HN Explorer

Discuss on HN: The State of Artifical Intelligence

8 points by davidkatz ↗ HN
Smart machines have been a promise of Computer Science for a while now. These days though I often find myself lost around what has actually been achieved.

Which problems have been solved, where has meaningful progress been made, and what do we desire but still seems far off?

Would someone care to start with an overview?

11 comments

[ 2.9 ms ] story [ 31.1 ms ] thread
zuckmitnick
AI is not an explicit subject like math and physics, which has solid principles. It's just about searching for a way of providing the machines an ability to "think", whatever the way you choose.

Different people understand the word "think" differently. Eg, for most people, AI is about algorithms, while for me, it is about recreating a human brain.

So, an overview is too complicated and multifarious, at least for me.

davidkatz
I hear you. Perhaps not an overview then, but an example of a solved or unsolved problem?
zuckmitnick
Well, most AI techs that exists right now are built upon rules. A very early example may be the chess machine "deep blue", who defeated the world champion Kasparov. A recent one could be Google's virtual brain, who "understood" the concept "cat" by analysing 10 million random frames from YouTube. Another example may be the one S4M mentioned, I'm not familiar with that.

While Google's virtual brain "understood" something, it still doesn't has the ability of analysing. This is an example of unsolved problems.

davidkatz
What exactly do you mean by the ability to 'analyse'. Could you expand on that a little?
zuckmitnick
This means something like 'reasoning'.

Eg, you ask the virtual brain 'why is the sky blue?' 'why is the sky sometimes grey?', and it may not tell you why.

Google's virtual brain can 'recognize' things from the images, but it cannot handle the logic problems or something like that.

S4M
I recall a discussion on HN [1] were a worm had been simulated, which might be a step forward to the long road of recreating a human brain as zuckmitnick mentions.

[1] http://news.ycombinator.com/item?id=4208454

zuckmitnick
Yeah, that is surely a step forward. But simulating a worm is way too far from simulating a human brain. And also I'm not sure if this is a right path (Just IMHO).

Luckily, I've been working on the framework of simulating the human brain and I've got a theory. In the next few weeks I'll be working on realizing this framework and release a demo. If this works, I will post it on HN. :)

S4M
Looks good, good luck with that! That seems a really interesting and challenging project, do you have a background in neuropsychology or another field that helps you understand how the human brain work? If you need help, you can shoot me an email, and I will see what I can do (my background is more in applied maths).
zuckmitnick
This is really challenging. I think my theory is completely new, and I'm not sure if it could work, so for about 4 years I've been working on this alone. If I can prove its correctness by creating my first workable demo, I will start my startup and I will need many people's help. I'd be very happy if guys like you can join me! Cuz my math is poor. LOL.

My major is software engineering. My background is not about neuropsychology but I have very much interested in it. I did it mainly in a psychological way and on a neuropsychologic base.

Sorry to reply you so late. I've been busy these few days and didn't get time for HN.

S4M
Well, if you want a bit of help, you can shoot me an email - in my profile - and I'll see what I can do (which, I am afraid, will not be much).
zuckmitnick
Yeah, that's nice. I will, when I need your help. Thx