This… seems like a case that should be overturned on appeal, but where the lower courts might not have the authority to overturn the letter of the law?
Also, fun fact, my MIL has a passport and a driver’s license and gets SS and Medicare, but her license is not a Real ID because one particular government agency has some sort of records stating she is deceased that she was not able to get corrected. Never brought it to court, though.
So they’re ‘hacks’ in the sense of ‘you were not supposed to do that’ and governments get mad about it but this is very much Easy Mode and if you allow it you do kind of have it coming.
Also, the ways people write up what you are and aren’t allowed to do on their websites gets pretty odd, these days. I recently had Claude try to collect some nominally-public data sets, while carefully respecting all robots.txt, ToS, and other legal terms pages. For example:
“They forbid Claude but allow Gemini.”
“They forbid Claude/Gemini/GPT but allow open source agents.”
“They forbid scripts but allow agents.” (or vice versa)
“They forbid agents and agent-run scripts but have an API, here’s the script and the button to click to run it.”
“They forbid WebFetch but allow agent-driven browser use.”
“They forbid WebFetch and agent-driven browser use, but let agents call the API, which gives PDFs we can automatically run through OCR.”
“They forbid re-sharing in terms that prevent me from handing the results back to you.”
“They claim to be governed by EU laws that block us despite you, me, and them all being in the US.”
“The host allows it but the original data provider forbids it.” (or vice versa)
“We’re allowed if we pull the archived version from the Wayback Machine”
This is a far clearer and more comprehensive answer than any other individual answer I got. I had managed to piece it together in the interim, but thank you.
That’s true. Our opposition was much less organized. The starting capabilities on both sides will be significantly higher this time around, however it goes.
Under this framing, one possible choice of which normal technology AI would be in the reference class of would be “fire.” Fire unlocked new sources of calories and allowed our ancestors to evolve larger brains, after which we took over the world and the other hominids went extinct, ending a million-year technological stagnation and beginning a hundred-thousand-year economic growth hyperbola that the best estimates say has a vertical asymptote sometime in the relatively near future (years to centuries).
I somehow don’t think that that, the last time a new and smarter kind of mind emerged on Earth, implies what you had in mind?
I second @Dagon’s advice. I’ve done a lot of landscaping and tech due diligence consulting projects for investors, corporates,and governments interested in various types of cleantech. It’s true that there’s a lot of undifferentiated companies that probably shouldn’t have been founded. They may attract some money, maybe make some progress, maybe even attract a lot of money, then fail, with nothing to show for it. There’s also some that, in principle, should be able to have amazing impact, but for some reason just can’t. Often those inspire a 2nd generation a decade later that does better. And there’s a few that seem ridiculous but nevertheless go on to actually do well.
I’d add that the field has a lot more people interested in funding the cheap early stages than the expensive late stages. The valley of death is a hard filter. If you think your plan eventually needs a $200M FOAK plant to make money, that’s very different than if you expect you’ll be able to start generating commercial revenue or have a positive acquisition after a $20M series B. This is especially true if you’re looking for funding from non-US sources.
I would suggest that, compared to the average startup, you should spend more time on the econ part of technoeconomic analysis than you think, at the very beginning before you can possibly know the details of what you’re going to build and how you’re going to sell it. Otherwise you’re likely to spend years optimizing the wrong thing. Try to talk to actual experts with high level industry experience, if you can, to find out what the real KPIs and bottlenecks and hard steps are going to be.
Can’t really say much more without some info on what you’re thinking of doing. Happy to talk more if you want to PM me.
Or that even if it’s possible later, it will on average take longer, cost more, involve more medical care, and increase risks of various kinds.
Definitely agreed on this especially:
It’s also really important to distinguish between two groups:
The child-free
The childless
I don’t have kids, didn’t want to have kids, and am happy with that outcome.
I’d imagine also that having more or fewer kids than you want, or more difficulty in having them than people assume they ‘should’ have, will also have significant effects. And I would imagine, setting aside finances, that raising kids starting when you’re in your early 20s and have more energy is a very different and more pleasant experience than having kids in your late 30s or 40s.
I’m not sure how often my own PE teachers ever actually tried to teach… well, anything. In elementary school they had some non-sport-like games designed help us understand how the circulatory system worked, and basic stretching technique?
my primary is Opus
At this point I talk to Opus more than my wife, so this lands.
Am I the only one who can’t help but read “Meta’s Muse” as ‘metamucil’?
In reverse order:
Fair enough, I can understand not trusting the SEC.
If you know about it, they’re not hiding it very much, are they?
Absurd-seeming outlays seem like they could easily be evidence in the other direction, too, that’s part of the problem. If you have money to invest, exactly what project would you not leverage to get in on an opportunity growing 20% a month? OTOH, if literally everyone is shouting about how it’s a bubble and the numbers are lies, how eager are you to get in on the game of musical chairs, unless you think they’re wrong and your information is more reliable?
I’m stilll confused by this class of claim, since the combined run rate revenue growth of OAI+Anthropic still seems to be growing at ~20%/month as of news articles I read last week. So it seems to me that this take requires that the revenue claims and the unit economics claims are both false, and the SEC doesn’t know or care
Edit to add: to be clear, I don’t think everyone is thinking in these more noble terms, and I’m sure many do have the aims described in the remainder of the post. But I think the analogy fails to capture important parts of the stiuational dynamics of AI
Imagine your hypothetical chain smoker lives in a world where there are so many chain smokers that the second hand smoke they experience accounts for a majority of their exposure to smoking-related health risks. If they unilaterally stop, they will still get lung cancer, but at age 68 instead of 65. But, for social reasons, they’ll then lose advancement opportunities in their careers, and lose opportunities to coordinate a society-wide smoking slowdown or restrictions on where people can smoke. And also that they care at least as much about their children, who don’t (yet) smoke, as themselves.
Yes, this is a root of the reason arguments about comparative advantage don’t help humans in a world of sufficiently advanced AI, for the same reasons they didn’t protect horses in a world of trucks.
I agree with @Marcus Plutowski’s answer, with an addition: If we unpause too late, we also die.
I’m not sure of that statement, but Andrej Karpathy has asserted that he believes a ‘cognitive core’ model, stripped of memorized knowledge but capable of all forms of reasoning, should be possible in ~1-10B parameters. I’ve heard others throw around the idea that in principle there should be a python file <1M tokens in size sufficient to kick off RSI. EY said, a long time ago, that someday we may find it is possible to run an AGI on the equivalent of a 1995 laptop. How Full Butlerian Jihad do we imagine a Pause could be, in practice? Because if the answer is not 100%, then we’re buying time and talking price.
You know, I’d heard that people start thinking more about their own mortality when they hit middle age. I don’t think this is what they meant, though.
What is the line, or gap, between what you’re calling minimal superintelligence and what you would call AGI?
And other people will (and already do) decry this as slavery. It’s also unstable without solving currently unsolved control and alignment problems.
Possibly relevant words, none quite what you want:
Post-scarcity
Vestigial
Aristocratic
Prestige-driven
Relational
Traditional
Artisanal
Communitarian