức nhất định tôi mới quyết định bắt đầu vào chủ đề được nhiều bạn quan tâm này.
Các thuật toán machine learning cổ điển khác vẫn có thể xuất hiện trong các bài sau của blog.
TechTalk via Machinelearningcoban
The US’s top 15 emerging jobs of 2020, according to LinkedIn
Too often the world of robotics seems to be a solution in search of a problem. Assistive robotics, on the other hand, are among one of the primary real-world tasks existing technology can seemingly address almost immediately. The concept for the technology has been around for some time now and has caught on particularly well…Diligent’s Vivian Chu and Labrador’s Mike Dooley will discuss assistive robotics at TC Sessions: Robotics+AI — TechCrunch