181–190 of 581 posts
Earlier quoted context omitted.
> Build fabs Such a misguided take. If it were that easy to build fabs to compete with TSMC, Intel would do it. China would do it. Hardware is way harder than software.
Or maybe, just maybe, TSMC has gaslit all of us to think it is much harder than it really is. If Taiwan isn't needed for its chips, what is the strategic interest of the USG?
Earlier quoted context omitted.
Using Claude 3.5 Sonnet in Cursor Composer already shows huge benefits for coding. I'm more productive than ever before. The models are still getting better and better. I'm not saying AGI is right around the corner or that we will reach it, but the benefits are undeniable. o1 added test-time compute. No need to be snarky.
There's no accounting for taste, but keep in mind that all of these services are currently losing money, so how much would you actually be willing to pay for the service you're currently getting in order to let it break even? There was a report that Microsoft is losing $20 for every $10 spent on Copilot subscriptions, with heavy users costing them as much as $80 per month. Assuming you're one of those heavy users, wo…
Earlier quoted context omitted.
>when the dust settles we'll have LLMs significantly more powerful than the current cutting edge -- which is already useful. LLMs, irrespective of how powerful, are all subject to the fundamental limitation that they don't know anything. The stochastic parrot analogy remains applicable and will never be solved because of the underlying principles inherent to LLMs. LLMs are not the pathway to AGI.
Arguing over terminology like "AGI" and the verb "to know" is a waste of time. The question is what tools can be built from them and how can people use those tools.
7 trillion is absurd. The recent advancements in AI are real, but not on the order of 7T.
they will be. Looking at the rate of progress (just tried o1mini), its inevitable.
Let's talk about that. GPT-3.5 (specifically text-davinci-003) was a massive leap over everything that came before it. GPT-4 was a massive leap over GPT-3.5. Everything since (GPT-4 Turbo, GPT-4o, GPT o1) feels like steady incremental improvement (putting aside multimodal capabilities) rather than massive leaps. I'm far from convinced that the rate of progress made with foundation models in 2022-2023 has continued in 2024 let alone from extrapolating that it will continue for the next several years.
He may have the last laugh. His traveling around the world and threatening to build fabs in the UAE and Taiwan is a diplomatic scheme to get the U.S. hawks into action. And lo and behold, ClopenAI has already hired CHIPS act people: "To bolster its efforts, OpenAI has hired Chris Lehane, a Clinton White House lawyer, as its vice president of global policy, along with two people from the Commerce Department who worked…
> Build fabs Such a misguided take. If it were that easy to build fabs to compete with TSMC, Intel would do it. China would do it. Hardware is way harder than software.
Earlier quoted context omitted.
I would pay hundreds of dollars per month for the combination of cursor and claude - I could not get my head around it when my beginner lever colleague said "I just coded this whole thing using cursor". It was an entire web app, with search filters, tree based drag and drop GUIs, the backend api server, database migrations, auth and everything else. Not once did he need to ask me a question. When I asked him "how lon…
Current LLMs fail if what you're coding is not the most common of tasks. And a simple web app is about as basic as it gets. I've tried using LLMs for some libraries I'm working on, and they failed miserably. Trying to make an LLM implement a trait with a generic type in Rust is a game of luck with very poor chances. I'm sure LLMs can massively speed up tasks like front-end JavaScript development, simple Python script…
He may have the last laugh. His traveling around the world and threatening to build fabs in the UAE and Taiwan is a diplomatic scheme to get the U.S. hawks into action. And lo and behold, ClopenAI has already hired CHIPS act people: "To bolster its efforts, OpenAI has hired Chris Lehane, a Clinton White House lawyer, as its vice president of global policy, along with two people from the Commerce Department who worked…
The CHIPS act has provisions to help fund fabs in allies.
This is why Biden and Modi announced a Feb dedicated to US and Indian defense systems [0] at the QUAD summit as well as as the designation of the UAE as a "Major Defence Partner" along with India [1] which includes tech transfer conditions.
A significant portion of the CHIPS and IRA is set aside to help with subsidizing international allies tech and innovation ecosystems in order to ensure they don't lean towards China [2]
[0] - https://www.bloomberg.com/news/articles/2024-09-23/biden-mod...
[1] - https://www.reuters.com/world/us/harris-plans-raise-gaza-cea...
[2] - https://cset.georgetown.edu/publication/agile-alliances/
The current AI hype wave has really hit a nerd soft spot - that we're steps away from AGI. Surely if a computer can make plausible-looking but incorrect sentences we're days away from those sentences being factually accurate! The winter after this is gonna be harsh.
Using Claude 3.5 Sonnet in Cursor Composer already shows huge benefits for coding. I'm more productive than ever before. The models are still getting better and better. I'm not saying AGI is right around the corner or that we will reach it, but the benefits are undeniable. o1 added test-time compute. No need to be snarky.
Earlier quoted context omitted.
"Plausible-looking but incorrect sentences" is cheap, reflexive cynicism. LLMs are an incredible breakthrough by any reasonable standard. The reason to be optimistic about further progress is that we've seen a massive improvement in capabilities over the past few years and that seems highly likely to continue for the next few (at least). It's not going to scale forever, but it seems pretty clear that when the dust se…
>when the dust settles we'll have LLMs significantly more powerful than the current cutting edge -- which is already useful. LLMs, irrespective of how powerful, are all subject to the fundamental limitation that they don't know anything. The stochastic parrot analogy remains applicable and will never be solved because of the underlying principles inherent to LLMs. LLMs are not the pathway to AGI.
Repeatedly, we’ve thought that humans and animals were different in kind, only to find that we’re actually just different in degree: elephants mourn their dead, dolphins have sex for pleasure, crows make tools (even tools out of multiple non-useful parts! [1]). That could be true here.
LLMs are impressive. Nobody knows whether they will or won’t lead to AGI (if we could even agree on a definition – there’s a lot of No True Scotsman in that conversation). My uneducated guess is that that you’re probably right: just continuing to scale LLMs without other advancements won’t get us there.
But I wish we were all more humble about this. There’s been a lot of interesting emergent behavior with these systems, and we just don’t know what will happen.
[1]: https://www.ox.ac.uk/news/2018-10-24-new-caledonian-crows-ca...