Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AGI fantasy is a blocker to actual engineering

tomwphillips.co.uk

Post reply on HN
Print view View original

291–300 of 683 posts

Re: AGI fantasy is a blocker to actual engineering

#291

Post by RaftPeople »

naasking wrote: post #160
IgorPartola wrote: post #21

It is ultimately a hardware problem. To simplify it greatly, an LLM neuron is a single input single output function. A human brain neuron takes in thousands of inputs and produces thousands of outputs, to the point that some inputs start being processed before they even get inside the cell by structures on the outside of it. An LLM neuron is an approximation of this. We cannot manufacture a human level neuron to be s…

> To simplify it greatly, an LLM neuron is a single input single output function. A human brain neuron takes in thousands of inputs and produces thousands of outputs This is simply a scaling problem, eg. thousands of single I/O functions can reproduce the behaviour of a function that takes thousands of inputs and produces thousands of outputs. Edit: As for the rest of your argument, it's not so clear cut. An LLM can…

> This is simply a scaling problem, eg. thousands of single I/O functions can reproduce the behaviour of a function that takes thousands of inputs and produces thousands of outputs.

I think it's more than just scaling, you need to understand the functional details to reproduce those functions (assuming those functions are valuable for the end result as opposed to just the way it had to be done given the medium).

An interesting example of this neuron complexity that was published recently:

As rats/mice (can't remember which) are exposed to new stimuli, the axon terminals of a single neuron do not all transmit a signal when there is an action potential, they transmit in a changing pattern after each action potential and ultimately settle into a more consistent pattern of some transmitting and some not.

IMHO: There is interesting mathematical modeling and transformations going on in the brain that is the secret sauce for our intelligence and it is yet to be figured out. It's not just scaling of LLM's, it's finding the right functions.

Re: AGI fantasy is a blocker to actual engineering

#292

Post by simianwords »

lynndotpy wrote: post #26
simonw wrote: post #12

Tip for AI skeptics: skip the data center water usage argument. At this point I think it harms your credibility - numbers like "millions of liters of water annually" (from the linked article) sound scary when presented without context, but if you compare data centers to farmland or even golf courses they're minuscule. Other energy usage figures, air pollution, gas turbines, CO2 emissions etc are fine - but if you com…

Farmland, AI data centers, and golf courses do not provide the same utility for water used. You are not making an argument against the water usage problem, you are only dismissing it.

who are you to determine the utility? we have the market for it and it has spoken.

Re: AGI fantasy is a blocker to actual engineering

#293

Post by energy123 »

IncreasePosts wrote: post #226
abathur wrote: post #174

Earlier quoted context omitted.

Agriculture feeds people, Simon. It's fair to be critical of how the ag industry uses that water, but a significant fraction of that activity is effectively essential. If you're going to minimize people's concern like this, at least compare it to discretionary uses we could ~live without. The data's about 20 years old, but for example https://www.usga.org/content/dam/usga/pdf/Water%20Resource%2... suggests we were us…

Agriculture is generally essential but that doesn't mean that any specific thing done in the name of agriculture is essential. If Americans cut their meat consumption by 10%, we would use a lot less water in agriculture and probably also live longer in general

Iran's ongoing water crisis is an example. One cause of it is unnecessary water-intensive crops that they could have imported or done without (just consume substitutes).

It's a common reasoning error to bundle up many heterogeneous things into a single label ("agriculture!") and then assign value to the label itself.

Re: AGI fantasy is a blocker to actual engineering

#294

Post by simianwords »

danaris wrote: post #137
simonw wrote: post #113

Earlier quoted context omitted.

Once again, the key thing here is to ask how MUCH value we get per liter of water used. If data centers and farms used the same amount of water we should absolutely be talking about their comparative value to society, and farms would win. Farms use thousands of times more water than data centers.

Yes, it is worthwhile to ask how much value we get. And a whole bunch of us are saying we don't see the value in all these datacenters being built and run at full power to do training and inference 24/7, but you just keep ignoring or dismissing that. It is absolutely possible that generative AI provides some value. That is not the same thing as saying that it provides enough value to justify all of the resources bein…

it is also a fraction of golf courses which you again ignore. this is just typical "don't do anything!!" ism. there's no argument here.. even if data centres used .00001 millilitre of water you would say the same thing.

Re: AGI fantasy is a blocker to actual engineering

#296

Post by simianwords »

buellerbueller wrote: post #156
simonw wrote: post #113

Earlier quoted context omitted.

Once again, the key thing here is to ask how MUCH value we get per liter of water used. If data centers and farms used the same amount of water we should absolutely be talking about their comparative value to society, and farms would win. Farms use thousands of times more water than data centers.

Once again, you are ignoring my (implied) argument: Humans NEED food, the output of agriculture. Humans do not NEED any of LLMs' outputs. Once everyone is fed, then we can talk about water usage for LLMs.

fine so do the same performative activism for

1. netflix

2. gmail

3. hackernews

4. discord

5. gaming

Re: AGI fantasy is a blocker to actual engineering

#297

Post by mikemarsh »

ACCount37 wrote: post #275
mikemarsh wrote: post #269

Earlier quoted context omitted.

They're not asking about telling the difference in collected data sets, data sets aren't weather. The question is can you tell the difference between the rain you see outside your window, and some representation of a simulated environment where the computer says "It's raining here in this simulated environment". The implied answer is of course, one is water falling from the sky and one is a machine.

You can't look at the "real weather" though. You can only look at the outputs. That's the constraint. Good luck and have fun. A human brain is a big pile of jellied meat spread. An LLM is a big pile of weights strung together by matrix math. Neither looks "intelligent". Neither is interpretable. The most reliable way we have to compare the two is by comparing the outputs. You can't drill a hole in one of those and se…

Even granting all of your unfounded assertions; "the output" of one is the rain you see outside, "the output" of the other is a series of notches on a hard drive (or the SSD equivalent, or something in RAM, etc.) that's then represented by pixels on a screen.

The difference between those two things (water and a computer) is plain, unless we want to depart into the territory of questioning whether that perception is accurate (after all, what "output" led us to believe that "jellied meat spread" can really "perceive" anything?), but then "the output" ceases to be any kind of meaningful measure at all.

Re: AGI fantasy is a blocker to actual engineering

#298

Post by dmix »

ACCount37 wrote: post #270
dmix wrote: post #242

I always found it funny OpenAI staff tried to delay the release of GPT to the world because they feared the consequences of giving the public such a power. Hearing stuff like this makes it even funnier: > In the pit, [Sutskever] had placed a wooden effigy that he’d commissioned from a local artist, and began a dramatic performance. This effigy, he explained represented a good, aligned AGI that OpenAI had built, only…

We don't know enough about consciousness to be able to conclusively confirm or deny that LLMs are conscious. Claiming otherwise is overconfident stupidity. Of which there is no shortage of that in AI space.

That's the sort of convenient framing that lets you get away with hand wavy statements which the public eats up, like calling LLM development 'superintelligence'.

It's good for a conversation starter on Twitter or a pitch deck, but there is real measurable technology they are producing and it's pretty clear what it is and what it isn't.

In 2021 they were already discussing these safety ideas in a grandiose way, when all they had was basic lego building blocks (GPT 1). This isn't just a thought experiment to them.

Re: AGI fantasy is a blocker to actual engineering

#300

Post by mysecretaccount »

naasking wrote: post #160
IgorPartola wrote: post #21

It is ultimately a hardware problem. To simplify it greatly, an LLM neuron is a single input single output function. A human brain neuron takes in thousands of inputs and produces thousands of outputs, to the point that some inputs start being processed before they even get inside the cell by structures on the outside of it. An LLM neuron is an approximation of this. We cannot manufacture a human level neuron to be s…

> To simplify it greatly, an LLM neuron is a single input single output function. A human brain neuron takes in thousands of inputs and produces thousands of outputs This is simply a scaling problem, eg. thousands of single I/O functions can reproduce the behaviour of a function that takes thousands of inputs and produces thousands of outputs. Edit: As for the rest of your argument, it's not so clear cut. An LLM can…

I agree with both of you, but scaling isn't feasible with this paradigm. You could need continent-sized hardware to approximate general intelligence with the current paradigm.
Post reply on HN