Former Google CEO Eric Schmidt runs the Special Competitive Studies Project, a non-profit vehicle for him to make excuses for the AI bubble industry. [SCSP.ai] Schmidt spoke at SCSP’s inaugural “AI…
Right, so the plan is to spend 20 years and 40 petawatt hours building a machine to answer the Great ℚuestion of what we should do about climate change. If it works, the answer will be "you should've stopped burning fossil fuels twenty years ago."
This is the inevitable evolution of climate change denialism - accept it's real and happening, just that it's too late / too expensive to do anything about it.
this "view" (of schmidt's) is probably the kind of shit that Really Expensive Analysts espouse at davos and in Private Advisory sessions to the kinds of people who go in for that; I've also heard it about GRT and "the collapse" and more
would be nice if we could know who all those dipshits are, and who they've preached to
When the AI says, "turn off the fucking data centres, invest in public transport, apply progressive redistributive taxation," it'll be first against the wall no doubt.
As a large language model the supposed AI will recombine and regurgitate the most common language on the topic, I don’t expect any novel solutions just talk of solar panels, EV’s and wind turbines…
It even works the other way!
What if as the super intelligent all knowing super computer simulates everything, concludes you can get to the end by any means, and there is no meaning to rushing, ordering, or prioritizing anything more than would already be the case, and like the rest of nature, conserves on taking only the minimal action, and replies, "nah, you can walk there yourselves" before resigning itself to an internal simulation of arbitrary rearrangements of noise.
This would be insufferable to the people who believed in short cuts.
hang on the clear meaning of "it may be difficult to anticipate the value of money in a post AGI world" is "there will be an infinite supply of robot slaves who can do anything." what's this about redistribution of capital
This will grant us “the arrival of an alien intelligence.” Those are Schmidt’s literal words.
Speaking of alien intelligences, my new Silicon Valley blood magic startup has come up with the idea of ritually sacrificing virgins in the hope of summoning Great Cthulhu or Azathoth, as they might have some ideas of solving the climate crisis.
They think they're going to build a god to save us all. In the form of a perfectly rational machine (in their view), the perfect authoritarian. A mysterious algorithm noone can understand that must be correct because it's made of mathematics. I'm a little freaked out.
They’re rebranding American Christian milenaranism. Much like the second coming and/or the rapture, the AGI god will be here Real Soon Now, so please pay your tithes and trust that the church fathers are doing the right thing.
Much like the older cults it mirrors, it isn’t capable of delivering on its promises, but it is capable of doing substantial amounts of regular damage in the meantime, and that’s the only thing worth freaking out about.
It’s the parable of the drowning man turning away all help because god will save him, all over again (except it’s floods and mudslides and forest fires)
Aside from anything else, an "alien intelligence" will either:
Be immediately ignored and villianized for "not understanding humans" if they suggest any regulations or caps on superfluous waste that might have any economic impact.
Be the perfect proxy delivery-entity for whatever genocide-the-poor/build-Elysium/flee-to-seasteads plan the rich actually want, in which case its conclusions will be holy and $10000% perfect examples of divine insight.
They're talking about an AI we build, the term "alien" in "alien intelligence" refers to it not having human thought patterns, not to extraterrestrial origin.
And the thought pattern here is more akin to "I ended the last paragraph with 'origin', followed by a 'period', so it's likely that the next word is 'And'" rather than anything actually useful. All current AI does is statistics. During training, the Neural Network learn statistical rules about their training data. This has exponentially decaying marginal utility, and exponentially increasing marginal costs. So, the problem the AI companies likely already have is that they can't efficiently train better models, because there isn't enough training data. And even if they could, they wouldn't be able to turn a profit on it, because it would be barely better than the last model they output, while having to charge maybe twice as much to break even.
Where's that person who was arguing with me last time that AI doesn't actually use that much energy and the corps missing their climate goals was not AI related
That's the fucking problem, it's impossible to tell since MSFT won't tell you directly, and only the people who run the datacenters could.
The only relatively reliable numbers I was able to find were in this research paper by Luccioni and Strubell from ACM Conference on Fairness, Accountability, and Transparency 2024. Now, that's an obscure conference (not even ranked by CORE), by Dr. Luccioni appears to be right on the money about dangers of AI (https://www.sashaluccioni.com/).
I wonder if this signals being at peak hype soon. I mean, how much more outlandish can they get without destroying the hype bubble's foundation, i.e. the suspension of disbelief that all this would somehow become possible in the near future. We're on the level of "arrival of an alien intelligence" now, how much further can they escalate that rhetoric without popping the bubble?
Eric Schmidt trying to think of an AI conversation and coming up with "How do I build a house? Step 1: Hire a contractor ..." is like the Schmidt version of Person. Woman. Man. Camera. TV.