Hacker News Reader

Hedepigon Cloudflare Global Network experiencing issues

Except on HN

HAHA!

Our servers are still down, though

Hedepigon Ask HN: Work on robotics or agents?

I mean, this could be phase 2 of career growth. Focus on personal growth first by surrounding by experts.

Hedepigon Show HN: Llama 3.2 Interpretability with Sparse Autoencoders

The static nature of LLMs is a compelling argument against the reasoning ability.

But, it can learn, albeit in a limited way, using the context. Though to my knowledge that doesn't scale well.

Hedepigon Show HN: Llama 3.2 Interpretability with Sparse Autoencoders

I do see your point, and it is a good point.

My observation is that the models are better at evaluating than they are generating, this is the technique used in the o1 models. They will use unaligned hidden tokens as "thinking" steps that will include evaluation of previous attempts.

I thought that was a good approach to vetting bad ideas.

Hedepigon Show HN: Llama 3.2 Interpretability with Sparse Autoencoders

If we do figure out how to vet these thoughts, would you call it reasoning?

Hedepigon Show HN: Llama 3.2 Interpretability with Sparse Autoencoders

> then it is a correct formula to say 'it's not "true reasoning"'

why is that point fundamental?

Hedepigon Show HN: Llama 3.2 Interpretability with Sparse Autoencoders

This is not totally my experience, I've debated a successful engineer who by all accounts has good reasoning skills, but he will absolutely double down on unreasonable ideas he's made on the fly he if can find what he considers a coherent argument behind them. Sometimes if I absolutely can prove him wrong he'll change his mind.

But I think this is ego getting in the way, and our reluctance to change our minds.

We like to point to artificial intelligence and explain how it works differently and then say therefore it's not "true reasoning". I'm not sure that's a good conclusion. We should look at the output and decide. As flawed as it is, I think it's rather impressive

Hedepigon Garak, LLM Vulnerability Scanner

I read the original comment as hyperbole. But can see why it was confusing.

Edit: that came out way more condescending than I intentended

Hedepigon How I write code using Cursor

Have you had a go with the o1 range of models?

Hedepigon Root cause of Alzheimer's may be fat buildup in brain cells, research suggests

I think you're right

The danger is two fold

1. People don't eat fatty foods which have solid evidence for their benefits (e.g. Virgin olive oil)

2. People substitute the lack of fats with sugar, which I believe (not an expert) has a lot literature linking it with obesity

Hedepigon Vesuvius Challenge 2023 Grand Prize awarded: we can read the first scroll

Argh, what has Black Mirror done to our sense of optimism?

(I agree with you).

Hedepigon Thomas Cochrane

I mentioned in another comment, I enjoyed the structure and found it much easier to process. This is despite mostly reading long form articles and never having been a huge twitter user.

Hedepigon Thomas Cochrane

On the contrary, I have never been a heavy twitter user, yet I would dearly love all articles I read to be broken down like this. I definitely find it easier to process a list structure like this.

Hedepigon VideoPoet A large language model for zero-shot video generation

For now

Hedepigon Magicoder: Source Code Is All You Need

> tends to give subjectively better responses to "bad prompts".

I wonder if a first pass with another model to expand these so called bad prompts into better prompts would work.

Hedepigon Llamafile lets you distribute and run LLMs with a single file

Thank you for this :)

Hedepigon Llamafile lets you distribute and run LLMs with a single file

I am currently tinkering with this all, you can download a 3b parameter model and run it on your phone. Of course it isn't that great, but I had a 3b param model[1] on my potato computer (a mid ryzen cpu with onboard graphics) that does surprisingly well on benchmarks and my experience has been pretty good with it.

Of course, more interesting things happen when you get to 32b and the 70b param models, which will require high end chips like 3090s.

[1] https://huggingface.co/TheBloke/rocket-3B-GGUF

Hedepigon Why aren't motherboards mostly USB-C by now? (2021)

At the risk of sounding a bit dumb, I shorted my motherboard with a USB A, it felt like I got it right, and pushed it in, at which point it shutdown. Fortunately there didn't seem to be any permanent damage

Hedepigon The first stable release of a memory safe sudo implementation

Does it not provide some protection against a buffer overflow?

Hedepigon The first stable release of a memory safe sudo implementation

Yes it would be prevented by the borrow checker.

Hedepigon First observation of a virus attaching to another virus

If it's some consolation, I've just listened to a recent podcast on This Week in Virology (as recommended by another comment on this post. And they are talking about research that uses the CMV virus which has certain properties that make it a very effective and crucially long lasting vector that can potentially protect us against all manner of infections and cancers.

Hedepigon First observation of a virus attaching to another virus

> they're usually not considered "alive" in the same way cells are

Still, it's up for debate whether they are alive, regardless whether it's the same way that cells are alive

Hedepigon First observation of a virus attaching to another virus

That page mentions the potential for weaponisation with a 100% fatality rate. Well thank you very much for another thing to keep me awake at night.

Hedepigon Grist is a modern, relational spreadsheet

From another comment, looks like at least 8 years. Sorry!

Hedepigon Being picky about a CSS reset for fun

Out of interest, have you built anything non-trivial with tailwind?

I've used CSS, and Tailwind and not worrying about naming everything correctly and not skipping around different files is a dream for me. Perhaps that's my brain, Idk.

Further I have gone over old code, and haven't had any problem with maintainability

Hedepigon AI hype is built on flawed test scores

That sounds like a good litmus test. Do you have a specific example you've tried?

My opinion is it isn't binary, rather it's a scale. Your example is a point on the scale higher than what it is now.

But perhaps that's too liberal a definition of "reasoning" , no idea.

We seem to move the goalposts on what constitutes human level intelligence as we discover the various capabilities exhibited in the animal kingdom. I wonder if it is/will be the same with AI

Hedepigon AI hype is built on flawed test scores

I have a hunch I am misunderstanding your argument, but does that mean the only way to build a "true reasoning machine" would be to just create a human.

I guess what I'm really asking, what would you expect to observe to make it not illusory?

Hedepigon Flexoki, an inky color scheme for prose and code

I'm glad I'm not the only one who thinks this. It's really easy to trash on someone for not getting it right, but IMO we should congratulate on the achievement of actually making something, and give feedback on how to make it better.

Hedepigon Flexoki, an inky color scheme for prose and code

Perhaps it's archaic, but I was using the word amateur to mean "for the love of"

Edit

> Not sure why you’re framing this thread by different HNers as an attack somehow and your last statement feels judgmental.

I wasn't talking about the criticisms themselves, heck, I agree with them. I was more referring to the fact someone felt entitled to a good user experience they got angry.

Perhaps I'm being oversensitive on behalf of the guy who put effort into releasing this. But this isn't isolated, often people get punished for putting something out there that does not have a certain level of polish, I think that's totally counterproductive.