241–250 of 340 posts
Honestly, who likes any hype in anything ever? Especially if you genuinely like and understand the thing being hyped.
This line: "this is my main argument against the valuation of frontier labs. It’s not that AI won’t create that much value, it’s that they won’t capture it." That is a very astute and concise way to explain everything about how the frontier labs are behaving and how they're trying to push more people to pay token rates for the best models. At the current subscription prices ($100 or $200 a month for a generous, thoug…
In 5-10 years an Apple Watch will run a Fable level model locally. I don’t think we (hackers) should worry too much about token cost inflation. The current wave of providers, that’s another story.
Earlier quoted context omitted.
No, it won't. We moved about order of magnitude that from 1990 to 2000. The thing is, it needs demand to drive it. Laptops have been roughly the same spec for the last 10 years because we don't need them to be bigger; there's no demand for a 16Tb RAM laptop because we don't have anything that could possible need that much RAM. Until LLMs came along, and we all want to run them locally, and so now there is a market fo…
Right but 2000 to 2010 didn't have similar progress, and especially 2010 to 2020 didn't. Sure, things have gotten better but not as much as the 1990 to 2000 leaps. And yes, laptop specs haven't changed much and this is partially because the need for spec changes wasn't present, but also during the last 20 years there has been tremendous pressure for efficiency in datacenters. Despite that, dennard scaling is dead sin…
I had a friend working in optical computing back in the late 80's that would wax lyrical about how optical computing was vastly superior to silicon back then. But it never took over because silicon worked well enough.
If we've hit the limits of silicon then there are other options. We would need to reinvent huge chunks of our tech stack, and that is incredibly expensive, but if the demand is there, we'll do it. The demand has never been there.
This line: "this is my main argument against the valuation of frontier labs. It’s not that AI won’t create that much value, it’s that they won’t capture it." That is a very astute and concise way to explain everything about how the frontier labs are behaving and how they're trying to push more people to pay token rates for the best models. At the current subscription prices ($100 or $200 a month for a generous, thoug…
Best comparison is Anthropic/OpenAI are AOL/Prodigy. Massive market capture, no moat. Little by little, the convenience and weight will be scraped off, but they (probably) won't roll over and die for quite a while. By the same measure, NVDA is Cisco, providing the backbone and capturing a ton of the early benefits, but soon becomes furniture while the excitement moves further up the chain.
I think we can start to see the outlines of this happening with LLMs as well. Local models have gone from being proofs of concept to something that is at least remotely comparable to contemporary models, and the gap there is closing faster than SOTA models are pushing it forward. Local models still suffer from high hardware requirements, but so did early image gen models where typical consumer hardware was insufficient for efficiently running them.
Earlier quoted context omitted.
Who is going to end up capturing all this value being generated is going to be very interesting. Back in 1980, who’d have thought MS would capture the majority of the value from PCs over the next 3 decades, and not IBM?
Was ms even making that much money compared to actual hardware manufacturers? Ms is licensing the os sure but I mean most of the spend was going to the actual workstation hardware and periphery I’d expect. Including stuff not directly tech like herman miller chairs.
I don't know how to apply those lessons to AI, as long as AI requires so much hardware to operate. If small models actually get capable enough, the shape of the industry changes drastically.
At least for me, the jump in productivity has resulted in building stripped down one-off software for my highly specific use-cases. You can use an LLM to create anything but you still need to know what it is that you're building, and you need to think through how everything should work or the LLM will just fill it with sausage. You can tell that the models are still quite jagged and limited by the mixed quality from…
This doesn't make sense, I enjoy making bread at home but it costs 10x and tastes like dog shit I dont want to spend my time perfecting the craft of making bread for my daily needs (maybe once in a while its a soothing activity), I want someone smarter than me to spend his entire life coming up and perfecting a solution and exerting more time and effort than I can afford and I am very happy to support him so I can st…
Eh? You enjoy making stuff at home that tastes like dog shit? That doesn't make sense at all.
BTW: I love making bread and it tastes amazing!
Earlier quoted context omitted.
Right but 2000 to 2010 didn't have similar progress, and especially 2010 to 2020 didn't. Sure, things have gotten better but not as much as the 1990 to 2000 leaps. And yes, laptop specs haven't changed much and this is partially because the need for spec changes wasn't present, but also during the last 20 years there has been tremendous pressure for efficiency in datacenters. Despite that, dennard scaling is dead sin…
The thing that we did in 1990-2000 was adopt new standards as the old ones became blocks on progress. I had a friend working in optical computing back in the late 80's that would wax lyrical about how optical computing was vastly superior to silicon back then. But it never took over because silicon worked well enough. If we've hit the limits of silicon then there are other options. We would need to reinvent huge chun…
Earlier quoted context omitted.
> that make the software useful, usable, and updated There is a lot of OSS software out there (e.g. in scientific communities) that I would say would barely qualify for each of those three attributes. The main reason it's valuable for the respective communities, is because it's the only thing that's available.
Developing scientific software is disproportionately hard though. Making it usable, useful and keeping it updated is even harder. There's two reasons for that. The math is generally very unorthodox and alien for a seasoned developer, and software development practices are equally alien for the scientist who can understand and evolve the math behind it. I have written a boundary element method evaluator for my Ph.D. n…
True, five years ago.
The advent of coding agents has changed everything, and often make modifying somebody else's bespoke software practical