nateb2022

> I continue to believe that the most concerning aspect of neural networks is that they really strengthen the argument that consciousness is the illusion.

Based on what? Just because you can model the weather using math doesn't imply at all that the weather is the product of a mathematical model itself. I don't see how reasoning is different.

We deliberately construct LLMs so that they approximate certain relations typical of reasoning. Suppose they become extraordinarily successful at doing so. What exactly have we demonstrated? That these effects can be artificially reproduced through such a mechanism. But how would it follow that the thing being reproduced must itself be nothing more than that mechanism?

Consider a calculator, a machine capable of valid transformations wrt the semantics of arithmetic.

Does a calculator reason its way to 2 + 2 = 4? No, it merely transforms.

An LLM is simply a much more complex version of this; except that its complexity hides a lot of what is otherwise reducible to a transformation of semantic, conceptual, and inferential relations. There are lots of NLP models prior to the introduction of modern "AI" that would make these representations more clear.

However I don't believe any of the aforementioned philosophers would ever consider reason itself as a transformation, as Kant especially spent a lot of time differentiating apprehension from transition.

> Knowledge and reasoning are inextricably interwoven in the weights of the neural network—there is no independent, explicitly represented set of beliefs.

I think an analogy would be helpful. As an LLM, reasoning through text, you wouldn't 'know' the idea of a man separately from, posterior to, the words man in say, French and English. As a human, you WOULD know the idea of man, and you would mean that idea when you say the French or English words for man.

For an LLM, although its approximation of knowledge would lead it to claim it knows they're the same thing, there would be differences in its weights that influence its usage of both the French and English words for man, and which may lead it to conclusions in one language it wouldn't reach in another. Because its knowledge/reasoning is interwoven to its knowledge, not prior to it.

> How does a set of beliefs help with reasoning?

You cannot have a syllogism without propositions.

> (eg more batteries included)

That's not the definition of a framework.

You call a library. A framework calls you.

Any source on the benchmarks/methodology besides the image? There's a ton of variance possible in llama.cpp's performance depending on how it was configured. I'd also like to see benchmarks against MLX.

> We made it closed source to reduce the potential for abuse (which would lead to Roblox blocking us again)

https://vinegarhq.org/Sober/FAQ/index.html#why-is-sober-clos...

I'd trust OP to make good architectural decisions/provide guidance even if AI mostly wrote the docs and UI. Looking into their GitHub, seems they are an engineering manager at Microsoft and contribute semi regularly to uBlue and Project Bluefin. Should be better quality than some random vibeslopper.

None of this speech is clear:

> Sentry's self-hosted RAM floor is the cost of a full suite. Design notes on an exception-only path: two containers, a DSN swap, and a dated RSS receipt.

This is the article's description. What is an exception-only path? I don't believe I've ever thought about designing one, and I don't understand its significant wrt the title whatsoever.

It's a really odd way to phrase a concept like "If you only care about logging errors, you can build 80% of Sentry for 20% of the engineering and infra costs."

> If you self-host on a small VPS and only need the next grouped exception, Sentry's RAM warning means the official install is a full suite. That suite is the undesired alternative.

This really reads like something a markov chain chatbot would write. Sentry's warning has nothing to do with installing a full suite. "That suite is the undesired alternative" is a really obtuse way to say bloated.

> A different job is narrower. When production throws, show a grouped issue with a stack you can read, and keep a crash loop from melting the box or the invoice. The rest of this note is that job.

A human would have written 'You don't need all of Sentry if you only need to log exceptions and stack traces.'

Seems like a sloplang.

Click this submission, navigate to the project's homepage by clicking the logo in the navbar, and use your browser's back button:

  Application Error
  Error: Unable to decode turbo-stream response at be (https://rescript-lang.org/assets/errorBoundaries-DO8A6pcs.js:2:8083)
  at async ge (https://rescript-lang.org/assets/errorBoundaries-DO8A6pcs.js:2:5991)
  at async Pt (https://rescript-lang.org/assets/jsx-runtime-C3U9CNM9.js:1:58170)
  at async Nt (https://rescript-lang.org/assets/jsx-runtime-C3U9CNM9.js:1:57906)
  at async Rt (https://rescript-lang.org/assets/jsx-runtime-C3U9CNM9.js:1:60260)
  at async Oe (https://rescript-lang.org/assets/jsx-runtime-C3U9CNM9.js:1:44211)
  at async ke (https://rescript-lang.org/assets/jsx-runtime-C3U9CNM9.js:1:45144)
  at async Se (https://rescript-lang.org/assets/jsx-runtime-C3U9CNM9.js:1:38537)
  at async be (https://rescript-lang.org/assets/jsx-runtime-C3U9CNM9.js:1:35508)

the phrasing 'boundaries' is very much an LLM tell; Claude definitely invented a load bearing footgun here. This happens on both Chrome and Safari; interestingly, on Safari the logo itself isn't clickable but substituting the Docs link suffices.

As someone very critical of mindlessly vibecoded software, I've worked with James (@prologic) years ago and very much respect his craft. AI assisted or not, I wouldn't describe him as a vibecoder. He's a great engineer.

Wang had been removed from his position as Editor-in-Chief, IEEE Trans on Intelligent Vehicles (https://ieee-itss.org/about/officers/past) prior to this; and IEEE has investigated multiple T-IV articles involving Wang:

  - https://ieeexplore.ieee.org/document/10486832
  - https://ieeexplore.ieee.org/document/10490245
  - https://ieeexplore.ieee.org/document/10499903
  - https://ieeexplore.ieee.org/document/10356797

So those retractions (seem to be peer review related) combined with the removal from the position of Editor-in-Chief, suggest something along the lines of editorial interference/manipulation of peer review.

> then we'd miss out on future efficiency improvements and savings.

Which I'm hoping will find its way back to our wallets but which I doubt will happen

Alberto (https://github.com/golang/go/issues/81639#issuecomment-58553...) found only a 75kB difference between the 'bloat' and your proposed patch in a stripped binary; and Roland (https://github.com/golang/go/issues/81639#issuecomment-58402...) contended:

> the symbols in your tester2 program account for ~67kb out of a 2.3mb binary

OP did find a "600k difference" in an unstripped comparison but if binary size was critical enough that 600k was a big deal, I'd assume users would be stripping the binaries outright or using a different language (tinygo perhaps) if a <100kb diff was on their list of concerns.

I think a lot of the focus on AI output poisoning future models' training has left their effects on our reasoning/ vocubulary neglected, which is why I brought it up.

How we define words, what connotations they evince and the qualities they exude, simultaneously limits and enables thought.

You're restating their point. Normies' output will still be bad compared to the output that good engineers would produce, with or without AI. You wouldn't want to use poorly built software regardless of origin.

HN data is accessible through the Firebase API: https://github.com/hackernews/api

You don't need to as much as parse HTML to read it. Building a client by hand in C even should be pretty painless as long as you don't write your JSON parser yourself.

charm makes some awesome libraries for building/serving CLIs over SSH: https://github.com/charmbracelet/wish

As an aside, 'meaningful' is an obtuse adjective to use there; and it's typical of text generated by Claude.[0] I seem to see it a lot more frequently than pre-AI (~2022). Maybe my next project will be statistically analyzing word frequencies in HN comments pre-AI and post-AI.

[0]: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7031138

Maybe next you can build something to prove a human created a HN submission

A bubble, like the housing bubble, involved people overpaying for things whose value became disconnected from reality. With AI, I'm aware of 0 people paying for an AI subscription against their will. Some people gravitate to cheaper models, some have no qualms paying SOTA prices. But regardless, I don't believe people in either camp would claim they're overpaying for something whose value has become disconnected from reality. On the contrary, the price of intelligence is rapidly decreasing.

I found these from 2024:

  - https://arxiv.org/pdf/2412.17199 
  - https://academic.oup.com/imrn/article/2024/16/11865/7704606

and re the Liouville function: https://en.wikipedia.org/wiki/Liouville_function

Conversely, Acer CEO could be downplaying 2030 shortage fears to protect investor confidence. Price discovery.

> Bespoke software can be orders of magnitude less complex.

Not in government. For payroll/HR, we're talking about hundreds of pages of legislative mandates, union bargaining rules, Title 5 statutory compliance, and FISMA/NIST regs. Just customizing something like PeopleSoft to fit is a large task in itself, let alone trying to implement bespoke software using an engineering pool that isn't even large enough to modernize the software it currently owns.

Bespoke sounds nice and works well in startups but that's not the context we're discussing. Check out Phoenix Pay: https://en.wikipedia.org/wiki/Phoenix_pay_system and understand the US is even more complex.

I think enterprise software in this vein used to have a quasi-monopoly due to the sheer work required to build software of its size (not enough engineers exist in the government to do so), and the difficulty for competitors to enter regulated markets and so pretty much 1-2 options to choose from.

AI now makes it possible to build this kind of software in-house, offering a 2nd choice, though it'll only be as good as the standards of the teams using it. Only time can tell.

There's incompetence in every major company including Oracle. PeopleSoft isn't known for being the most modern or secure thing out there. Less reliance on 3rd party software like this will be a good thing going forward.

> Maintained by the Sovereign Systems & Security Architecture Team.

Hmmm... must be some "team" with this user (https://github.com/mc493) contributing only this repository in all of three years. Complete slop.

I see no proof of care with the first commit 9 hours ago, and every commit by this "security architecture team" unsigned except for the AI output paste commits to README made thru the web editor.

Linux on ChromeOS is basically as capable as WSL now: https://chromeos.dev/en/linux

Although I would definitely prefer macOS or Windows + WSL (since I play a decent bit of games), ChromeOS is a lot more pleasant experience now than in the past.

https://googlebook.google/ has more info with no paywall

I think this part of their comment:

> The claims seem a bit overstated though, since the metrics mentioned are cherrypicking refusal count and KL divergence, both of which make the outcome seem the most dramatic.

is right out of an LLM. It's the kind of language I'd expect out of a thinking trace also mentioning "boundaries" and "oracles" and "contracts."

[dupe] https://news.ycombinator.com/item?id=45945587 (10 months ago, 387 comments)