gobdovan

Not really, but sounds interesting. Would you care to share some sites that offer better llms.txt than main web page? Or talk about some piece of info you easily found on llms.txt that was hard to navigate to on the regular website?

Perhaps we should not go into any even older professions on HN.

Notice how you describe a perfectly intelligible concept like 'unicorn in space' as meaningless, even though there's a clear picture you can form in your mind about it and you can imagine the exact evidence that would convince you of their existence.

We can meaningfully talk about 'unicorns in space' since it's analytically intelligible and merely syntetically unverified.

For someone who only cares about utility of findings, I suppose there's little value in analysing the phenomenology of consciousness. I suspect that for a similar profile of thinkers meditation started being interesting when they saw there's some downstream utility in 'emotional well-being'. So why bother analysing an uncomfortable, hard problem when there's no clear benefit to it?

I just googled 'Daniel Dennet about meditation' and, surprise-surprise: 'Daniel Dennett acknowledged that meditation had practical value for "settling and centering" the mind. While he tried and saw benefits in practices like Transcendental Meditation, he largely discarded the mystical "aura" surrounding it, viewing the practice through his strictly materialist and evolutionary framework'

I started reading the first part of your comment before opening the article and thought you were mocking AI bros. I then read the rest of your comment and was sure you're misrepresenting TFA. I clicked on the article and started at it in disbelief.

I am making a narrower claim than the Chinese Room argument. I actually see consciousness via mechanistic processes as entirely possible.

My point is about current LLMs specifically, which the article is clearly referencing. For a present day transformer, I can write down in my notebook everything the model ever sees as input, plus weights and architecture notes, and can compute the next token with pen and paper, just extremely slowly.

This does not prove that a computation cannot be conscious. But if the same transition from prompt to next token can be decomposed into an explicit sequence of arithmetic operations, then the burden is on the defender to explain: Where, in this process, consciousness is supposed to enter?

Mind that the Chinese Room experiment is comparing the mind of a Chinese speaker to a different mechanistic symbolic procedure. I am, however, executing the exact same mechanistic process, whether it is done on a GPU or with pen and paper.

My hope is that some magical consciousness process emerging from electricity circulation or whatever people believe the mechanism of consciousness would be in the case of LLMs obviously becomes implausible, unless you hold a particularly strong form of substrate-independence, stronger than what most substrate-independence supporters would need to accept.

> The Only Programming Language Built on Mathematics, Not Fashion

Had to reread the title again since I thought I opened a different article about TLA+.

As for SQL, if you're referring to DBMS systems, here's what E.F. Codd, inventor of relational algebra, had to say about them and the departure from his work: https://thaumatorium.com/articles/the-papers-of-ef-the-coddf...

Yeah, I think it would be an interesting thought experiment to replace parts of a brain with Neuralink-like devices until you have a fully digital brain, although I doubt it would end up anything like a current LLM.

Although I get that it's a metaphor, I really dislike the "some people are born with a GB300 while others have an L40S" part. Even as a joke, it ranks people like hardware tiers, which is dehumanising and uncomfortably close to eugenic language. On top of that, the analogy also breaks down if you try to implement it literally.

For an LLM, you have clear stages of mostly feedforward computation over finite numbers and a perfect way to reconstruct the computation.

For meat, even if you model it under a purely Newtonian approximation, you need to simulate at least the immediate closed system around it which is continuous, thermodynamic, chemical and so on. You'd need to choose an arbitrary time step and update enormous amounts of coupled physical state to get an inexact simulation of a minimal slice of reality.

You would have a much harder time obtaining even a substrate-independent dead organism, comapred to LLMs that are already substrate-independent, which is basically what my notebook example shows.

I suppose it's mostly about clear communication. If you are reviewing books for a movie, the job does not seem to be "will this become a successful adaptation?" so much as "what is the strongest movie latent in this book, and how do I communicate that to the people who can act on it?". Those people would then try to imagine how that script would be portrayed on the screen, what the acting would be like, what the scenes would look like and where the material would break under translation. Given you all have some understanding about what makes a great script and what makes a great movie, you make a pipeline that has multiple experts controlling different aspects of the transformation and generating the strongest final product based on the original book, which, from book adaptations I saw, most of the time is just a thin seed rather than a forced blueprint.

If it later turns out the material was not adaptable in the way you thought, I'd imagine that is not just a binary miss, since the reader, producer, writer and executives can discuss and try to see where their judgement failed and what went wrong. I get that the hard feedback is sparse, but it doesn't have to be researche-grade measurements as much as it has to be good judgement, constant reality checks, even if just from proxies, and good taste. I'd be curious if this sounds close to what you were doing.

PS: there's this Dalton + Michael YC advice for startups which seems relevant: when outcomes are highly uncertain, you can't judge the result-only whether you acted logically, ethically and treated people well along the way.. (https://www.youtube.com/watch?v=XgcdvIj5I-k)

You can take the weights and model description, write them down on a notebook, then, by hand, compute the next token. Try to do the same with meat.

> I honestly don’t know what it even means to be conscious

> subjugate other cultures (assuming you mean they're not conscious in other's minds)

Have you ever considered you might be a philosophical zombie? [0]

[0] https://en.wikipedia.org/wiki/Philosophical_zombie

Chess has set rules and is a closed world with a set objective. In maths, you make up the rules. For every clearly defined problem that everybody cares about and is not yet proven, there's someone who first who recognised the importance of that given problem and managed to define it clearly enough for it to be recognised as such.

There's also the separate, less glamorous issue that people don't want to talk about, which is proof reliability. [0] If you have systems to help you formalise the problems and leave an algorithm or AI or whatever solve it in a verifiable way, that's a win for both the mathematicians and the rest of the world.

The deeper question is whether AI can replace the human role in deciding what mathematics should be done and what concepts matter. If that's automated, then yeah, we're screwed.

[0] https://lamport.azurewebsites.net/tla/proof-statistics.html

Why wouldn't Meta also directly sell data if nobody cared about this distinction?

They did have immense networks and were constantly communicating with one another. In On the Origin of Species, I was amazed how Darwin basically knew all the top breeders and naturists form South America, Arab Gulf, Oceania, which he mentions by name. He mentions his peers from the first page in the introduction and the correspondence and responses he got from them.

There's not many geniuses without an ecosystem around them that produced them. And even if there are, how would we know about them if they weren't well connected enough to start mattering?

It's fun to think about why that is. Most compilers explore the code into sea of nodes, with explicit relation kinds (depends-on, computes, effect relations etc) that humans don't have access to in surface languages. Then they just reduce and shorten a lot of the unnecessary edge chains, sometimes duplicate code that improves scheduling and a lot of not-so-semantically relevant stuff to get a dataflow that is guaranteed to have the same external behaviour with the code they're compiling. So basically compilers work on representations that humans almost never see, with different objectives that humans have. If humans were to code directly in dataflow networks and could find a way to keep them tidy and neat, I think humans would have a chance to beat solutions generated from surface code and then compiled automatically.

I had the same thought about ML engineering a while back, when Google released the AutoML suite, that was banned from Kaggle competitions. At the time, it seemed obvious to me that the closer you were to the models, the easier it was for them to replace your work, since most of the work on models was itself hill-climbing, grind searching and mutation search. So, the more your work is an explicit, measurable search loop, the more automatable it is.

Same with prompts, most attempts seem to be fidgeting with the models till they get your intend right, which is also a matter of hill-climbing, subtle mutation, and so on.

If I were to clarify anything from the article, I'd probably say that I'd rather do the factorisation of programming roles by how long they already existed. If someone is an AI engineer and his work only became relevant a month ago, very probably it will be obsolete in another month. If they do the same thing for the past 10 years, changes are that their skills would be useful for another 10 years to come.

As a former 7th grader, I liked the style. If anything, the over-the-top-ness makes it less bitter and more of a self-parody. It's honest about its own immaturity, making everything lighthearted again.

It's also using only under-specific swearwords like 'm..f*king', which is not really instigating violence, attacking any characteristic directly, just exaggerated profanity to the point of unseriousness.

I'm not saying the style is good or that everyone should tolerate it, I'm saying only that for me the exaggeration softens out the sentiment. I'd argue it's also what you sing up given the URL.

Exactly, cops can impose force on you and their jobs requires them to do it. If they wouldn't have body cams, it would be hard to see if coercion was justified or not. Surveilling powerless folks works in the other direction. It only serves to make them easier to discipline and teaches them that deviations from approved behaviour will be deterministically punished. And I also don't believe anyone is currently paid to act on their best behaviour as a simple citizen. Although this kind of monitoring works quite well for children under 6, you don't want paternalistic institutions, since you will never outgrow them.

I think he's arguing about OpenAI vendoring specifically, where OpenAI has a lot of subprocessors, but AWS doesn't and there's not really a 3rd camp to choose from, yet. But even there you can't just choose AWS as I tried to illustrate in uncle comment.

This is too vague to respond to meaningfully.

Yeah, I understood the original point. And I'm tired of it.

I'm just tired of the 'everyone follows their immediate incentives while the system stays incoherent' as the de facto reality. I think shedding some light over the actual mechanics would maybe make someone consider 'perhaps we shouldn't allow our acquisition team just turn off their brain and choose the default to cover their bottoms; maybe vendors are worth more decision investment via actual thinking instead of performatively ending up on the default choice after a little ritualistic game of "eeny, meeny, miny, AWS"'.

I think it's worth pointing out that Jeff Bezos would fight this tooth and nail from happening in his companies. He popularised 'process as proxy'. Yet AWS as sold to external enterprises is the exact proxy Bezos warned against internally. Do what Bezos does, and even what Bezos preaches, just don't do by default what Bezos sells.

You're onto something, but from some Romanian local investigations and public statements into similar cluster of laws [0], it's just something banks lobby for, so they can have more control and consolidate their place in any transaction chain. Maybe govt benefits a little, but when you see limited cash withdraws or commercial transactions requiring a bank system, it's usually private gains at play.

[0] (Romanian article) https://economedia.ro/parlamentarii-usr-au-depus-un-proiect-...

Have you considered checking the actual AWS contract and the limited liability they explicitly stipulate in contracts and even linked docs from marketing materials?

If you read the fine print, you'll notice something funny. You are largely responsible for data loss, SLA claims require you to present concrete evidence, and the remediation you accepted is usually credits for future spend on specifically the same product you lost your data on.

And AWS fine print is actually quite reasonable compared with, say, GCP, where the SLA seems mostly useful so the enterprise acquisition team can say "they have SLA, I can't get fired for choosing them since I did my due diligence!", while GCP can say "you already accepted the proposed remedy when signing the contract, sue us and we'll just point you to it. Thanks for your trust.". [0]

[0] https://docs.cloud.google.com/storage/docs/storage-classes

^ Standard multi-region or dual-region storage has a 99.95% availability SLA, regional Standard has 99.9%, and regional Nearline, Coldline, or Archive can be as low as 99.0%. The credits are 10%, 25%, or 50% of the monthly bill for the affected service tier, with 50% as the aggregate monthly cap, applied to future use. Google also says the customer must request the credit within 30 days or forfeit it.

Interestingly, many other insects need blood to reproduce because they cannot produce some of the required proteins on their own. Some common flies do this too, including horseflies, black flies, sand flies, and others. Some famously transmit disease, like the tsetse fly.

It makes me wonder if this kind of technology is deployed, where should the stop line be? And I don't think it's a trivial question.

In plain English, they made mosquitos like repellent.

Now imagine Netherlands has a reputation in Europe as having the most bland and utilitarian cuisine.

If you want to see the extent of his respectfulness and depth, with his courage for self-analysis added on top, My Best Fiend would be my choice. It's a documentary about his relationship with Klaus Kinski, who was a pretty unhinged actor. It's fascinating to watch Herzog hold his own with Kinski, and to observe the strange mental space Herzog is in, somewhere between crystal-clear vision and complete madness.

Yeah, but there's attempts to fix it. The Cholet paper (https://arxiv.org/abs/1911.01547) is a good attempt, shifting from measuring ~skill to measuring ~acquiring skill. It's the framework behind ARC-AGI benchmarks.

[0] JR Flynn: The mean IQ of Americans: Massive gains 1932 to 1978.

[1] Legg, S., Hutter, M.: A collection of definitions of intelligence

[2] François Chollet, On the Measure of Intelligence