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

231–240 of 683 posts

Re: AGI fantasy is a blocker to actual engineering

#231

Post by belter »

simonw wrote: post #168
belter wrote: post #159

Earlier quoted context omitted.

I am surprised by your analytical mistake of comparing irrigation water with data-center water usage... They are not equivalent. Data centers primarily consume potable water, whereas irrigation uses non-potable or agricultural-grade water. Mixing the two leads to misleading conclusions on the impact.

That's a really good point - you're right, comparing data center usage to potable water usage by towns is a different and more valid comparison than comparing with water for irrigation.

The problem is many data centers are in areas where water systems are supply constrained... - https://spectrum.ieee.org/ai-water-usage

Re: AGI fantasy is a blocker to actual engineering

#232

Post by ACCount37 »

HarHarVeryFunny wrote: post #189
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…

Assuming you want to define the goal, "AGI", as something functionally equivalent to part (or all) of the human brain, there are two broad approaches to implement that. 1) Try to build a neuron-level brain simulator - something that is a far distant possibility, not because of compute, but because we don't have a clear enough idea of how the brain is wired, how neurons work, and what level of fidelity is needed to ca…

Or

3) Pour enough computation into a sufficiently capable search process and have it find a solution for us

Which is what we're doing now.

The bitter lesson was proven right once again. LLMs prove that you can build incredibly advanced AIs without "understanding" how they work.

Re: AGI fantasy is a blocker to actual engineering

#233

Post by Capricorn2481 »

hamdingers wrote: post #215
the__alchemist wrote: post #97

Earlier quoted context omitted.

I will go meta into what you posted here: That people are classifying themselves as "AI skeptics". Many people are treating this in terms of tribal conflict and identity politics. On HN, we can do better! IMO the move is drop the politics, and discuss things on their technical merits. If we do talk about it as a debate, we can do it when with open minds, and intellectual honesty. I think much of this may be a reactio…

> IMO the move is drop the politics, and discuss things on their technical merits. I'd love this but it's impossible to have this discussion with someone who will not touch generative AI tools with a 10 foot pole. It's not unlike when religious people condemn a book they refuse to read. The merits of the book don't matter, it's symbolic opposition to something broader.

Okay, but a lot of people are calling environmental and content theft arguments "political" in an attempt to make it sound frivolous.

It's fine if you think every non-technical criticism against AI is overblown. I use LLMs, but it's perfectly fine to start from a place of whether it's ethical, or even a net good, to use these in the first place.

People saying "ignoring all of those arguments, let's just look at the tech" are, generously, either naive or shilling. Why would we only revisit these very important topics, which are the heart of how the tech would alter our society, after it's been fully embraced?

Re: AGI fantasy is a blocker to actual engineering

#234

Post by hamdingers »

dwohnitmok wrote: post #13

> And this is all fine, because they’re going to make AGI and the expected value (EV) of it will be huge! (Briefly, the argument goes that if there is a 0.001% chance of AGI delivering an extremely large amount of value, and 99.999% chance of much less or zero value, then the EV is still extremely large because (0.001% * very_large_value) + (99.999% * small_value) = very_large_value). This is a strawman. The big AI n…

Do they believe that, or do they need to project confidence in that vision to hit investment targets? I can't tell. Probably a mix.

Re: AGI fantasy is a blocker to actual engineering

#235

Post by ACCount37 »

zgk7iqea wrote: post #193
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…

it is an architecture problem, too. LLMs simply aren't capable of AGI

Why not?

A lot of people say that, but no one, not a single person has ever pointed out a fundamental limitation that would prevent an LLM from going all the way.

If LLMs have limits, we are yet to find them.

Re: AGI fantasy is a blocker to actual engineering

#236

Post by AndrewKemendo »

tomwphillips wrote: post #222
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…

Hey Simon, author here (and reader of your blog!). I used to share your view, but what changed my mind was reading Hao's book. I don't have it to hand, but if my memory serves, she writes about a community in Chile opposing Google building a data centre in their city. The city already suffers from drought, and the data centre, acccording to Google's own assessment, would abstract ~169 litres of water a second from lo…

None of which have to do with AI or AGI.

Nestle is and has been 10000x worse for global water security than all other companies and countries combined because nobody in the value chain cares about someone else’s aquifer.

It’s a social-economic problem of externalities being ignored , which transcends any narrow technological use case.

What you describe has been true for all exported manufacturing forever.

Re: AGI fantasy is a blocker to actual engineering

#237

Post by snowwrestler »

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…

Water and power are local issues. And data center use of water and power is already, currently having local impact on politics. I saw ads about it during the election cycle that just concluded. Candidates had to answer questions about it at debates and in interviews.

People are using these arguments for the simple reason that they demonstrably resonate with average people who live near data centers.

They probably don’t resonate with people who have plenty of income and/or do not compete with data centers locally for resources.

Re: AGI fantasy is a blocker to actual engineering

#238

Post by derektank »

abathur wrote: post #174
simonw wrote: post #134

Earlier quoted context omitted.

Also from that article: > U.S. data centers consume 449 million gallons of water per day and 163.7 billion gallons annually (as of 2021). Sounds bad! Now let's compare that to agriculture. USGS 2015 report: https://pubs.usgs.gov/fs/2018/3035/fs20183035.pdf has irrigation at 118 billion gallons per day - that's 43,070 billion gallons per year. 163.7 billion / 43,070 billion * 100 = 0.38 - less than half a percentage p…

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…

The vast majority of water in agriculture goes to satisfy our taste buds, not nourish our bodies. Feed crops like alalfa consume huge amounts of water in the desert southwest but the desert climate makes it a great place to grow and people have an insatiable demand for cattle products.

We could feed the world with far less water consumption if we opted not to eat meat. Instead, we let people make purchasing decisions for themselves. I'm not sure why we should take a different approach when making decisions about compute.

Re: AGI fantasy is a blocker to actual engineering

#239

Post by macleginn »

I think he got it backwards. Whisper, incredible things chatbots can do with machine translation and controlled text generation, unbelievably useful code-generation capabilities (if enjoyed responsibly), new heights in general and scientific question answering, etc. AI as a set of tools is just great already, and users have access to it at a very low cost because these people passionately believe in weirdly improbable scenarious and their belief is infectious enough for some other people to give them enough money for capex and for yet other people to work 996 if not worse to push their models forward.

To put it another way, there were many talented people and lots of compute already before the AI craze really took off in early 2020s, and tell me, what magical things were they doing instead?

Re: AGI fantasy is a blocker to actual engineering

#240

Post by thewebguyd »

slightwinder wrote: post #187
randallsquared wrote: post #123

Earlier quoted context omitted.

The problem is more one of scale: a million liters of water is less than half of a single Olympic-sized swimming pool. A single acre of alfalfa typically requires 4.9 - 7.6 million liters a year for irrigation. Also, it's pretty easy to recycle the data center water, since it just has to cool and be sent back, but the irrigation water is lost to transpiration and the recycling-by-weather process. So, even if there's…

> The problem is more one of scale: I think the source is the bigger problem. If they take the water from sources which are already scarce, the impact will be harsh. There probably wouldn't be any complaints if they would use sewerage or saltwater from the ocean. > Also, it's pretty easy to recycle the data center water, since it just has to cool Cooling and returning the water is not always that simple. I don't know…

> If they take the water from sources which are already scarce, the impact will be harsh.

Surprised I had to scroll down this far to see this mentioned.

The water use argument is highly local and depends on where we are building these data centers. Are you building in the great lakes region with plenty of fresh water and no water scarcity issues (yet)? Ok fine.

But we aren't building there. We're building in Arizona, Nevada, Nebraska, Iowa putting further stress in an area that water scarcity is already an issue, or soon going to become one due to long term drought conditions. Or Texas, which already has problems with their power grid.

We're building in these locations because they're cheap. If we're going do to this, we need to not let the bottom line be the sole driving decision of data center locations. If it's not profitable to build elsewhere, don't build it until you've figured out how to make it efficient enough to where it is profitable.

Post reply on HN