I think there are still many other barriers to entry, but yes...

...unless they were able to lock it down as IP...

Or if there's a massive breakthru in training efficiency, which is actually fairly plausible given how insanely inefficient it is currently.

NVDA doesn't capture that much of the upside (longer-term) if the monopoly can't be maintained for inference side and training hits a plateau. Or (medium-term) if bubble pops and already-procured chips become available cheap.

But the wrap can be a lot newer than the truck. Seeing new colors doesn't nec mean those are new trucks.

The colors are wraps. Tesla sells some in California, but most are 3rd-party.

Not heated, apparently, but inserting a heatpad should be easy enough:
desknest.com

Saw an ad for one just recently, actually...

"Technological evolution is fast, efficient, accelerating and better by design."

Is it maybe that line in particular that you've now lost confidence in? Here in the age of generative slop?

Sure.

Or else, what?

The MSM has been entirely bought out by authoritarians, so they will do no such thing, you need to stop waiting for that.

It's not the baseline norm, it's still a sick minority seizing levers of power, as it has always been. Urgent to be reckoned with, but don't cede what they have no claim to.

I, for one, would be there for it. 👍

Like, yeah, I suppose you can help more shrimp lives directly with a dollar than you can human lives. But shouldn't you maybe consider the higher-order impacts of things like Rhodes scholarships and MacArthur grants?

The EA crowd tries to do that much, but again I think they're not considering enough bits. At least from what I've ever seen, they're just comparing the direct cost effectiveness of a marginal donation to one predefined charity vs. another, without considering agency and iterated returns.

Quantification is an important part of that, but instead of just probability, what you usually should be considering is cost/benefit or risk/reward.