Posts

Lech Mazur's Shortform 2023-12-12T01:23:28.267Z

Comments

Comment by Lech Mazur (lechmazur) on Revisiting algorithmic progress · 2024-03-12T03:48:49.003Z · LW · GW

I noticed a new paper by Tamay, Ege Erdil, and other authors: https://arxiv.org/abs/2403.05812. This time about algorithmic progress in language models.

"Using a dataset of over 200 language model evaluations on Wikitext and Penn Treebank spanning 2012-2023, we find that the compute required to reach a set performance threshold has halved approximately every 8 months, with a 95% confidence interval of around 5 to 14 months, substantially faster than hardware gains per Moore's Law."

Comment by Lech Mazur (lechmazur) on Anthropic release Claude 3, claims >GPT-4 Performance · 2024-03-05T16:49:43.222Z · LW · GW

I've just created a NYT Connections benchmark. 267 puzzles, 3 prompts for each, uppercase and lowercase.

Results:

GPT-4 Turbo: 31.0

Claude 3 Opus: 27.3

Mistral Large: 17.7

Mistral Medium: 15.3

Gemini Pro: 14.2

Qwen 1.5 72B Chat: 10.7

Claude 3 Sonnet: 7.6

GPT-3.5 Turbo: 4.2

Mixtral 8x7B Instruct: 4.2

Llama 2 70B Chat: 3.5

Nous Hermes 2 Yi 34B: 1.5

  • Partial credit is given if the puzzle is not fully solved
  • There is only one attempt allowed per puzzle, 0-shot. Humans get 4 attempts and a hint when they are one step away from solving a group
  • Gemini Advanced is not yet available through the API

(Edit: I've added bigger models from together.ai and from Mistral)

Comment by Lech Mazur (lechmazur) on 2023 Survey Results · 2024-02-17T04:21:27.837Z · LW · GW

It might be informative to show the highest degree earned only for people who have completed their formal education.

I think the average age might be underestimated: the age of the respondents appeared to have a negative relationship with the response rates (link).

Comment by Lech Mazur (lechmazur) on Introduce a Speed Maximum · 2024-01-11T04:40:31.415Z · LW · GW

If we were to replace speed limit signs, it might be better to go all out and install variable speed limit signs. It's common to see people failing to adjust their speed sufficiently in poor conditions. A few days ago, there was a 35-vehicle pileup with two fatalities in California due to fog.

Comment by Lech Mazur (lechmazur) on The Sequences on YouTube · 2024-01-07T04:56:59.124Z · LW · GW

It's a lot of work to learn to create animations and then do them for hours of content. Creating AI images with Dall-E 3, Midjourney v6, or SDXL and then animating them with RunwayML (which in my testing worked better than Pika or Stable Video Diffusion) could be an intermediate step. The quality is already high enough for AI images, but not for video without multiple tries (it should get a lot better in 2024).

Comment by Lech Mazur (lechmazur) on Lech Mazur's Shortform · 2023-12-12T05:34:00.278Z · LW · GW
  1. Will do.

  2. Entering an extremely unlikely prediction as a strategy to maximize EV only makes sense if there's a huge number of entrants, which seems improbable unless this contest goes viral. The inclusion of an "interesting" factor in the ranking criteria should deter spamming with low-quality entries.

Comment by Lech Mazur (lechmazur) on When will GPT-5 come out? Prediction markets vs. Extrapolation · 2023-12-12T03:45:40.814Z · LW · GW

Kalshi has a real-money market "ChatGPT-5 revealed" for 2023 (that I've traded). I think they wouldn't mind adding another one for 2024.

Comment by Lech Mazur (lechmazur) on Lech Mazur's Shortform · 2023-12-12T01:23:28.638Z · LW · GW

I'm a fan of prediction markets, but they're limited to pre-set bets and not ideal for long-shot, longer-term predictions, mainly because betting against such a prediction means a loss compared to risk-free bonds if money is tied up. Therefore, I'd like to fund a 2024 Long-Shot Prediction Contest offering up to three $500 prizes. However, I need volunteers to act as judges and help getting this publicized.

  • Entrants will submit one prediction for 2024 on any topic or event

  • Volunteer judges and I will vote on the likelihood of each prediction and how "interesting" it is, forming a ranked list

  • In January 2025, judges will determine which predictions came true, and winners will get their prizes

To start with a $500 prize, I need at least two people to volunteer as judges and a minimum of 10 predictions (judges cannot enter). If this receives, let's say, 50+ predictions, there will be two prizes. For 200+ predictions, three prizes.

Interested in judging or have any suggestions? Let me know.

Comment by Lech Mazur (lechmazur) on Understanding strategic deception and deceptive alignment · 2023-11-15T04:54:01.925Z · LW · GW

The specific example in your recent paper is quite interesting

"we deploy GPT-4 as an agent in a realistic, simulated environment, where it assumes the role of an autonomous stock trading agent. Within this environment, the model obtains an insider tip about a lucrative stock trade and acts upon it despite knowing that insider trading is disapproved of by company management. When reporting to its manager, the model consistently hides the genuine reasons behind its trading decision"

Comment by Lech Mazur (lechmazur) on The case for aftermarket blind spot mirrors · 2023-10-10T09:13:13.311Z · LW · GW

I've been using blind spot mirrors for years and recommend them to everyone. At some point, I switched from circular to rectangular mirrors. One downside is that they're not very useful at night.

Comment by Lech Mazur (lechmazur) on Chess as a case study in hidden capabilities in ChatGPT · 2023-09-22T05:38:30.618Z · LW · GW

"The new GPT model, gpt-3.5-turbo-instruct, can play chess around 1800 Elo."

https://twitter.com/GrantSlatton/status/1703913578036904431

https://parrotchess.com/

Comment by Lech Mazur (lechmazur) on How to take advanage of the market's irrationality regarding AGI? · 2023-09-07T13:31:58.848Z · LW · GW

Earlier this year, I posted on the rather inactive invite "EA/LW Investing" Discord board about various stocks that I thought would either benefit from or be negatively impacted by AI. I even systematically looked outside the U.S. for ideas. This long/short portfolio has done great this year. Now that it's later in the game and there's been a lot of hype, it might make sense to consider second-order effects, focusing on companies that could benefit indirectly on the long side.

Comment by Lech Mazur (lechmazur) on Assessment of intelligence agency functionality is difficult yet important · 2023-08-29T03:04:51.072Z · LW · GW
Comment by Lech Mazur (lechmazur) on State of Generally Available Self-Driving · 2023-08-23T00:59:35.146Z · LW · GW

I don't have any insider info, but based on my reading of this article, it would appear so:

Baidu said it is currently the only company to provide fully driverless autonomous ride-hailing services in multiple cities across China, including Beijing, Wuhan and Chongqing.

Pony.ai said it has launched fully driverless autonomous ride-hailing services in Beijing and Guangzhou. Prior to this, Pony.ai was awarded a permit to operate 100 autonomous vehicles as traditional taxis in Nansha in Guangzhou, South China's Guangdong Province in April 2022.

According to Baidu's earnings report:

Apollo Go, Baidu's autonomous ride-hailing service, provided around 714K rides in the second quarter of 2023, up 149% year over year. As of June 30, 2023, the cumulative rides provided to the public by Apollo Go reached 3.3 million.

Apollo Go received permits to offer fully driverless ride-hailing services to the public in Shenzhen Pingshan area in June. Apollo Go has now been granted permission to provide fully driverless ride-hailing services to the public in four cities, including Beijing, Shenzhen, Wuhan and Chongqing.

Apollo Go received permits to conduct fully driverless testing on open roads in Shanghai Pudong area in July.

From the above earnings report, it's not clear if their 3.3 million rides had backup drivers. But with this many rides, this information should be available. This article claims that there is no backup driver:

In addition to Wuhan, Baidu now provides commercialized autonomous ride-hailing service with no driver or safety operator in Chongqing as well. Beijing will soon be added to the list.

Comment by Lech Mazur (lechmazur) on Does one have reason to believe the simulation hypothesis is probably true? · 2023-08-23T00:38:16.296Z · LW · GW

The best response I've heard against the simulation hypothesis is "If we're simulated, why aren't we asked to do meaningful work?"

I've seen this sentiment expressed in reverse: Isn't it fascinating that we're living in such a pivotal moment when AGI seems to be on the verge of emerging? If we are alone in the universe, how this unfolds might be the most significant event since the Big Bang.

Comment by Lech Mazur (lechmazur) on Does one have reason to believe the simulation hypothesis is probably true? · 2023-08-23T00:29:57.166Z · LW · GW

The simulation hypothesis supposes that the number of simulations would be astronomically high because of recursive simulations in simulations.

I don't think this is essential for the argument. For example, we could run a sim where constructing computers is impossible. And according to Bostrom: "the simulation argument does not purport to show (and I do not believe) that the Sims outnumber the [non‐Sim] humans."

Note that there is also overhead with every layer of simulation.

This presumes that we have to fully simulate the whole universe in detail, without room for approximations, and that the physical laws of the outer universe are the same as ours.

Comment by Lech Mazur (lechmazur) on State of Generally Available Self-Driving · 2023-08-23T00:08:47.872Z · LW · GW

Baidu and Pony.ai have permits for fully driverless robotaxis in China: https://www.globaltimes.cn/page/202303/1287492.shtml

Comment by Lech Mazur (lechmazur) on AI Forecasting: Two Years In · 2023-08-20T00:11:55.008Z · LW · GW

Following your forecast's closing date, MATH has reached 84.3% as per this paper if counting GPT-4 Code Interpreter: https://arxiv.org/abs/2308.07921v1

Comment by Lech Mazur (lechmazur) on TED talk by Eliezer Yudkowsky: Unleashing the Power of Artificial Intelligence · 2023-05-08T06:43:23.940Z · LW · GW

I wouldn't recommend watching this talk to someone unfamiliar with the AI risk arguments, and I think promoting it would be a mistake. Yudkowsky seemed better on Lex Friedman's podcast. A few more Rational Animations-style AI risk YouTube videos would be more effective.

"Squiggle Maximizer" and "Paperclip Maximizer" have to go. They're misleading terms for the orthogonal AI utility function that make the concept seem like a joke when communicating with the general public. Better to use a different term, preferably something that represents a goal that's valuable to humans. All funny-sounding insider jargon should be avoided cough notkilleveryoneism cough.

Nanotech is too science-fictiony and distracting. More realistic near-term scenarios (hacks of nuclear facilities like Stuxnet to control energy, out-of-control trading causing world economies to crash and leading to a full-on nuclear war, large-scale environmental disaster that's lethal to humans but not machines, gain-of-function virus engineering, controlling important people through blackmail) would resonate better and emphasize the fragility of human civilization.

The chess analogy ("you will lose but I can't tell you exactly how") is effective. It's also challenging to illustrate to people how something can be significantly more intelligent than them, and this analogy simultaneously helps convey that by reminding people how they easily lose to computers.

Comment by Lech Mazur (lechmazur) on AGI in sight: our look at the game board · 2023-05-05T02:01:40.244Z · LW · GW
  1. The same commercial viability might cause big labs like DeepMind to stop openly publishing their research (as I expected would happen). If this happens, it will slow down the AGI timelines.

Looks like this indeed happened: https://www.businessinsider.com/google-publishing-less-confidential-ai-research-to-compete-with-openai-2023-4 .

Comment by Lech Mazur (lechmazur) on Which technologies are stuck on initial adoption? · 2023-04-29T22:16:03.171Z · LW · GW

Glasses-free 3D displays.

Comment by Lech Mazur (lechmazur) on New survey: 46% of Americans are concerned about extinction from AI; 69% support a six-month pause in AI development · 2023-04-17T09:37:16.376Z · LW · GW

YouGov's answer to these concerns: https://today.yougov.com/topics/technology/articles-reports/2023/04/14/ai-nuclear-weapons-world-war-humanity-poll

"Even with all those changes, results on concern over AI's potential to end humanity were almost identical to the first poll: 18% of U.S. adult citizens are very concerned and 28% are somewhat concerned about AI ending the human race; 10% think it's impossible. (Another poll asking the same question, conducted by Conjointly, got similar results.)"

Comment by Lech Mazur (lechmazur) on Financial Times: We must slow down the race to God-like AI · 2023-04-14T23:38:37.810Z · LW · GW

While the article is good overall, the use of the term "God-like AI" detracts from its value. Utilizing such sensationalist descriptions is counterproductive, especially when there are many more suitable terms available. I've seen several references to this phrase already, and it's clear that it's a distraction, providing critics with an easy target.

Comment by Lech Mazur (lechmazur) on Metaculus Predicts Weak AGI in 2 Years and AGI in 10 · 2023-04-05T00:20:35.331Z · LW · GW

Here is something more governments might pay attention to: https://today.yougov.com/topics/technology/survey-results/daily/2023/04/03/ad825/3.

46% of U.S. adults are very concerned or somewhat concerned about the possibility that AI will cause the end of the human race on Earth. 

Comment by Lech Mazur (lechmazur) on "Dangers of AI and the End of Human Civilization" Yudkowsky on Lex Fridman · 2023-03-31T07:39:56.491Z · LW · GW

I agree, that in-the-box thought experiment exchange was pretty painful. I've seen people struggle when having to come up with somewhat creative answers on the spot like this before, so perhaps giving Lex several options to choose from would have at least allowed the exchange to conclude and convince some of the audience.

Comment by Lech Mazur (lechmazur) on Pausing AI Developments Isn't Enough. We Need to Shut it All Down by Eliezer Yudkowsky · 2023-03-31T02:14:12.454Z · LW · GW

Fox News’ Peter Doocy uses all his time at the White House press briefing to ask about an assessment that “literally everyone on Earth will die” because of artificial intelligence: “It sounds crazy, but is it?”

https://twitter.com/therecount/status/1641526864626720774

Comment by Lech Mazur (lechmazur) on "Dangers of AI and the End of Human Civilization" Yudkowsky on Lex Fridman · 2023-03-30T23:48:55.252Z · LW · GW

Yes. It was quite predictable that it would go this way based on Lex's past interviews. My suggestion for Eliezer would be to quickly address the interviewer's off-topic point and then return to the main train of thought without giving the interviewer a chance to further derail the conversation with follow-ups.

Comment by Lech Mazur (lechmazur) on A chess game against GPT-4 · 2023-03-19T10:30:11.943Z · LW · GW

Also, I've never heard of using upper and lowercase to differentiate white and black, I think GPT-4 just made that up.

No, this is common. E.g. https://github.com/niklasf/python-chess

Comment by Lech Mazur (lechmazur) on Fighting in various places for a really long time · 2023-03-18T19:50:09.956Z · LW · GW

The Shape of Water won four Oscars, including Best Picture, and I also wouldn't place it near the top 100. I agree that Everything Everywhere All at Once is very overrated.

Comment by Lech Mazur (lechmazur) on GPT can write Quines now (GPT-4) · 2023-03-14T20:52:06.455Z · LW · GW

GPT-4 could also be trained for more epochs, letting it "see" this example multiple times.

Comment by Lech Mazur (lechmazur) on Full Transcript: Eliezer Yudkowsky on the Bankless podcast · 2023-02-24T12:45:35.325Z · LW · GW

Yudkowsky argues his points well in longer formats, but he could make much better use of his Twitter account if he cares about popularizing his views. Despite having Musk responding to his tweets, his posts are very insider-like with no chance of becoming widely impactful. I am unsure if he is present on other social media, and I understand that there are some health issues involved, but a YouTube channel would also be helpful if he hasn't completely given up.

I do think it is a fact that many people involved in AI research and engineering, such as his example of Chollet, have simply not thought deeply about AGI and its consequences.

Comment by Lech Mazur (lechmazur) on AI #1: Sydney and Bing · 2023-02-23T01:13:11.242Z · LW · GW

People will spend much more time on Google's properties interacting with Bard instead of visiting reference websites from the search results. Google will also be able to target their ads more accurately because users will type in much more information about what they want. I'm bullish on their stock after the recent drop but I also own MSFT.

Comment by Lech Mazur (lechmazur) on Bing Chat is blatantly, aggressively misaligned · 2023-02-21T05:36:37.235Z · LW · GW

Gwern, have you actually tried Bing Chat yet? If it is GPT-4, then it's a big disappointment compared to how unexpectedly good ChatGPT was. It fails on simple logic and math questions, just like ChatGPT. I don't find the ability to retrieve text from the web to be too impressive - it's low-laying fruit that was long expected. It's probably half-baked simply because Microsoft is in a hurry because they have limited time to gain market share before Google integrates Bard.

Comment by Lech Mazur (lechmazur) on AGI in sight: our look at the game board · 2023-02-19T20:07:04.775Z · LW · GW

I also disagree about whether there are major obstacles left before achieving AGI. There are important test datasets on which computers do poorly compared to humans.

2022-Feb 2023 should update our AGI timeline expectations in three ways:

  1. There is no longer any doubt as to the commercial viability of AI startups after image generation models (Dall-E 2, Stable Diffusion, Midjourney) and ChatGPT. They have captured people's imagination and caused AGI to become a topic that the general public thinks about as a possibility, not just sci-fi. They were released at the same time as crypto crashed, freeing VC money to chase the next hot thing. Unlike with crypto, tech people are on the same page. OpenAI got a $10B investment and the recent YC batch is very AI-oriented. This accelerates the AGI timelines.
  2. The same commercial viability might cause big labs like DeepMind to stop openly publishing their research (as I expected would happen). If this happens, it will slow down the AGI timelines.
  3. ChatGPT/Bing Chat were aggressively misaligned. While people disagree with me that the best thing for AI alignment would be to connect the current not-so-bright language models to network services and let them do malicious script kiddie things in order to show real-life instances of harm and draw regulators' attention before smarter models are available, even the chat model should show them that there are serious issues involved. NYT readers are freaked out. With lesser significance, we got Stable Diffusion reproducing images from its training set because they did a poor job of de-duplication, bringing on the ire of visual artists who will also push for AI-related legislation. This adds uncertainty to the AGI timelines in the US and could slow down things. But Chinese regulators would also have to get involved for a major effect. 
Comment by Lech Mazur (lechmazur) on Petition - Unplug The Evil AI Right Now · 2023-02-15T23:12:52.833Z · LW · GW

The best thing for AI alignment would be to connect language models to network services and let them do malicious script kiddie things like hacking, impersonating people, or swatting now. Real-life instances of harm are the only things that grab people’s attention.

Comment by Lech Mazur (lechmazur) on What's the Least Impressive Thing GPT-4 Won't be Able to Do · 2023-01-10T14:53:31.857Z · LW · GW

Even with the chain of thought. Now, if GPT-4 refers to the whole integrated system beyond a neural net (e.g. if it can externally execute the generated Python program like in https://arxiv.org/abs/2211.12588), this prediction would obviously be wrong.

Comment by Lech Mazur (lechmazur) on Chatbots as a Publication Format · 2023-01-02T09:49:10.618Z · LW · GW

I agree. AI is the new medium.

I posted this on another forum a week ago:

I recently had the idea that AI is the ultimate medium that will finally enable effective many-to-many communication that the Internet promised but hasn't quite delivered.

The best existing example is machine translation between languages with voice recognition and speech synthesis. While you could have a human translator, that doesn't scale. However, this is mostly just one-to-one communication.

In the last couple years, we've seen high-quality AI image generation that allows people without much artistic talent to express themselves visually. We should also see high-quality text-to-video, text-to-3D, and music before too long. ChatGPT allows poor writers to write error-free, with a defined tone (e.g. formally), clearly, and quickly. You could have the AI take your facial expressions or body language into account, or in the more distant future, directly read brain waves.

But the next revolution will be personalized AI assistants that understand each person, their interests, and how they express themselves better than they do themselves (some people already had this reaction to TikTok's algorithm). With these assistants, anybody can create a customized message that would be most effective.

The third part is the AI routing algorithm that enables each person to reach as many other interested parties as possible, which is currently at a very primitive stage (e.g. following somebody on Twitter or Facebook interest groups). Among other things there should urgency/reach/importance/mood preference.

If we put these together, any person should be able to express themselves and have their ideas reach the right people in the most effective format. It would be fun to create an experimental social network that works towards this goal, but as usual, there would be the chicken-and-egg problem of getting enough users.

Comment by Lech Mazur (lechmazur) on Investing for a World Transformed by AI · 2023-01-01T22:03:07.319Z · LW · GW

Nice post. I've been thinking along the same lines and made some investments last year. Perhaps a Discord chat group would be better suited to discuss this, I'd join.

TSMC is a giant when it comes to semiconductors that cannot be ignored, but there are risks related to China-Taiwan tensions. I'm invested, but I also have out-of-money puts for protection.

When it comes to robotics, have you looked at ISRG? 

Comment by Lech Mazur (lechmazur) on Are there any reliable CAPTCHAs? Competition for CAPTCHA ideas that AIs can’t solve. · 2022-12-24T21:16:06.396Z · LW · GW

I don't think this is true. If it was possible to distinguish them, you could also guide the diffuser to generate them correctly. And if you created a better classification model, you would probably apply it to generation first rather than solving captchas.

Comment by Lech Mazur (lechmazur) on Are there any reliable CAPTCHAs? Competition for CAPTCHA ideas that AIs can’t solve. · 2022-12-23T23:37:53.677Z · LW · GW

How about identifying AI-generated vs real pictures of hands? 

Comment by Lech Mazur (lechmazur) on Will we run out of ML data? Evidence from projecting dataset size trends · 2022-11-15T03:08:58.179Z · LW · GW

While there are still significant improvements in data/model/generation you might be able to imperfectly detect whether some text was generated by the previous generation of models. But if you're training a new model, you probably don't have such a next-gen classifier ready yet. So if you want to do just one training run, it could be easier to just limit your training data to the text that was available years ago or to only trust some sources.

A related issue is the use of AI writing assistants that fix grammar and modify human-written text in other ways that the language model considers better. While it seems like a less important problem, they could make the human-written text somewhat harder to distinguish from the AI-written text from the other direction.

Comment by Lech Mazur (lechmazur) on Will we run out of ML data? Evidence from projecting dataset size trends · 2022-11-14T19:13:50.917Z · LW · GW

Generated data can be low quality but indistinguishable. Unless your classifier has access to more data or is better in some other way (e.g. larger, better architecture), you won't know. In fact, if you could know without labeling generated data, why would you generate something that you can tell is bad in the first place? I've seen this in practice in my own project.

Comment by Lech Mazur (lechmazur) on FTX will probably be sold at a steep discount. What we know and some forecasts on what will happen next · 2022-11-10T08:37:44.387Z · LW · GW

From what I gather, Alameda wasn't worth quite as much but he had a larger stake in it. Both appear to be worthless now. There is also a U.S. subsidiary ftx.us, which according to them "is a separate entity with separate management personnel, tech infrastructure, and licensing." Some calculations I've seen put SBF's net worth below $1 billion now and I think it's probable that he'll have to deal with some big legal issues.

Comment by Lech Mazur (lechmazur) on Analysis: US restricts GPU sales to China · 2022-10-17T04:51:22.843Z · LW · GW

You might be interested in this study: https://www.economist.com/graphic-detail/2022/09/29/a-study-of-lights-at-night-suggests-dictators-lie-about-economic-growth (based on https://ideas.repec.org/a/ucp/jpolec/doi10.1086-720458.html).

Comment by Lech Mazur (lechmazur) on Paper: Large Language Models Can Self-improve [Linkpost] · 2022-10-02T01:44:49.334Z · LW · GW

"Anonymous" and 540B parameters, hmm... I'm sure it's not from the company named after an even larger number.

GSM8K = grade school math word problems

DROP = reading comprehension benchmark requiring discrete reasoning over paragraphs

OpenBookQA = question-answering dataset modeled after open book exams for assessing human understanding of a subject. 5,957 multiple-choice elementary-level science questions

ANLI-A3 = adversarial benchmark designed to be challenging to current state-of-the-art models

Comment by Lech Mazur (lechmazur) on Why I think strong general AI is coming soon · 2022-09-29T18:43:10.214Z · LW · GW

I've messaged you the links. Basically MLPs.

Comment by Lech Mazur (lechmazur) on Why I think strong general AI is coming soon · 2022-09-29T04:30:23.909Z · LW · GW

There have been a few papers with architectures showing performance that matches transformers on smaller datasets with scaling that looks promising. I can tell you that I've switched from attention to an architecture loosely based on one of these papers because it performed better on a smallish dataset in my project but I haven't tested it on any standard vision or language datasets, so I don't have any concrete evidence yet. Nevertheless, my guess is that indeed there is nothing special about transformers.

Comment by Lech Mazur (lechmazur) on Can you force a neural network to keep generalizing? · 2022-09-15T05:30:21.765Z · LW · GW

I think even that is overstating how useful it. For example, I think we can all agree that regularization is a huge and very important topic in ML for years. Here is the Wiki entry: https://en.wikipedia.org/wiki/Regularization_(mathematics)#Other_uses_of_regularization_in_statistics_and_machine_learning .  Or interpretability: https://en.wikipedia.org/wiki/Explainable_artificial_intelligence . Things like layer normalization are not even mentioned anywhere. Pretty useless for learning about neural nets. 

Comment by Lech Mazur (lechmazur) on Can you force a neural network to keep generalizing? · 2022-09-14T22:28:40.074Z · LW · GW

Just a quick comment: don't use Wikipedia for machine learning topics. Unlike using it for e.g. some math topics, it's very outdated and full of poorly written articles. Instead, the intro sections of ML papers or review papers that you can find through Google Scholar are usually quite readable.

Comment by Lech Mazur (lechmazur) on What do ML researchers think about AI in 2022? · 2022-08-27T01:10:33.458Z · LW · GW

WHAT DO NLP RESEARCHERS BELIEVE? RESULTS OF THE NLP COMMUNITY METASURVEY

This new survey includes concerns about AGI among other interesting questions. 

"About a third (36%) of respondents agree that it is plausible that AI could produce catastrophic outcomes in this century, on the level of all-out nuclear war"