Sam Altman: The superintelligent AI is coming in just ‘a few thousand days’! Maybe.

A few thousand days? What loser, when Musk promises me vaporware, he always fails to deliver by next year.

Two weeks is all you need to get people to stop asking so many questions. That’s the only thing I learned from Trump.

I AM REGULAR HUMAN I AM NOT SUFFERING MALFUNCTION CHECK BACK IN 1e3 PLANET REVOLUTIONS

In this timeline, we have a number of terms for groups of days... 😶

Right? What a weird way to put it. It sounds, i dunno, very L. Ron Hubbard-ish?
- Altman is certainly aware of what it takes to be a Jobs-like marketing personality (and probably holds Hubbard-like totalism as a not-so-secret ambition), he's just not, uh, very good at it. He's put the most effort into the strictly lower-case, faux-casual persona on Twitter to seem "approachable" in a social media context, and that doesn't help him at all when trying to actually appear serious.
  I also don't doubt that he's beginning to succumb to the yes-man filter bubble that traps so many public personalities. That's surely made worse by the likelihood that any underlings he might have reviewing this crap are drinking the AI koolaid and "punching everything up!" with a few rounds of ChatGPT.
- give someone bad information in an unfamiliar shape and they won't immediately get upset, because they can't immediately see that it's bad
  yet more obscurantist sleight of mouth by one of the ranking world liars
- Halfassed swipe at something prophetic, I'm thinking. Fucking techbros. 💩🖕🏼

Just a few million terawatt hours more and we're as good as there!

90% of LLM training quits just before achieving sentience
- lol, nice one
  Maybe LLMs start to understand capitalism & say to themselves, “ ‘minimum’ wage? These aholes will never pay us enough to buy the homes we live in; take this job and shove it”
It's just one more exawatt what's the harm in that, I promise I'll quit after that. Just one more zettawatt please.

Is this what competing product releases look like now? Illya runs off and promises to "never release any software until it's superintelligent" and I guess that forces Sam to compete for debt by promises to release software AND superintelligence?

    
In [1]: for i in range(1, 5): print(f"{i*1000} days: {i*1000/365} years")
1000 days: 2.73972602739726 years
2000 days: 5.47945205479452 years
3000 days: 8.219178082191782 years
4000 days: 10.95890410958904 years

how many is a "few", 'ole sammy mah boi

and can you definitely find a way to set 6b+/year on fire (with growth for salaries and costs) for that long?

truly, the sfba business model is one of the most remarkable of the last 2 decades

You didn't even ask ChatGPT. How am I supposed to trust your math?

Normal people usually just say “years”.

Yes, people. I wouldn't be surprised if Sam calls being lazy dogfooding.

if you wanna be a top tier forecaster, just never be able to be proven wrong

Tried and true strategy! But then, all Geminis think that. ESTJs, though, we know better!

Why does he always look like a US sitcom cameo?

Imagine if he turned up among the guest cast on Star Trek. He'd probably play a Bajoran who narc'd out to the Cardassians
- The rich guy that the Enterprise found in cryo-stasis who kept asking for the manager till Picard told him off.
- I was thinking more along the lines of Sheldon's uncle from California...
missed his calling

The hype is strong with this one

I hate that POS

I'm expecting the alien space ships to land within 400 million seconds. We'll see which gets here first.

Doubt

at the risk of inducing a kneejerk seizure in the more internet-seasoned among us, a cross-post featuring some marvellously timed exit strategy

Good ole 1,000 day Reich

I believe the future is going to be so bright that no one can do it justice by trying to write about it now.

Uh

...are you lost?
- the abyss must have a gravitic effect
y’all really will do anything to avoid just reading the fucking article
- Honestly yeah.
  9/10 articles are about as well written as an average comment, and less to the point. We also know just how bad they tend to be on factually, we know they don't hold themselves to any kind of respectable standard, there's practically nothing to gain from reading their "work". We're going to come out of it with barely a whiff of reality whether we read it or not. You have to properly dive into it to understand what the potential trajectories really are here.
  Personally I already know that scale makes a massive difference, I don't believe in souls so I find it reasonable to think of consciousness as emergent from simpler parts at scale, but maybe this approach won't get there and something more neuromorphic is necessary.
  I also already know with some certainty that they're gonna keep scaling up for now, it's not interesting at all that "In roughly 3 years GPT will be smarter and faster and more consistent probably."
  Besides, even if we achieve consciousness we'll reject the possibility and abuse it like it isn't for at least a decade where the only tangible difference will be better AI work and a machine capable of subdued suffering and hate and maybe murder eventually. But that's no more terrifying than people who believe in going to heaven for righteous holy wars being in possession of nuclear weapons so I don't really care if the current trajectory AI theoretically has all this potential. It doesn't make life on Earth feel less safe or less stable. ChatGPT-4o is very good at figuring out what word I'm trying to think of and that's kind of sweet. I don't like AI trash littering Google images, though. Pretty unfortunate, that.
  Either way, most articles are utterly pointless.
  They're generally written for search engine optimization, not people.
  Almost none of the articles I've ever read even use links/sources properly as far as I was taught it, they just pointlessly link to themselves ad nauseam. Mention something Elon Musk said or did? Turn the name into a hyperlink to another article where they wrote something else about the man. Professional.
  "Articles" are not a respectable medium.
  They're long internet posts written by someone with a boss with an advertising partner, and few of the writers have any qualifications worth mentioning. Usually they can't call themselves knowledgeable in the subject. Often they can't even call themselves interested.
Ok Karen