News for hackers. Stuff that matters.
Re:How Will AI Learn Next? by robbrown451 on Friday October 06, 2023 @03:27PM (#37795110)
AlphaZero demonstrates that more human-generated data isn't the only thing that makes an AI smarter. It uses zero human data to learn to play Go, and just iterates. As long as it has a way of scoring itself objectively (which it obviously does with a game like Go), it can keep improving with literally no ceiling to how much it can improve.

Pretty soon ChatGPT will be able to do a lot of training by iterating on its own output, such as by writing code and analyzing the output (including using vision systems).

Here's an interesting thing I noticed last night. I have been making a lot of images that have piano keyboards in them. DALL-E 3 makes some excellent images otherwise (faces and hands mostly look great), but it always messes up the keyboards, as it doesn't seem to get how black keys are in alternating groups of two and three.

But I tried getting chatgpt to analyze an image, using its new "vision" capabilities, and the first thing it noticed was that the piano keys were not properly clustered. I said nothing about that, I just asked it "what is wrong with this image" and it immediately found that. What if it could feed this sort of thing back in, using similar logic to Alpha Zero?

That's just a tiny hint of what is to come. Sure, it typically needs human generated data for most things. It's already got thousands of times more than any human has looked at. It will also be able to learn from human feedback, for instance a human could tell it what it got wrong in a response (whether regular text, code, or image), and explain in natural language where it deviated from what was expected. It can learn which humans are reliable, so it can minimize the number of paid employees doing RLHF, using them mostly to rate (unpaid) humans who choose to provide feedback. Even if most users opt out of giving this sort of feedback, there will be plenty to give it new, good information.

Re:How Will AI Learn Next? by realistic2020 on Friday October 06, 2023 @04:03PM (#37795583)
With Alpha Go, you have a clear objective -- to win a game. How does that work for creative outputs?
Re:How Will AI Learn Next? by Jerrrry on Friday October 06, 2023 @04:14PM (#37795731)
>The discriminator in a GAN is simply a classifier. It tries to distinguish real data from the data created by the generator. It could use any network architecture appropriate to the type of data it's classifying.

https://developers.google.com/machine-learning/gan/discrimin...

Re:How Will AI Learn Next? by ToValueFunfetti on Friday October 06, 2023 @04:13PM (#37795727)
The same way we do it. Verifying that an output is good is far easier than producing a good output. We can write a first draft, see what's wrong with it, make changes, and iterate on that until it's a final draft. And along the way we get better at writing first drafts.
Re:How Will AI Learn Next? by robbrown451 on Friday October 06, 2023 @05:02PM (#37796383)
Read the last paragraph. You still have humans, but their input is more akin to a movie reviewer than a movie director/writer/actor/etc. It still takes skill, but it takes a lot less time.

RLHF typically employees humans, and that can be time consuming in itself, but less time consuming than creating content. And their efforts can be amplified. If they are actually rating unpaid humans, that is, users, who are willing to provide feedback and are also prompting the system. Plenty of people are happy to do this for free, and some of it happens, just as a byproduct of them doing what they're already doing, creating content and choosing, which comes out good and which one doesn't. Every time I am working through a coding problem with chatGPT, and it makes mistakes and I tell her about those mistakes, it can be learning from that.

People can also come up with coding problems that can run and test itself on. As a simple example, I imagine it's trying to write a sorting algorithm. It can also write a testing function simply tests that it is correctly sorted. They can also time its results, they can count how many steps had to do in that sense it can work just like Alpha zero, where there is an objective goal, which is to do it with the least clock cycles, and there's a way to test whether and how well it is a achieving that goal. While that may be a limited number of programming problems that that works for, by practicing on that type of problem it will presumably get better at other types of problems, just like humans do.

This is exactly what large language models do, they find a way to objectively test their writing ability, which is by having them predict words and things that they've never seen before. In a sense it's different from actually writing new creative content, but it is practicing skills that you need to tap into when you are creating new content. Interestingly, a lot of people will dismiss them as simply being word predictors, but that's not really what they're doing. They're predicting words when they're training, but when they're actually generating new content, they're not "predicting" words (you can't predict your own decisions, that doesn't make sense), they are choosing words.

Re:How Will AI Learn Next? by spacecadet on Saturday October 07, 2023 @04:05AM (#37799943)
I got her writing pretty advanced programs that generate fake data sets and self score those data sets. Fun little project to see what would happen.
Re:How Will AI Learn Next? by riku_iki on Saturday October 07, 2023 @03:15AM (#37799695)
> With Alpha Go, you have a clear objective -- to win a game. How does that work for creative outputs?

there are still tons of potentially valuable applications with clear objective: win stock market, create new material or design to maximize some metrics, etc.

Re:How Will AI Learn Next? by mindwok on Saturday October 07, 2023 @03:35AM (#37799798)
I think similar to humans, creativity will be an emergent behavior as a result of the intelligence needed to pass other tests. Evolution doesn't care about our art, but the capabilities we use to produce it also help us with survival.
Re:How Will AI Learn Next? by skybrian on Saturday October 07, 2023 @08:26AM (#37801136)
This isn’t directly about creativity, but I suspect a lot of training will happen in simulated environments. A sandboxed Python interpreter is a good example. There are plenty of programming questions to train on.
Re:How Will AI Learn Next? by nonameiguess on Friday October 06, 2023 @04:27PM (#37795932)
That works fine for purely text-based or digital knowledge domains. So, sure, many types of programming, probably most game play, certainly all video game play, many types of purely creative fictional writing.

I don't want to downplay those applications, but the killer breakthrough that the breathless world imagines and has wanted since Turing first talked about this is accurately modeling physical reality. "Invent a better engine" and what not. Without being physically embodied and being able to conduct experiments in the real world, you can't bootstrap that, short of simulating physics from first principles, which is not computationally feasible. You're inherently relying on some quorum of training material produced by embodied sources capable of actually doing science to be factually accurate.

Re:How Will AI Learn Next? by alexpetralia on Saturday October 07, 2023 @06:31AM (#37800558)
Not dissimilar to how large organizations operate today! Humans operate at the edge collecting sensory data (making measurements, inputting forms, etc.) and the "brain" is a giant management and software apparatus in the middle.
Re:How Will AI Learn Next? by ooterness on Saturday October 07, 2023 @02:56AM (#37799595)
Unfortunately, I think the current strategies for RLHF are a huge contributor to hallucination / confabulation.

In short, they're paying contract workers for quantity, not quality; they don't have time to do independent research or follow up on citations. Unsurprisingly, the LLM optimizes for superficially convincing bullshit.

Re:How Will AI Learn Next? by quickthrower2 on Saturday October 07, 2023 @06:08AM (#37800462)
> Among those who labeled demonstration data for InstructGPT, ~90% have at least a college degree and more than one-third have a master’s degree.

Source: https://huyenchip.com/2023/05/02/rlhf.html#demonstration_dat...

Re:How Will AI Learn Next? by yeck on Saturday October 07, 2023 @08:13AM (#37801056)
Controlling for academic experience probably raises the average accuracy of labelling, but by how much? Clearly having a degree will not make you omniscient in your major, let alone other subjects.
Re:How Will AI Learn Next? by mdekkers on Sunday October 08, 2023 @01:05AM (#37807998)
Are they getting paid on quantity or quality?
Re:How Will AI Learn Next? by robbrown451 on Sunday October 08, 2023 @04:51PM (#37814292)
Most jobs I've known factor in both. I would assume they have processes in place that incentivize quality. Some it is as simple as you have a manager that will fire you if you produce crap.

With billions in funding, and bad results causing bad press etc, you think that OpenAI would not have given this a bit of consideration?

Re:How Will AI Learn Next? by mdekkers on Sunday October 08, 2023 @10:49PM (#37816453)
> I would assume they have processes in place that incentivize quality.

> you think that OpenAI would not have given this a bit of consideration?

Those are just assumptions though. The issue is not “this was labelled as a shoe, but it’s a car”, the issue is about depth vs superficially, which is harder to verify. See also https://www.theverge.com/features/23764584/ai-artificial-int... for a well-sourced article on the subject.

Re:How Will AI Learn Next? by robbrown451 on Sunday October 08, 2023 @04:47PM (#37814263)
"In short, they're paying contract workers for quantity, not quality;"

How do you know this?

Just taking a wild guess, but I'd think a company with billions of funding, and a ton of people trying to find flaws in what they are producing, would have some processes in place to incentivize quality as well as quality.

What you are suggesting is that a company that produces a product that is based on balancing trillions of floating point numbers makes core business decisions in the most simplistic black and white terms. "Hey, lets just go with a one and a zero on this." Bizarre assumption.

Maybe I'm just good at prompting, and I'm not trying to trick it, but I don't see this "superficially convincing bullshit." Can you show me a chat where you have sincerely prompted it and gotten something that matches that description?

I often see responses that are better than I could have given even if given hours to research and compose them, and I'm a pretty good writer and researcher.

Here, since I'm asking you to share one where it fails as you say it does by creating "superficially convincing bullshit", I'll share several where it succeeds.

https://chat.openai.com/share/523d0fec-34d3-40c4-b5a1-81c77f...

https://chat.openai.com/share/e09c4491-fd66-4519-92d6-d34645...

https://chat.openai.com/share/233a5ae2-c726-4ddc-8452-20248e...

https://chat.openai.com/share/53e6bda1-fe97-41ce-8f5c-89d639...

https://chat.openai.com/share/19f80ea9-e6be-4ac3-9dd4-7ea15c...

Re:How Will AI Learn Next? by ooterness on Monday October 09, 2023 @12:24PM (#37822022)
Never forget that RLHF is driven largely by sweatshop labor:

https://www.washingtonpost.com/world/2023/08/28/scale-ai-rem...

These jobs are overwhelmingly paid by task, which puts a lot of pressure to go fast.

I assert the entire "hallucination" phenomenon is a side effect of these practices. When ChatGPT makes up a fake fact with fake sources to back it up, it's largely because such lies are rated very highly by the underpaid humans who aren't incentivized to follow up on sources.

Re:How Will AI Learn Next? by robbrown451 on Monday October 09, 2023 @07:45PM (#37826881)
"I assert the entire "hallucination" phenomenon is a side effect of these practices. When ChatGPT makes up a fake fact with fake sources to back it up, it's largely because such lies are rated very highly by the underpaid humans who aren't incentivized to follow up on sources."

It seems like with billions of investment, they could figure that out. It's commonly discussed as an extremely difficult problem to solve and the most important problem to solve in the most talked about industry on the planet. I'm having a problem believing that its something that is so easy to solve.

Are you suggesting that even with that much money, they have to do things the way things are "overwhelmingly" done, as opposed to being able to say "hey, we need it done this way instead, because it's important and we can pay for it."

It just seems pretty bizarre to think that the highest of high tech, that is massively funded, doesn't have the clout to just fix that in a heartbeat, if that's really where the problem is.

Re:How Will AI Learn Next? by dr_dshiv on Saturday October 07, 2023 @04:58AM (#37800184)
We need alphago for math problems. Anyone know of a project like this?
Re:How Will AI Learn Next? by rhdunn on Saturday October 07, 2023 @06:37AM (#37800586)
With AlphaZero there are clear evaliation metrics -- you win, lose, or draw the game given specific rules. With chess, there is even a way of detecting end-game threats via check. The zero human data approach works here because of that, allowing the computer to find optimal strategies.

With natural language you don't have that unaided feedback evaluation metric. Especially when given idioms, domain specific terms, etc.

This is slow and hardwork because you need to process some text, evaluate and correct that data, retrain and repeat with the next text. You also need to check and correct the existing data, because inconsistencies will compound any errors.