Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

John Carmack: I’m going to work on artificial general intelligence

facebook.com

Post reply on HN
Print view View original

531–540 of 930 posts

Re: John Carmack: I’m going to work on artificial general intelligence

#531

Post by randomsearch »

Greed wrote: post #512
randomsearch wrote: post #491

Earlier quoted context omitted.

The VR vs AI comparison is interesting to me, because I think both technologies have come in “waves”. However, I think this is the last VR wave - it’s going to be on a steady gradient to ubiquity now - whilst I believe AI will winter again and there are many more waves to come, and decades (centuries?) to pass before AGI. Reasoning being: VR is just making what we have better. Better screens, better refresh, better b…

Out of curiosity, why do you see VR as being a necessary step towards the creation of AGI? Those two don't seem related at all in any way that I can discern.

Maybe “necessary” is too strong, but “likely pivotal” is better.

If VR becomes widespread, and amazingly high quality, then almost everything we do will migrate to VR.

Once that is the case, we will have an unprecedented amount of data about human behaviour, and near endless data for training, experimenting, and testing AIs.

The problems of AI will become much easier to formulate: “replace this person in this VR scenario, interaction” etc. This will help drive research by giving clear goals.

More pragmatically, it just removes a lot of barriers to research and accidental difficulties ie you’ll just be able to fire up a VR rather than worrying about how your robot is going to pick things up or access real world data etc

Re: John Carmack: I’m going to work on artificial general intelligence

#532

Post by olalonde »

piinbinary wrote: post #258

What's a useful definition of what AGI is? If you have a computer that does some clever AI stuff, what criteria do you look for to decide that it has a general intelligence?

It's generally used to mean human level intelligence (e.g. an AI that can learn any task a human can). An often cited criteria is passing the Turing test.

Re: John Carmack: I’m going to work on artificial general intelligence

#533

Post by jacquesm »

xamuel wrote: post #374
noelsusman wrote: post #370

Earlier quoted context omitted.

But the fact that there's so much low hanging fruit is kind of the point. AGI as a field is still in its infancy. It's pure research and will be for a long time.

So if you want "Carmack's Theorem" or "Noelsusman's Theorem" to appear in high school math textbooks 500 years from now, this is the best time to jump in. (Assuming there are still textbooks 500 years from now.) "Doom" probably won't be remembered by then.

I won't be around but 500 years from now the chances are much better than even that Doom will still be around in some form and a couple of people might even know who wrote it back in the stone age of technology. Not so sure about Carmack's Theorem.

Re: John Carmack: I’m going to work on artificial general intelligence

#534

Post by jacobush »

unityByFreedom wrote: post #458
bitexploder wrote: post #433

Earlier quoted context omitted.

I am not convinced current AI is the approach to an AGI. I think it is at least feasible someone outside of this sphere has a reasonable shot at it. It feels like many AI researchers get caught up in refining existing techniques that amount to fancy statistics algorithms and data crunching, but not AGI. Current AI techniques may be synthesized or used in part in some AGI but it’s clear there is a revolutionary step t…

> I am not convinced current AI is the approach to an AGI This is a non-sequitor. I didn't argue that current AI theory, or Kaggle, will lead us to AGI. > Kaggle is almost just an optimization fest Public machine learning competitions have produced a lot of innovative learning techniques. If Kaggle is so easy, and AGI so hard, it would follow that anyone tackling AGI would have some experience applying machine learni…

Ok, I'll take a shot: "it would follow that anyone tackling AGI would have some experience applying machine learning competitively in some public space".

No, that would absolutely not follow. (I'm a pretty good devil's advocate, but I can't with this one.)

And given AGI would come from some completely new breakthrough not related to the current practice of "machine learning", competitions may be completely moot. They may be great for finding the nice increments in the state of the art of machine learning, but they are unlikely to help much with AGI.

I could even imagine a stumbling AGI being very stupid compared to just about any machine learning solution thrown at it - yet being undeniably AGI. Like a dog not being very good at DOTA, Star Craft or Chess, yet it undeniably possesses some kind of general intelligence.

Re: John Carmack: I’m going to work on artificial general intelligence

#535

Post by fastball »

Animats wrote: post #313

This is encouraging. If you're going to work on artificial general intelligence, a reasonable context in which to work on it is game NPCs. They have to operate in a world, interact with others, survive, and accomplish goals. Simulator technology is now good enough that you can do quite realistic worlds. Imagine The Sims, with a lot more internal smarts and real physics, as a base for work. Robotics has the same issue…

Sooooo, The Talos Principle?

https://en.wikipedia.org/wiki/The_Talos_Principle

Re: John Carmack: I’m going to work on artificial general intelligence

#536

Post by The_rationalist »

ilaksh wrote: post #64
The_rationalist wrote: post #11

I hope he'll go the hybrid symbolic and neural network way (causal and statistical), instead of just statistical. AGI needs a type system... I hope I'll achieve AGI before him but it's nice to know there's some real competition! (because, reader, there are almost 0 researchers seriously trying to achieve AGI in a not totally bullshit way. Only opencog and Cyc comes to mind).

Are you really that sure that the approaches to increasing the generality of AI being taken by LeCun (self-supervised model learning), Hinton (capsule networks) and Bengio (state representation learning) are all "total bullshit"?

From my reading, Hinton capsule networks seems far from being enough, it could at best be an incremental improvement. And is unrelated to English semantic parsing, it seems specialized for computer vision.

Re: John Carmack: I’m going to work on artificial general intelligence

#538

Post by sobani »

hnaccy wrote: post #521
a1studmuffin wrote: post #488

Earlier quoted context omitted.

Game AI has nothing to do with AGI (or even regular AI) beyond the surface level description OP provided. The reason game AI hasn't progressed in the last few decades isn't because technology is holding us back - after all we can already achieve impressive machine learning feats using current-gen GPUs - it's because bad NPC AI is by design , so players can learn to overcome them and succeed. Very few people want to p…

You're assuming game AI means an agent that directly competes with the player. Plenty of games have NPCs with scripted routines, dialog, triggers, etc that could be improved either by reducing the dev cost to generate them without reducing quality or reacting to player behavior more naturally.

Except in those cases it's even more important that the NPCs don't do anything unexpected. Those NPCs are like actors in a stage play, you don't want them to come up with their own lines and confusing the audience.

Don't forget there is a certain randomness with 'more natural' and with randomness you're going to invite Murphy to the party.

Re: John Carmack: I’m going to work on artificial general intelligence

#539

Post by jacobush »

TeMPOraL wrote: post #525
snowwrestler wrote: post #426

Earlier quoted context omitted.

As I like to say: lots of people are working on making a car that is smart enough to drive itself wherever a human wants to go. How many people are working on a car smart enough to tell humans to fuck off, it doesn’t feel like driving anywhere today?

Self-driving cars and AGI are two different targets. We don't want a car that has a mind and can argue for itself. We want a car that's smart, but otherwise just a domesticated animal. We want to turn cars into horses.

Not even horses. That would be cruel to the car. At most, like Rat Things. (They have their built in entertainment when they are not in active use.)

Re: John Carmack: I’m going to work on artificial general intelligence

#540

Post by TeMPOraL »

solipsism wrote: post #383
xamuel wrote: post #319

I've been dabbling in AGI and it seems like the field has a lot of low-hanging fruit. I'll bet Carmack can offer some significant contributions. I'll take an opportunity to plug a paper I recently published on comparing relative intelligence. The punchline will illuminate the low-hangingness of the fruit in this field. Suppose X and Y are AGIs and you want to know which is more intelligent. For any interactive reward…

For any interactive reward-giving environment E, you could place X into E and see how much reward X gets; likewise for Y. If X gets more reward, you can consider that as evidence of X being more intelligent. That's an odd definition of intelligence. By that definition, a bird is more "intelligent" than a human at the task of opening a nut. Seems like "fitness" would be a much more appropriate term. It seems especiall…

> Nothing about what you just described is general.

But once you start integrating over many environments, it seems to make more sense, and the "generality" would be with respect to the set of environments being considered.

Post reply on HN