HNHacker News
TopNewBestAskShowJobs

salamo

408 karma · joined November 27, 2022

https://blog.lukesalamone.com
submissionscomments
salamo··on Golden Gate Claude
I'm wondering if this might be a cleaner way to incorporate symbolic knowledge bases into the model. In other words, with RAG we have to retrieve entire passages and ask the generator to pick out the relevant bits. Instead of doing that, if we had a knowledge base with clear-cut relationships there should be a way of "boosting" the relevant features once those features were identified in the model.
salamo··on Golden Gate Claude
Absolutely. "Refusals" all seem to follow a similar direction in the LLM residual stream, so preventing the model from representing this direction prevents it from refusing.

https://www.lesswrong.com/posts/jGuXSZgv6qfdhMCuJ/refusal-in...

salamo··on GPT-4o's image generation capabilities
Is this GPT4-o not the same as the one available on the pro plan? I can access it now.
salamo··on GPT-4o's image generation capabilities
The text is very visually appealing, but it still cannot draw chess boards properly. We may need to wait for AGI for that one.
salamo··on Veo
Yeah "AI complete" is a bit tongue-in-cheek but it is a fairly spectacular failure mode of every model I've tried.
salamo··on Veo
The first thing I will do when I get access to this is ask it to generate a realistic chess board. I have never gotten a decent looking chessboard with any image generator that doesn't have deformed pieces, the correct number of squares, squares properly in a checkerboard pattern, pieces placed in the correct position, board oriented properly (white on the right!) and not an otherwise illegal position. It seems to be an "AI complete" problem.
salamo··on Development Notes from xkcd's "Machine"
Really cool, I really enjoy the "follow" feature. Unfortunately, the balls get stuck sometimes and I have to follow a different one. I guess a few bugs slipped through the mod queue.
salamo··on AI Photo Geolocation
Neat demo. It seems like there may be a few things happening in tandem.

I uploaded a picture of a forest and it came back with visually similar images. So the first thing it might be doing is some kind of KNN, and if the pictures have location labels associated applying some sort of weighted average to determine GPS coordinates. This is pretty cool.

I also tried flipping the image horizontally, and it came back with the same images. So their embedding isn't based off of exact matches (good) and seems to be invariant to some basic translations (good). It also seems like it's directly extracting visual features from the image. This can be done with something like Blip[0].

Then I uploaded a screenshot from Magic School Bus. It still extracted information to guess the "location" of the cartoon (San Francisco, which is wrong). So that's probably how it works.

I also found the text output is similar in some ways with OpenGVLab InternViT [1]. So perhaps this or something like it is being used to extract features.

And of course there may be an LLM on top of these extracted features with some sort of prompt template. But I should add that the text explanation is the least useful part of the result, since it is unreliable and less informative than the "boring" similarity metrics above.

[0] https://huggingface.co/Salesforce/blip-image-captioning-larg... [1] https://internvl.opengvlab.com/

salamo··on Show HN: Chess bot predicts your blunders to beat you faster
No, just chessmate.ai
salamo··on Amazon virtually kills efforts to develop Alexa Skills
I've been waiting for a while for the tech to catch up, but it should be more or less possible now to use Whisper and an LLM to run a virtual assistant locally. Virtual assistants always seemed like a privacy nightmare to me so I never pulled the trigger.

But if it's all local? It would be a great product. No internet unless I turn on a physical switch, no ads, no monthly fee. Just do the thing I want you to do. Maybe at the moment it will still need to talk to my GPU. But we're getting there.

salamo··on Avoid blundering: 80% of a winning strategy
Chess.com is also definitely using an evaluation-centric definition to label moves as blunders. The issue is that this definition is also some function of the change in winning probability.

> So what is a blunder for a strong player may merely be a mistake for a weaker player.

Statistically this intuition appears to be correct. Your winning probability is still more than 25% when down a queen against an 800 rated player, but under 10% if playing a 2200: https://web.chessdigits.com/articles/when-should-you-resign#...

So it would make sense for the definition to take into account the opponent's Elo rating.

salamo··on Avoid blundering: 80% of a winning strategy
As a side note, I think this blunder-centric approach to chess is underappreciated, especially at low to mid levels where the vast majority of players peak. In addition to working on tactics, players should also specifically work on not blundering.

What this means in practice would be something like an "anti-tactic". In normal tactics, we need to find the best move which will usually win the game or give a decisive advantage. But an anti-tactic is a fight with your intuition: you want to make a move but that move is actually a blunder. There are several correct answers but one or two attractive but very wrong answers.

Of course the "rest of the owl" is how to determine which moves are very attractive to certain players. That is something I am working on now.

salamo··on Avoid blundering: 80% of a winning strategy
> The problem is that winning in startups is often a tail event, which means reducing blunders doesn't have much impact on the overall probability.

Maybe you could compare it to playing chess against a much higher rated player. In that sense, minimizing your weaknesses is still useful. And looking at the "blunders" the author lists we can definitely say that making these errors is worse than not making them.

One place where this analogy breaks down is that chess is a zero-sum perfect information game. Business is neither zero-sum (competition can encourage innovation and better products) nor perfect information. In chess, we can calculate for each of the candidate moves and have theoretically perfect knowledge of the future state of the board. In business and many other situations, this kind of thing isn't possible. The upshot is that several of the "blunders" the author lists may only turn out to be blunders after the fact.

salamo··on Grandmaster-Level Chess Without Search
The advantage of this approach that we can run many simultaneous computations on the GPU/TPU. Instead of using maybe a dozen CPU threads, we can approximate the value of a few thousand positions at the same time.
salamo··on Grandmaster-Level Chess Without Search
I think this is an interesting finding from a practical perspective. A function which can reliably approximate stockfish at a certain depth could replace it, basically "compressing" search to a set depth. And unlike NNUE which is optimized for CPU, a neural network is highly parallelizable on GPU meaning you could send all possible future positions (at depth N) through the network and use the results for a primitive tree search.
salamo··on RestGPT
It's actually really interesting to see GOFAI techniques like planning used in conjunction with LLMs.
salamo··on Ask HN: Tell us about your project that's not done yet but you want feedback on
Thanks for the feedback! Yes, the installation process is a bit barebones at the moment, a pip install would be a lot cleaner.

As far as the wiki link, it's a reference for the code sample I use in the demo.

salamo··on Ask HN: Tell us about your project that's not done yet but you want feedback on
https://github.com/lukesalamone/code_nitro

Convert your code to an image from the command line. No internet connection needed, although it can also download and convert code from Github too. Made it because currently the best alternative (carbon-now-cli) sends your code over the wire and there's a lot of situations where that's not desirable.

salamo··on Ask HN: Could you share your personal blog here?
https://lukesalamone.github.io

I usually write about machine learning & statistics topics as well as chess AI. I'm most proud of the search bar, which is all performed client-side so it's lightning fast.

salamo··on Ask HN: What has your personal website/blog done for you?
Hey, thanks for reading my post, I'm glad you liked it! I would say the purpose of alphabet chess is to be unequal, but in a semi-predictable way. That's why streamers will use the "egg opening": it puts them at a disadvantage and brings everyone out of "book".

> positions where the engine may consider the position to be almost equal, but it could be objectively worse for one of the players

That is a good point, centipawns aren't the best metric for the subjective experience of playing. Engines are really good at defending, finding the "only move" defense where a human will struggle. And the mental work of finding "only moves" is pretty taxing. If you put someone under that kind of pressure for long enough, they will break.

The analysis is also flawed because it doesn't take into account the fact that the opponent is also playing a word. If I know that my opponent must play a "G" move on his next turn, I should play a move to take advantage of that fact. Stockfish's game tree is not eliminating impossible moves and will tend to overrate the opponent's position.

(As a concrete example, suppose I can make a move M1 where the opponent's only move to not lose immediately is an F move, but if the opponent plays it they are at an advantage. In this situation the opponent must play a G move, a fact which a human might be able to guess. Stockfish will not play M1, but instead play a different move M2 that at least maintains equality. So the stockfish evaluation is pretty flawed.)

salamo··on Ask HN: What has your personal website/blog done for you?
I have a few ML-related demos on my blog that I created in school to help me to understand some fundamental concepts. I have pointed to them during interviews and I think they helped to demonstrate a depth of knowledge that words alone could not.
salamo··on Brazilian researchers find plastic rocks on remote island
Sure, but petroleum is also high in energy. There are even some bacteria that consume it. Yet petroleum is still available in huge deposits.
salamo··on TikTok “teen” AI filter is hitting hard for Gen-X users
https://en.wikipedia.org/wiki/AI_effect

> Researcher Rodney Brooks complains: "Every time we figure out a piece of it, it stops being magical; we say, 'Oh, that's just a computation.'"

salamo··on TikTok “teen” AI filter is hitting hard for Gen-X users
Actually it would be really interesting to see code that attempts to e.g. classify MNIST digits from first principles. No neural nets, no GBDTs, no clustering. And no cheating by trying to translate the above into if-then statements. Seems next to impossible.
salamo··on New Vehicle Safety Report Reveals Teslas Using Autopilot Are 10x Safer
> In the 3rd quarter, one crash was recorded for every 6.26 million miles driven using Autopilot. Teslas not using Autopilot technology logged one crash for every 1.71 million miles driven.

Autopilot might be safer, but this statistic is just a textbook example of selection bias. In order to compare one to the other, crashes in places autopilot doesn't work need to be excluded.

salamo··on Disqualified from a National Web Design Competition for Using GitHub
> OVERVIEW Applying leadership and 21st century skills, participants are required to design, build, and launch a website and present a given topic pertaining to technology.

I think if the purpose of the competition is to apply 21st century skills then they shouldn't ban 21st century tools.

salamo··on Ask HN: YC co-founder matching is bad
> kinda worried about how many people are looking to start an "ai startup" rather than solve a problem which might involve AI

My issue is that I am working on a project (I wouldn't call it a startup exactly) which does directly involve AI, so I attract attention from the exact types of people you mentioned, fair weather fans. If you have crypto in your profile and are now interested in AI, that tells me all I need to know.

I've been able to scare most of them away by not having a functioning landing page and directing them to a repo instead. Since I am looking for another technical person this has been a great filter.

salamo··on Sorting 400+ Chrome tabs in seconds
It only sorts on the title as far as I can tell.
salamo··on Sorting 400+ Chrome tabs in seconds
> The idea is simple: Give GPT3 a list of items and ask it to return a list of categories those items belong to.

60 seconds to sort 400 tabs is a hell of a price to pay. It needs to be worth it. Do you really need semantic search? Do you really need semantic search that queries a huge proprietary model?

(A really cool feature that Chrome doesn't have would be the ability to search the content of the webpage as well.)

What about something simpler, both architecturally and computationally. Start with something boring, like BM25. In my experience it works surprisingly well. If that's not good enough, then running a free model from Huggingface would be the next best option.

The benefit of this approach is that you can do it in the browser, in real-time (or at least locally with a downloaded model). Even 400 tabs will be a piece of cake. Plus you won't need to proxy precious API calls to a service which may become unavailable or very expensive with no warning.

salamo··on ChatGPT passes MBA exam given by a Wharton professor
There's even a specific term for shifting AI goalposts: https://en.wikipedia.org/wiki/AI_effect
← PreviousPage 3 of 4Next →