mm321

My experience with the late SO was, that the original question used to be ten years old and filled with responses that were no longer relevant to the current state.

What amazes me is that an LLM can actually make sense out of all this crappy training data and often produces meaningful answers. (maybe that's a sign that we already have AGI)

well, some pension funds would also suffer. Ordinary people will suffer as a result, similar to the crisis of 2008-2009.

> Let someone else take on the research and training costs, and charge a massive premium for your hardware.

the linked article is from July, it is almost three months old.

Now there may be a very relevant recent story; it has the potential to reduce the flow of new open weight models. This could change quite a few expectations, which are based on open weight models:

"China’s STAR 50 Index Falls 40%, Erasing $480 Billion as AI Stocks Crash"

https://www.technology.org/2026/10/27/china-ai-stock-bubble-...

https://www.reuters.com/world/china/china-ai-bubble-has-alre...

Reuters says there is a run, but also deeper reasons:

"There has been no obvious selling trigger in China, although traders said high prices, tax bills and the threat of the US or other nations banning Chinese equipment in data centers all contributed."

So far I don't see any discussions of this story on HN, interestingly.

Funny animated cartoon from the Soviet Union (has subtitles). You must see this:

“Robbery, Italian Style”

https://www.youtube.com/watch?v=vktTgOr4M1s

(it has a wiki page https://en.wikipedia.org/wiki/Ograblenie_po... )

Marcello Mastroianni is performing a bank heist, it goes well - but in the end he has to return all the money he borrowed. And that's where he is left with nothing...

they got the same responses in the article, because the system is caching its responses, per user prompt.

The text production of LLMs has an element of randomness. This whole genre of 'look what all AI models are saying' is a lie.

A not very well known fact: older versions of the numpy python library still require Fortran for its build process. I think newer versions don't need the GNU Fortran compiler any longer.

https://numpy.org/doc/1.19/user/building.html

I think it started with Russian invasion of Ukraine. Now we got a second cold war, multiple wars in the middle east are caused by this war (the Vega and Wagner groups trained Hamas for 7/10/2023, as Russia is creating conflicts via her alliances, these are creating diversions of mil. resources from the larger war). Nowadays most political discussion in Europe center around war / internal security. Everything goes according to the script in Mad Max fury road:

  My name is Max.
  My world is fire and blood.
  Why are you hurting these people?
  It's the oil, stupid.
  - Oil wars.
  - We are killing for guzzoline.
  The world is running out of water.
  Now there's the water wars.
  Once, I was a cop.
  A road warrior searching
  for a righteous cause.
  - to the terminal freak-out point.
  Mankind has gone rogue, terrorizing itself.
  Thermonuclear skirmish.
  The earth is sour.
  Our bones are poisoned.
  We have become half-life.
  As the world fell... 
  ...each of us, in our own way

actually Matt mentions the same problems as Scott Aronson. If an LLM works on a codebase with lots of tiny modules, or a code base that has ten data layers added by previous runs, then poor LLM will make bad decisions.

Now what happens to mathematics, if prior art gets filled up with spaghetti code like proofs that no one can verify? Maybe there is some hallucination somewhere in the middle that has been left unchallenged by the other agents? Nobody can tell.

> Is anybody getting good code out of these things reliably without a bunch of additional legwork?

Matt Pocock says that he knows how, but it comes with a lot of context fiddling.

https://www.youtube.com/watch?v=v4F1gFy-hqg

"Software Fundamentals Matter More Than Ever"

This lecture is very impressive. However I haven't managed to reach AI enlightenment, yet.

Very interesting. I wonder, if there are plans to add French in a future version, or will Kolibri remain bilingual? (English/German). The domains include industry/aerospace, so there would be some demand for French, presumably.

Good luck. I once knew how to fix shift/reduce reduce/reduce errors, but that was a long time ago. With recursive descent parser you need to check for left recursion, which is somewhat less tricky.

Thanks! Curious detail: no "What books are you reading now?" between 2019 and 2023. Were people too busy during the pandemic?

the default writing style of 'Google AI mode' is less tiresome than that of Claude or ChatGPT. I wonder if the difference is due to training or due to prompting.

I like his still of prompt creation "Give me a list of styles to choose from" then choosing one of the available options. SKILL.md files seem to be built upon assumptions of what the language model already knows and understands. Can it be that most skills are therefore tied to a particular LLM vendor/version combination ? The same prompt might be interpreted completely differently by a slightly different language model.

Here the author is asking the machine about what it knows, so there is less of an element of doubt.

"""OpenAI called the model a "generational leap" for areas such as cybersecurity, professional work, software engineering, and science, with the company's president Greg Brockman saying that it could eventually be seen as the arrival of artificial general intelligence"""

https://en.wikipedia.org/wiki/GPT-6_Astra

according to Sam Altman from last year, an LLM should 'solve quantum gravity', if it is to count as AGI. Did anything change since then?

https://www.businessinsider.com/sam-altman-openai-david-deut...

We usually see this kind of discussion around weekends. To me it looks a bit like psychological warfare, designed to drive up the mega valuation of some IPO candidates, in other words: to make LLM based systems more valuable by bashing an entire class of workers.

I don't quite believe in the doom scenario: a lot of stuff depends on the correct functioning of software. However language model don't have a complete model of reality, they won't be able to evaluate if a given change is sufficient. Therefore systems based on language models will not be able to provide complete solutions - for some time to come. There will still be quite a bit of work in software development.

It didn't come up with a counterexample for the P versus NP problem, I wonder if they just didn't ask about it...

> "Everything is a Remix" is a good watch on youtube that explains this

Not completely. Novelty used to be a major thing, when Humans did it. Another important criteria used to be if that new thing makes sense at all. Here the language model has a problem, as it doesn't have the means to evaluate this criteria.

I think this is a veiled advertisement for Norway. Half of the world is engulfed in wars and terror (half across the road towards Mad Max), however Norway is worrying about the environmental effect of salmon feces on its fjords.

Interesting link, thanks. What about the contradiction between the next two sentences?

“a human may select or arrange AI-generated material in a sufficiently creative way that ‘the resulting work as a whole constitutes an original work of authorship.’”

"copyright would extend to the material the human author contributed but would not extend to the underlying AI-generated content itself."

The linked document is mentioning the act of selecting AI generated images for a comic book as an example, where the result is adding material that is copyrightable on its own merit. I am not sure if the same line of reasoning would apply to programming.

"in one early case, for instance, the Office found that the selection and arrangement of AI-generated images with human-authored text in a comic book were protectable as a compilation."

I think that AI is adding an extra layer of politics to just about everything. It is as if we are entering a phase of super-extra politics, as everyone is trying to figure out what should come next.

Instead of a war with machines we get an eternal war among lawyers and managers. Don't know which prospect is worse.

At some point there will be no one left to review Claude generated code. I think that will be an interesting time, given our reliance on software based systems...

A linked-in post on CS enrollment at universities:

https://www.linkedin.com/posts/eric-pop_has-computer-science...

"In the past two years, the drop was ~20% at UC Berkeley and ~30% at UC Davis, while UC San Diego is the only one still rising."

On the other hand: LeetCode traffic rankings are up, so some people still solve programming problems by hand - or they hope to get a job, by being able to solve them.

https://www.semrush.com/website/leetcode.com/overview/

I think this is a confusing time. It will probably take a while, until things become clearer.

Wasn't Claude supposed to do all the coding by now? Did we have a change of mind or did we run out of tokens?

Wonderful, thanks! These are the Joo Janta 200 Super-Chromatic Peril Sensitive Sunglasses, powered by AI :-)

I suspect that distillation attacks may be slightly exaggerated. Most of the training data used during fine-tuning is now synthetic data. You can't just repeat the same stuff twice, therefore another LLM is writing a text book that is explaining a topic in detail, ideally without any gaps in the material.

Actually here is the full talk on youtube, the quote is at 20:26 https://www.youtube.com/watch?v=t9HmOz8H0qI

I saw a video on Linked-in. It's a talk by Steve Jobs from 1983, where he is anticipating, what an LLM can do now (at 4:03 in the video). https://lnkd.in/p/ddWU7cTp

My thoughts on the subject: What would Aristotle have said? The LLM is simulating Aristotle's response, based on syntactic and pragmatic dependencies between words and phrases taken from his works. Often you will get a sensible answers, but sometimes you won't. It will require some understanding of the larger context to tell the difference, you again need to read some books in order to tell the difference and in order to engage in a meaningful Socratic dialogue. Without human scrutiny you will get wrong results, and you need some real training to handle that.

just recalled: some significant open source projects are being rewritten in Rust, with the help of Claude. I don't know if these efforts will be successful in the long run, but in some way these news headlines may be creating a media dynamic as part of the IPO preparations?

[1] https://news.ycombinator.com/item?id=48870966 pgrust passes 100% of the Postgres regression tests

[2] https://news.ycombinator.com/item?id=48837877 Rewriting Bun in Rust

[3] https://news.ycombinator.com/item?id=48789325 My AI-built PHP engine in Rust passes 17% of PHP-src tests, renders WordPress (ekinertac.com)

Speaking of financing: how is the Anthropic IPO going, what is the timeline? They filed over a month ago, no news since. (I would have expected some spectacular news headlines that would be designed to fuel public interest in the impending IPO, but can't detect anything of substance)

Wow. Now did you try to check the setup with something like Claude Fable? Will it find issues, what kind of issues? Another question: how many tokens did this effort cost? Did you learn new prompting tricks?