Amal

• Amal Nov 5, 2022, 6:59 PM

  2 points

  0

  in reply to: Amal ’s comment on: What will the scaled up GATO look like?

  I should also make a prediction for the nearer version of GATO to actually answer the questions from the post. So if a new version of GATO appears in next 4 months, I predict:

  80% confidence interval: Gato will have 50B-200B params. Context window will be 2-4x larger(similar to GPT-3)
  
  50%: No major algorithmic improvements, RL or memory. Maybe use of perceiver. Likely some new tokenizers. The improvements would come more from new data and scale.
  
  80%: More text,images,video,audio. More games and new kinds of data. E.g. special prompting to do something in a game, draw a picture, perform some action.
  
  75%: Visible transfer learning. Gato trained on more tasks and pre-trained on video would perform better in most but not all games, compared to a model with similar size trained just on the particular task. Language model would be able to descripe shape of objects better after being trained together with images/​video/​audio.
  
  70%: Chain of thought reasoning would perform better compared to a LLM of similar size. The improvement won’t be huge though and I wouldn’t expect it to gain some suprisingly sophisticated new LLM capabilities.
  
  80%: It won’t be able to play new Atari games similarly to humans, but there would be a visible progress—the actions would be less random and directed towards the goal of the game. With sophisticated prompting, e.g. “Describe first what the goal of this game is, how to play it, what is the best strategy”, significant improvements would be seen, but still sub-human.

• Amal Nov 5, 2022, 5:15 PM

  7 points

  1

  on: All AGI Safety ques­tions wel­come (es­pe­cially ba­sic ones) [~monthly thread]

  Isn’t the risk coming from insufficient AGI alignment relatively small compared to vulnerable world hypthesis? I would expect that even without the invention of AGI or with aligned AGI, it is still possible for us to use some more advanced AI techniques as research assistants that help us invent some kind of smaller/​cheaper/​easier to use atomic bomb that would destroy the world anyway. Essentially the question is why so much focus on AGI alignment instead of general slowing down of technological progress?

  I think this seems quite underexplored. The fact that it is hard to slow down the progress doesn’t mean it isn’t necessary or that this option shouldn’t be researched more.

• Amal Nov 5, 2022, 12:32 PM

  4 points

  1

  in reply to: Daniel Kokotajlo’s comment on: What will the scaled up GATO look like?

  I see. I will update the post with some questions. I find it quite difficult though to forecast on how percentages of the performance metrics would improve, compared to just predicting capabilities as the datasets are probably not so well known.

• Amal Nov 5, 2022, 12:18 AM

  5 points

  3

  on: What will the scaled up GATO look like?

  Ok, I was thinking about this a bit and finally got some time to write it down. I realized that it is quite hard to make predictions about the first version of GATO as it depends on what the team would prioritize in development. Therefore I’ll try to predict some attributes/​features of a GATO-like model that should be available in next two years, while expecting that many will appear sooner—it is just difficult to say which ones. I’m not a professional ML researcher so I might get some factual things wrong, so I would be happy to hear from people with more insight to correct me.

  First the prediction regarding the size of the model: I would expect a GATO-like architecture to have a larger amount of commercial success/​usefulness than e.g. GPT-3 and so the investment should also be higher. Furthermore, I would guess that there would be several significant improvements to the training infrastructure e.g. by companies such as Cerebras/​Graphcore etc. Therefore I estimate the model to use somewhere between 10-100x more compute compared to GPT-3. This might result in the model having more parameters, a larger context window, or most likely both. I predict the most likely context window size to be ~40000 (10x more) and param count 1T (6x compared to gpt). Regarding the context window—I think there will be some algorithmic improvements, so it won’t be the same as before (see bellow).

  Since GATO is multimodal, I would expect the scaling laws to change a bit due to the transfer learning. E.g. since the model won’t need to find all information about the shape of the objects in the text, but instead can just look at the images, it should be much easier for it to answer questions such as “Can scissors be inserted into a glass bottle?” and so require a significantly smaller amount of data. Thus the scaling laws would also need to be multi-dimensional to answer what is the optimal ratio of audio/​text/​image/​video to achieve best results. For example to improve the language model part of GATO, we may counterintuitively need to train on more images, instead of text.

  I predict that GATO would be trained on text, images, audio and video. I belive they would try to do also image/​audio/​video generation which they didn’t do in the current version, by essentially just trying to predict what is the next image/​video token, instead of using diffusion models. The context window size seems too small for video, however I believe there are two reasons why it won’t be a huge problem. First, by using something like Perceiver where more processing is happening on tokens seen recently and just a small amount of computation is used on older far away tokens, the context window could be increased significantly (or some other kind of memory could be added).

  Second, I don’t think that the model needs to see the whole image/​video. When humans look at something, only a very small part of the image is sharp and the rest is blurry. Similarly, I think that Gato will get image information with just a small amount of tokens that describe a small rectangle of the picture/​video clearly and some small number of tokens decribing the blurred rest. There would further be action tokens describing the “eye movement” when the focus shifts to a different part of the image. In this way I think GATO will be able to watch/​generate videos or read/​write long books.

  Futhermore, I think that in general, RL could be used to make the model “think slow” when predicting the tokens. For example, instead of the task of predicting the next token, GATO could be trained on a RL task “find a set of actions by which you find what the next token is”. So to predict what the next image token, next word, or next action in a game should be, it would first look around the image/​video/​book and only after collecting the relevant information, it would emit the right token. Possibly it could also emit tokens summarizing the information it has seen so far from a large amount of data that it might need in the future. Of course it would probably still be trained on Atari games(likely now with actual RL) or in some coding environment with predefined inputs/​outputs, but I think these will be much less significant compared to the “information finding RL”. Maybe a smaller feature would be that GATO would be able to emit commands modifying its context window and deleting/​adding tokens to it.

  So some capabilities that I would predict in 2 years: Generation of images,video,audio,text, even quite long ones. E.g. size of a book, 5min. long video etc. Instead of “Let’s think step by step” we would have “Let’s sketch the solution”, to draw diagrams etc. Gato would be able to reasonably operate a computer using text commands—e.g. search on the internet, use paint, IDE/​debugger and so on. It will be much better at coding by combining various prompting methods and would perform in top 10% of competitive programmers (compared to Alphacode being in 50th percentile). Solve some IMO problems (probably wouldn’t get gold medal, but maybe bronze by ignoring combinatorics). Act like a smart research assisant—e.g. finding relevant papers, pointing their strengths/​weaknesses, suggesting improvements/​ideas. Learn to play entirely new (atari) games similarly fast as humans—this would probably however require RL instead of just being a prediction model. Complete IQ test and get above average result.

  Capabilities I don’t expect: generating novel jokes, outperforming best humans at long-term planning, research, math or coding. Image/​Video generation wouldn’t match reality. Similarly, AI generated books wouldn’t sell well. Empathy—it wouldn’t make a very good friend. It will be slow and expensive—so real-time robotics will probably still be a challenge. Also, It won’t be a very reliable doctor, despite being connected to the internet.

• Amal Oct 26, 2022, 4:37 PM

  6 points

  4

  on: What will the scaled up GATO look like?

  this has generated much less engagement than I thought it would...what am I doing wrong?

What will the scaled up GATO look like? (Up­dated with ques­tions)

• Amal Oct 24, 2022, 6:09 PM

  4 points

  2

  on: AGI in our life­times is wish­ful thinking

  thanks for this post! I think it is always great when people share their opinions about the timelines and more people(even the ones not directly involved in ML) should be encouraged to freely express their view without the fear that they will be held accountable in the case they are wrong. In my opinion, even the people directly involved in ML research seem to be too reluctant to share their timelines and how they impact their work which might be useful for others. Essentially, I think that people should share their view when it is something that is going to somehow influence their decision making, rather than when they feel it crosses some level of rigour/​certainty, therefore posts like this one should receive a bit more praise (and LW should have two types of voting also for posts not just comments).

  While I disagree with the overall point of the post, I agree that there is probably a lot of wishful thinking/​curiosity driving this forum and impacting some predictions. However, even despite this, I still think that AGI is very close. My prediction is that TAI will happen in the next 2-5 years(70%) and AGI in the next 8 (75%). I guess it will be based on something like scaled-up GATO pre-trained on youtube videos with RL and some memory. The main reason for this is that deep-learning was operating on a very small scale just two years ago(less than a billion parameters) which made it very difficult to test some ideas. The algorithmic improvements to me seem just too easy to come up with. For example, almost all important problems e.g. language,vision, audio, RL were solved/​almost solved in a very short time and the the ideas there didn’t require much ingenuity.

  Just a slight exaggeration—if you take a five year old and ask him to draw a random diagram, chances are quite high that if scaled up, this is a SOTA architecture for something. It is just hard to test the ideas, because of the engineering difficulty and the lack of compute. However, this is likely to be overcomed soon with either more money being thrown on the problem or architecture improvements—e.g. Cerebras and Graphcore seem to be doing some promising work here.

• Amal Jun 1, 2022, 3:53 PM

  2 points

  on: Deep­Mind is hiring for the Scal­able Align­ment and Align­ment Teams

  Hi Rohin, how long does it usually take to hear back if selected for the next stage? I applied two weeks ago but didn’t receive any other mail yet, so I was just curious if I still have a chance or was not selected.