thrwayaistartup

I think that ascribes way too much meaning to hallucinations, which are the artifact of a big fancy markov chain doing what you'd expect a big fancy markov chain to do.

Looking back in 25 years, the "Hallucination Problem" will sound a lot like the "Frame Problem" of the 1970s.

Looking back, it's a bit absurd to say that GOFAI would've got to AGI if only the Frame Problem could be solved. But the important point is why that sounds so absurd.

It doesn't sound absurd because we found out that the frame problem can't be solved; that's beside the point.

It also doesn't sound absurd because we found out that solving the frame problem isn't the key to GOFAI-based AGI. That's also beside the point.

It sounds absurd because the conjecture itself is... just funny. It's almost goofy, looking back, how people thought about AGI.

Hallucination is the Frame Problem of the 2023 AI Summer. Looking back from the other side of the next Winter, the whole thing will seem a bit goofy.

Even in the parts of the US with 203X goals, many of the statutory "EV" definitions include vehicles with large gas tanks. E.g., the Toyota RAV4 Plugin Hybrid qualifies for EV tax credits in most states with such credits.

Yes, that's the problem with US political-economics right now. You've (failed to?) identify the "political" part of "political-economics".

There are a ton of software development jobs in those sectors! Especially if you include all the "software" jobs that are less like the sort of programming at a typical software shop and more like "Math/CS/Eng PhDs writing MATLAB code".

I think we will see that transition throughout the economy, actually: software will continue to be very important, but the marginal value add of someone writing software will depend on their having deep domain expertise.

Montana has a massive amount of coal reserves, several productive oil fields, and a lot of refining capacity. If you close the oil refineries and leave coal in the ground, a lot of communities in Montana are going to become ghost towns. The lives of the children of people in those towns is going to suck -- poverty, drugs, and an early death if you can't get out.

The fact that those jobs are offset by growth in solar and energy jobs doesn't help without a centralized wealth transfer, which ain't gonna happen.

The problem with US political-economics right now is that the future of the country depends on making a series of "society wins, some regions win, some regions lose (win-win-lose)" decisions. Because the societal win is net, it could be made into a very attractive win-win-win situation. But everyone knows from the last half century of globalization that taking win-win-lose and turning it into win-win-win is a false promise that won't pan out.

I think it's an artifact of (1) being an all-day-meeting-person, and (2) the majority of work-product being present in PPT form.

(1): Taking notes in PowerPoint is sensible if you want an entire conference room to be able to read the notes, as they are being taken, from their seats, and without any fuss about font sizes and so on.

(2): Additionally, since the final "product" will be a PPT, having the notes already in a PPT file means you get to skip some copy/paste steps.

It's still a bit silly and a bit of a cargo cult in the online-first world, but I get how it happens.

> Not OneNote, not Word, not even freakin' Notepad.

Hey, I'll take PPT brain damage over the even worse brain damage that causes some orgs to do everything in a buggy ass Salesforce version of google docs... give me MBA brain-rot bullet points over MBA brain-rot fancy versions of five paragraph essays any god damn day...

I am about to start working on a non-profit project -- not a startup, but similar in terms of resources dedicated to the project and how we hope it will scale.

One of our big questions is whether it makes sense to rent or to buy for training/finetuning/RLHF. The advantage of renting is obvious: I don't think that this phase of the project will last very long, and if it turns out that the idea is a success we'll have no problem securing funding for perma-improvement infra.

The possible advantage of buying is that we would then have the hardware available for inference hosting. We do expect some amount of demand in perpetuity. Having that ongoing cost as small as possible would allow us to continue serving the "clients" we KNOW would benefit a lot from our service with minimal recurring revenue.