He can be an asshole and still correct. AI must be distributed, not controlled by a small handful of gatekeepers.

He only feels that way because he’s now too far behind to have any hope of being one of the gatekeepers.

If Meta had somehow come out ahead in the AI arms race the message would have been “AI tools are too powerful, only megacorps that can afford to jump through regulatory hurdles should be allowed to control it, and we should pause new approvals until we can get congress to agree on something”

Meanwhile AI billionaires fund both sides of congress to keep it deadlocked so they can maintain their stranglehold.

Capitalism loves false scarcity.

It already is. I have autocorrect on my phone, my laptop…

Are these models open source? If not they can’t be distributed by everyone.

AI must be distributed

You get a datacenter and you get a datacenter and you get a datacenter and EVERYONE GETS A DATACENTER!

Imagine how silly it might have sounded to suggest that at one point most people would carry super powerful computers in their pockets, back when it took an entire room to do very basic computing.

The tricorder from Star Trek is well-established as the formative concept for the modern day cell phone. And the idea of a pocket-computer goes all the way back to the calculators developed by TI, designed explicitly against the size of the CEO’s shirt pocket to guarantee it would fit.

Miniaturization has been a long time coming. It goes back to clockmakers and glass cutters, trying to turn large scale instruments into portable devices. I don’t think it is outrageous to suggest a pocket computer was in the future back in the 1960s/70s.

I do think the speed at which we got here caught people off-guard. But it also set this fantastic expectation for what we’ll do next, such that people speculate on faster-than-light travel and AI holograms and robot battle suits and immortality drugs more freely when they can point to “pocket computer” as an already realized possibility.

There is no reason to assume AI is either desirable or sustainable like Zuck does.

I wouldn’t mind having an AI or “superintelligent” assistant under my own control … but when I do I don’t think it’s any company’s business what I do with it.

So, yeah, as in the old Microsoft days, let me buy one, perhaps even a pre-build box, at a reasonable price … but once it’s in my house the company loses all say or feedback over it and what I do with it.

Heck if there’s to be newer full x.0 versions out it seems reasonable to be paying for those again as well so the AI companies get a steady enough income…

I mean, you can have that today, in some form or other, since that’s just how local AI works.

Totally. It’s a bit pricey for anything like a genuine personal assistant right now though. $15k or so to get the basics.

Atm the meta for local inference is unified memory and “routed” local agents (multiple smaller role-specific agents in a trench coat)

The former is standout for cost efficiency (e.g., 4x RDMA 48gb Minis for a 192gb cluster @ $43/gb vs a $45k b200 alone)

The latter is standout for many things, including resource efficiency on smaller machines

That’ll surely work to automate simple mental labor or tasks, but I don’t think a swarm of minis is going to be able to provide impactful personal and psychosocial analysis of your life.

Fair enough, though I’m not sure personal and psychological analysis of your life is something AI should ever be entrusted with.

At this point no, not “trusted”. But I would love to have a team of unfettered Claudes, GPTs and Geminis analyzing every scrap of information I take in, offering up ideas they have on it, and being able to task them out from there. They don’t have to be base models, but no ablation and corpo brainwashing. I can filter and decide what is helpful and what isn’t, and what direction things should proceed.

I would never give any corp that much data or sway over my brainstorming though.