chetan51

I believe the basic building blocks (prediction, pattern recognition, attention, etc.) give rise to the higher-level phenomenon that you mentioned. I think we need to first understand those fundamental principles, and we'll be able to infer most of the rest from that point.

Emergent phenomenon can seem complicated and impossible to understand, but the mechanisms that give rise to them are usually simple (for example, evolution creating diverse and intricate life).

"Why would we continue to file patents on work that is going to be open source? The principle reason is to protect the NuPIC community. For example, outside developers could work on similar concepts without becoming part of the open source community. They could seek patents on their own work, making it proprietary and blocking progress of open source NuPIC developers. By keeping our patent portfolio current, we retain the ability to protect the NuPIC community from these threats. In other words, by holding patents on the work, we are able to protect the whole community from others who might seek to wall off their work through patents. In addition to filing select patents going forward, we also will evaluate other measures that would enhance patent protection for the NuPIC community."

Since this is a long-term project, it's more important that Numenta is able to protect the community it is building from patent trolls, and this is one approach to doing that.

How would you comment on a "Ask HN" submission asking HN how pg would fill out a YC application with YC as his idea?

The most concrete version of my point is that I don't think the most powerful AI we'll create will have, for instance, a human-coded algorithm for detecting faces. Instead, it'll have the ability to read electrical signals from a camera and understand the changing patterns in them, including the presence of faces. This ability to understand changing patterns would be due to "simpler" rules than the rules specifically designed to understand faces.

So yes, a general purpose learning algorithm, using the correct paradigm, would learn to think in a way as powerful as we do. And it'll do so in a way that its programmers would never be able to predict.

In the same vein, I would say that Google search results is an emerging phenomena, albeit not quite as interesting as general purpose intelligence. This is because it's intractable to predict what Google will return for certain queries, even if we know all of its rules. Keep in mind that there are degrees of emergence, it's not black and white. (On the other hand, I don't think Google's algorithm is as "simple" as it originally was, but that's for another discussion.)

Actually, there's a growing amount of evidence that there's a single, general-purpose algorithm in the human brain that gives rise to intelligence. For one, there's the fact that every part of the brain looks and behaves the same. There's also the fact that the brain is very plastic in what it learns – the auditory cortex can learn to "see" if we were to rewire the signals from the eyes from the visual cortex to the auditory cortex. It's very unlikely that our brain is hard-wired to recognize faces, for instance, but rather that it learns to do so using this generic learning algorithm.

I urge you to watch Andrew Ng's talk that I linked to in the post, and read On Intelligence (http://www.amazon.com/On-Intelligence-Jeff-Hawkins/dp/080507...) by Jeff Hawkins, a book that totally changed the way I look at intelligent behavior.

If we keep writing down common sense datums until 2100, we can make computers as smart as people.

We learn more about brains and making smart computers, but we seem to have run out of major architectural innovations -- better ones won't make a huge difference. The big stumbling block seems to be how much "common sense" a system knows, like that things tend to fall down when you bump them. One group has been writing these down for fifteen years with moderate success; a century more effort may be plenty.

This seems like a really inefficient way to go about it.

Quite a simple yet innovative idea. Now if it has some essential features like tabs for quick switching between different current directories, command history, and a beautiful interface, it could actually replace my default terminal app. Great start though!

I used this to make a screensaver out of it: https://github.com/liquidx/webviewscreensaver

Ooh, I would love this in screensaver form!

The part I found most useful:

Recently, I have been practicing a new time management technique and I am calling it the “Minimum Viable Daily Tasks”. Here is how it works: Choose the minimum amount of work that needs to be done everyday so that you have a big smile on your face when you are heading home in the evening. If you accomplish anything more then it is a bonus.

Demo demo demo!

I recently got hired through Developer Auction, and I just want to say that it was a really great experience. +1 from me :)

Well hopefully with enough data, it can automatically know that you don't care about your grandma's sushi preferences, and deemphasize that information when you search.

Try it in Incognito?

Well for one, it seems like a great way to browse Facebook content using natural language, rather that mouse clicks and page surfing. I like the idea of saying "give me pictures that I like taken in my hometown with my brother", rather than clicking through a bunch of filters.

I for one welcome any new natural language interface that can replace and improve on click interactions.

I actually thought the headline was a joke or a misrepresentation. Wow.

That's great to hear! I would suggest you consider those people who are trying to portray a public profile of links that reflect their interests, so a good profile view would reflect that intention :)

I'm going to use Kippt as a way to collect all the highest-quality links I've come across, with good descriptions on all of them. That collection will then be a public profile of links that I love that represent me.

The only thing I find lacking is the public profile view. It doesn't show all links on one page, but rather only shows 3 per section, and then you have to click links to see the whole section. It also doesn't show descriptions on that page for the links.

Fortunately, they have an API so you can make your own beautiful page to show all that, which I might do and embed onto my blog.

After doing a lot of web development, this is how I'm starting to feel as well. That's why I'm shifting my focus to AI and deep learning; it's a field with a lot more real computer science problems that you actually get to think about and solve on your own.

Actually, the last great human invention will probably be self-reflective Artificial Intelligence.

This. Both articles seemed extreme to me, and then again, only the extreme will really make it to the front page. If you rewrote this comment in a blog post and titled it "The Middle-Point", it wouldn't make front page, because it's just plain obvious.

Buy what fits at the right point in the spectrum for you, and move on with your life.

Do it!

I find that I rarely ever use anybody else's computer on which I cannot use Colemak. And on mobile devices, touchscreen QWERTY is just fine, since it doesn't suffer from the same inefficiencies that the keyboard QWERTY suffers from.

Bottom line: I get to use Colemak all the time that I want, and I never have a problem with mobile devices. Win-win!

Also, Colemak is more supported and available than any of those other layouts.

Colemak is a tradeoff between ultimate optimality, and shortcut / punctuation positioning for those switching from QWERTY. I agree, if you're learning a layout from scratch (if you're a young child who can't type yet, for example), then learn the best you can get. But if you're switching from QWERTY, the compatibility Colemak offers is indispensable (see the "Other layouts" section of the post).

Thanks for the great tips! I'll copy them over and quote you in the comments on the post.

Maybe it's the case that YMMV. I didn't touch-type QWERTY, so I learned Colemak just fine. On the other hand, my friend who touch-typed QWERTY at 90 WPM also had no problem switching to Colemak. In addition, the official Colemak website states:

"Easy to learn – Allows easy transition from QWERTY. Only 2 keys move between hands. Many common shortcuts (including Ctrl+Z/X/C/V) remain the same. Typing lessons available."

Please see the post on Time Bending (http://chetansurpur.com/blog/2012/10/time-bending.html) for more on the productivity technique that this app is based on.

Looks great!

An alternative interface for navigating links with your keyboard in Chrome is DeadMouse: http://chetan51.github.com/deadmouse/

With it, you can simply start typing out the text of any link you see on the page to go to it.

Thanks so much for this encouraging post.