← Back to context

Comment by atleastoptimal

15 hours ago

The guy from SemiAnalysis said this too. If he buys Anthropic now at $2T and they go to zero, he's fucked, but if they go to $20T, we're all fucked.

They're not really a "good investment" unless they are literally worth more than combined GDP of the entire planet, multiple times over. I don't think I'm buying shares of Weyland-Yutani here, but they sure are acting like it.

  • Why wouldn't the same apply to Spacex? That one sold and is holding it's IPO price.

    • It's down 8% over past 6 months while the S&P500 is up 7%. It's definitely so far been not a great investment, unless you managed to get in and out during that brief huge upswing

      6 replies →

    • If I understand correctly, the argument isn't that it can't _hold_ it's price, but that it's not worth investing in if the goal is _meaningful gains_.

    • The fact its price is only holding, as opposed to having maintained its initial growth, is an indicator (for me anyway) that the bubble is pretty close to its maximum size, and it could only take very few more breaths, cough Anthropic and OpenAI cough, to reach critical mass (mixing my metaphors).

      But, as they say, bears sound smart, bulls make money.

    • As others said, the market is generally unhinged currently, with enough extremely wealthy people needing somewhere to put money that doesn't really matter to them, and I suspect some level of "investing in SpaceX public stock" is done to appease/appeal-to Elon.

If they grow to 20T then buying the index will also be a good investment but with a good risk hedge.

  • Isn't it likely that inference providers like Anthropic continue to capture more and more of the share of labor as revenue? Everything up makes sense for a year or two as we automate more and more labor, but at some point unemployment will drag the whole market down. It's difficult to see it working out any other way, unless we come up with something like a token tax.

    • > Isn't it likely that inference providers like Anthropic continue to capture more and more of the share of labor as revenue?

      Not if open weight models turn inference into a commodity that's sold at cost.

    • They will capture share of labour as the labour works today.

      There will be, as is always the case, some amount of new labour skills that emerge and I can't say for sure what percentage or amount that might be.

      Looking at tech work being increasingly well the last 20-30y, we can imagine new ways of working might be pretty valued.

      1 reply →

Good point - in a way, if we get AGI, investors have financed the way to their potential undoing.

The term singularity is worn out, but the physics parallel is quite apt - a point where the known gears of economics stop working, and it's unknowable what's there until you get there.

What if, here me out, you got to share in the profits of the companies you buy shares in? Let's call those payments "dividends". Theb it doesn't matter how much shares go up in value.

> if they go to $20T, we're all fucked

Maybe I’m naive but I expect any kind of “we’re just taking over the world now” event with any AI company will be met with enough resistance from 99.99999% of the world’s population that they’ll have to make it a sweet deal for us if they want to survive.

  • Not if they start making killbots.

    • Our killbot manufacturing process needs 50% of the oxygen in the earth's atmosphere. Thankfully, unlike GPUs and RAM, which we needed to purchase from their manufacturers, we can extract oxygen out of the air at outrageous volumes and minimal cost. This effort has already begun, therefore by the time you read this, you are unlikely to have the ability to disagree, much less protest. And if you do somehow manage to protest, say hi to one of our many killbots, who are likely to also be in attendance, en masse.

    • Actually we can do classical theory of strategic interaction here. There will be a range of acceptable negotiated settlements to both parties since both parties have some estimate E of the expected balance of power resulting after a war. If A is the amount of surplus value Anthropic expects to lose in a war and H is expected amount of value humanity expects to lose in a war, the the range [max(E - H, E - A), min (E + H, E + A)] gives a whole continuum of mutually acceptable settlements.

      Now there will be bargaining friction given that any power sharing agreement with Anthropic will likely increase their future capacity and thus the future expected balance of power, but maybe AI has a long enough utility horizon that we should all expect to be dead by the time it realizes the long-term expectation advantage

  • > resistance from 99.99999% of the world’s population that they’ll have to make it a sweet deal for us if they want to survive.

    The "dark-enlightenment" billionaires have been planning for a "collapse" for years, and their plans involve bunkers, acquiring citizenship in remote places, pervasive surveillance and armed dogbots/drones, the last 2 powered by AI, naturally.

    • I love the bunker building. Where is Dario's bunker?

      Altman already has a couple I think.

      Thiel has gone to Argentina but has a backup in New Zealand (w/ Citizenship).

      Zuckerberg just bought an Irish castle (a Medieval bunker of sorts) but probably has others.

      Musk probably has plans for a Moon bunker complete with Pleasure-Dome. Investor funded of course. Not sure how cosmic rays affect sperm however.

      Brin couild opt for Blimps - imagine floating above the reeks and the wrecks like a Meth from Altered Carbon.

      Bezos - does he have a bunker?

      2 replies →

    • Sadly true, even to the point of paying for not only the pilot to fly their helicopter to the island in the inevitable collapse, but also the pilot's family. Not as a token of generosity of course, no, because otherwise the pilot wouldn't fly them.

      2 replies →

the technology isn't the company yeah? maybe AI will take over the world but it's unlikely to be a single firm