The AI dooooooom thread

[Tamas]

Cue more global warming?

Marketing materials are not going to generate themselves, not unless you can host all these AI scripts.

[DGuller]

I don't understand how AI data centers are going to create tons of jobs.

There are things more important than jobs, regardless of whether it will or will not generate jobs.  I know that this forum is full of skeptics of AI, but I would make a strong argument that it's a skepticism out of fear rather than out of information.  AI is advancing at a frightening speed, and the only thing scarier than advanced AI is advanced AI that China or Russia got to first.

[The Minsky Moment]

AI data centers will "create" jobs in the sense that the structure of employment will change. Because most of the world outside Africa is transitioning to steady state or negative population growth, the total number of jobs in much of the world - including ALL of the developing world and nearly all the mid-income countries - cannot increase and must decrease in the long-term, barring massive immigration from the few demographic growth countries left (highly unlikely in the present political configuration).

The real question is what will work look like in a world dominated by mid-21st century AI and quantum computing tech? Will there be incremental change or a wrenching paradigm shift.  I still think the former is more likely but it's too early to tell.

[Valmy]

I don't understand how AI data centers are going to create tons of jobs.

There are things more important than jobs, regardless of whether it will or will not generate jobs.  I know that this forum is full of skeptics of AI, but I would make a strong argument that it's a skepticism out of fear rather than out of information.  AI is advancing at a frightening speed, and the only thing scarier than advanced AI is advanced AI that China or Russia got to first.

Well that's a different argument.

A big AI data center is in my town, what sort of job can I get there?

[DGuller]

AI data centers will "create" jobs in the sense that the structure of employment will change. Because most of the world outside Africa is transitioning to steady state or negative population growth, the total number of jobs in much of the world - including ALL of the developing world and nearly all the mid-income countries - cannot increase and must decrease in the long-term, barring massive immigration from the few demographic growth countries left (highly unlikely in the present political configuration).

The real question is what will work look like in a world dominated by mid-21st century AI and quantum computing tech? Will there be incremental change or a wrenching paradigm shift.  I still think the former is more likely but it's too early to tell.

I'm actually probably more afraid of what transformation AI will bring than the technophobes here, because I have higher expectations of what AI can accomplish.  I think the worries about jobs are quaint, I think we have much bigger issues to fear.

My main worry is that AI will be the totalitarian government's wet dream.  It will make obsolete the whole notion that "the government doesn't have resources to listen to all of our phones all the time", because yes, a single AI system would be able to do that.  Not only would it be able to do that, but it would be a single entity, so it would be able to synthesize all the wiretapping of all the people itself, without having to have a system for all the spies to communicate their findings and analysis and agree on it.

Another fear is that AI will bring a single point of failure to many systems that people interact with, which is already a problem but will be a much bigger one.  If you apply for a job and the interviewer for some stupid reason doesn't think you're a good fit, oh, well, there are hundreds of other interviewers out there.  If all the interviewers are using the same AI vendor to make such decisions, then you're just fucked if you're an unlucky residual of that model.

Yet another fear that worries AI researchers is that things will get dangerously unpredictable when AI will become smart enough to work on improving its capabilities.  I don't think anyone can even predict what will come out of it, but obviously there is a lot of potential for very bad outcomes (and very good outcomes too, but it's wise to be risk averse with existential issues).

[Valmy]

Ok just to clarify I was just surprised by a claim that AI data centers are going to bring hundreds of thousands of jobs. Unless we are talking about people just building the physical space, I am not sure what jobs those data centers would generate by themselves. The AI being done there and it's impact on jobs is something else.

[crazy canuck]

And my concerns are exactly the opposite of DGuller.

The obvious flaws in AI are ignored to the detriment of us all and the reliance on AI will be disastrous.

But people like the colour will continue to ignore the data, although they claim to the specialist and analyzing data.

Who cares if AI generates ridiculously obvious errors it's all good right?

[crazy canuck]

As just one example, despite the fact that a lawyer was recently disciplined by our loss society for presenting a brief to our court with phantom cases generated by AI, we are beginning to see more and more communications from lawyers, which are using AI.

How can we be so sure that lawyers are so stupid as to do that? Because we're seeing more and more cases that actually don't exist being cited in their communications.

But hey, label me as a technophobe rather than somebody who actually cares about you know accuracy and reliability.

[Sheilbh]

I don't understand how AI data centers are going to create tons of jobs.

There's two sides I think. There will be jobs created by building and maintaining new physical infrastructure - and all of the material inputs it needs (e.g. with Trump, oil and gas). I think that will apply to data centres. If they're being built all over because AI needs it, then it'll be like any other nationwide infrastructure building boom.

On the AI piece there's vast amounts of money flooding into it. The bet is it will have transformative impacts on productivity (reducing the number of employees needed) - basically like robots and mechanisation did to blue collar workers but to the white collar workers. That will obviously have huge social and economic consequences (and is why the tech bros love both AI and UBI). My instinct is that's right and it's a hugely transformative technology. Of course that might be wrong - it might not be that technology. In which case we've had huge waste (for now) of capital which will cause a lot of economic shocks as it unwinds. Those seem to me to be the two alternatives.

Of course the other background to this is that China is also working on all of this technology and also implementing them. So I think, if you believe there is competition with China (or technically advanced authoritarian states more generally), then I don't think the idea of there being a route where we can park this holds up.

My own view - having looked at a lot of AI for work - is there's discrete areas where it's clearly already incredibly useful. Particularly around code, data designers, scientists, engineers etc. The other areas are still, at this stage, a little inchoate in my experience. I think that is why, for example, Google are just looking at releasing it and integrating it into loads of tools (at the minute at no or low cost) to see where it works and how from which they'll build products. So I think on the one hand we need to move the hype from "this could destroy us all" to "at the minute it's an extension on Software-as-a-Service tools" while keeping in mind we're only just starting to explore how it can be used.

I think Google's Notebook LM (which is fantastic) is a really interesting example. They released that to consumers - it's got great features (like citations etc). I think they were surprised by how positive the reaction was and businesses coming to them to say they'd love an enterprise version with enterprise T&C protection. They're now launching that (or have just launched it). But I don't think they expected it to be a big hit until it was rolled out and people could see how they could use it. I think we'll see a lot of that - which is why (and I could be totally wrong) I think the platforms with products and tools people are already using like Google will be able to take advantage of it more initially because they'll be able to integrate and see how it's used. I think that's a clearer path than a "general" AI like OpenAI and I suspect they may instead end up becoming basically infastructure for others (like the AWS of AI - which is not a bad place to be).

[Valmy]

I guess. But surely that would be true no matter where the data centers are located? I guess I thought it was important we build the data centers here so we get sweet data center jobs.

But I guess it is just jobs created building buildings and power production.

[crazy canuck]

So a couple of months of employment for the construction trades (it doesn't take long to pop one of those up) and not much if any increase in employment for the source of energy. The number of employees needed to supply energy is also decreasing due to technological advances.

It reminds me of the times politicians say that pipelines create jobs. Well, for a short period of time during pipeline construction.  But the whole purpose of a pipeline is to transport something at a greater volume and lesser cost overall, and that includes labour costs.

But as JR said, that is not the issue at all.  The question is the impact the reliance on AI will have.

[Valmy]

But as JR said, that is not the issue at all.  The question is the impact the reliance on AI will have.

Right. Obviously that is the biggest issue.

Just when somebody says something like "we will build the data centers here and create hundreds of thousands of jobs" I wonder if there is something I am missing. I don't think a huge staff is required in a data center.

[DGuller]

I guess. But surely that would be true no matter where the data centers are located? I guess I thought it was important we build the data centers here so we get sweet data center jobs.

But I guess it is just jobs created building buildings and power production.

If it helps, "create jobs" reasoning is total bullshit 100% of the time, AI or not.

[Sheilbh]

Just when somebody says something like "we will build the data centers here and create hundreds of thousands of jobs" I wonder if there is something I am missing. I don't think a huge staff is required in a data center.

It's not.

It's a modern version of building a road or, say, a bridge to nowhere. Again if the US decides it's going to build hundreds or thousands of data centres around the country with also new power infrastructure then that construction will employ lots of people, for the duration of that build. Each will also employ a few people in the running.

But I think that's broadly true of all infrastructure investment projects, no? We had something like 75,000 people working on the construction of Crossrail. In that sense it creates jobs. It only needs (I'd guess) a few hundred or thousand working on the platforms and trains (which I think are driverless) once it's actually built. The main benefit is whether it's valuable afterwards - a bridge to nowhere not, a Tube line with 200 million + journeys a year is. Data centres (and associated power infrastructure) to support the adoption of AI...the jury's out.

Edit: I'd add that I suppose Trump using that language doesn't surprise me. I imagine Trump towers (like all construction projects) talk about the number of jobs created while it's being built, not the number of people actually needed/employed by the building afterwards.

[Valmy]

If I was an ordinary Joe and I heard something called a "data center" was going to employ people I would think it would be like a manufacturing facility or something where lots of workers would end up working there. It is very misleading to use that to mean there will be jobs in construction and utility industries where there is not exactly a lack of employment in already.