api

8 ms·

121.
by api 28d ago
There's a deep failed idea here. It's the idea that technology is an alternative to politics, period. I think that whole idea has failed. You can't avoid building technologies that could be used by a police state, because lit
122.
by api 28d ago
Isn't a MoE model basically a cascading tree of smaller models or some variation of that?
123.
by api 29d ago
Wait there’s a mouse that requires an account? I’ve been using just Linux and Mac long enough to not know this, and I’m glad.
124.
by api 29d ago
I'd say the code I see coming out of leading models today is okay. Not great, but not terrible. It's also easy to get LLMs to code in more efficient languages, and you can also goad them via prompts and CLAUDE.md type stuff into
125.
by api 29d ago
I agree with the sentiment, but there is a devils' advocate argument: Most compute loads are bursty. Most PCs and phones sit idle almost all the time. It's an inefficient use of capital. In the cloud you can be idle and then sudde
126.
by api 29d ago
This is part of why I think the data center build-out is a bubble. We've barely scratched the surface when it comes to hardware optimization. We'll see exponential improvements in energy efficiency and speed over the next decade.
127.
by api 29d ago
I'm glad I splurged years ago and got an M1 Max with 64 GiB RAM. Still fast enough for everything, don't plan to upgrade until this cools off.
128.
by api 29d ago
Memory in the cloud isn't free either.
129.
by api 29d ago
You really think Lobster Jesus is what most of this is doing? It's doing code, analytics, training, research, operating services. The silly slop stuff is a small rounding error of usage.
130.
by api 29d ago
That's true. Making them pay enough for water might mean the data center doesn't even get built there. It gets built somewhere more logical with abundant cheap water.
131.
by api 29d ago
I've done a bit of research into it, and I think it's a mix. The energy thing is exaggerated. Want to compensate for a week of heavy AI coding usage? If you drive to work, telecommute for one day. Skip one long-distance DoorDash o
132.
by api 29d ago
Industry build-out in the USA? It's raising our energy and water costs. Lack of industry build-out in the USA? We're losing our industrial base to the Chinese. Social media always promotes the angry take because that's what m
133.
by api 1mo ago
It was really the touch screen UI, which was brought to market by Apple. Always been curious if it was developed there or if like the desktop windowed UI it was pioneered elsewhere. Prior efforts were menu oriented or attempts at desktop UI
134.
by api 1mo ago
This is new? There were tons of fake news sites, fake blogs, and sock puppet and troll farm accounts before AI. I guess the new angle might be that AI scrapers have even less discernment than humans, so you don’t have to put much effort int
135.
by api 1mo ago
Ahh, yes, the cosmic horror hypothesis about how anesthesia works. You can't move and experience everything but remember nothing. But hey, I don't have cancer anymore. So if I had to endure four hours of agony that I don't re
136.
by api 1mo ago
I had some serious surgery a few years ago and paid attention when I went under to see if I could recall anything. Nope. You don’t exist for that time. Anesthesiologist counts down 10, 9, 8, 7, and then I was teleported to the recovery bed.
137.
by api 1mo ago
All cults are for-profit in some sense. Sometimes it's not money though. Sometimes it's getting the leader(s) laid or getting them their "narcissistic supply." I've heard tons of rumors about a get-laid angle to the
138.
by api 1mo ago
I've read some of it. It's a cult. The net effect of their recommendations is to mandate a monopoly on AI, which oh so happens to be them and their friends. That's what most cults' ideologies boil down to: reasons to giv
139.
by api 1mo ago
You're getting lots of hate because people assume this will only subtract, that people will use tools like this to replace their creative process. Anyone who does that isn't creative to begin with, and isn't trying to be crea
140.
by api 1mo ago
Not a new sentiment. This is over two thousand years old: https://www.anthologialitt.com/post/the-god-thoth-and-the-in... Every invention is a Faustian pact. It amplifies certain things and atrophies others. Personally
141.
by api 1mo ago
He would legitimately try this. There’s no limit to the cartoonish narcissism. It’s absurd and would fail of course.
142.
by api 1mo ago
I keep saying Linux can win on desktop by just sitting still while Windows sucks more and more.
143.
by api 1mo ago
Closely related: one of the biggest AI superpowers I've found is testing. "Please implement a comprehensive unit testing suite for this, making sure to test ..." You still have to check the work and shouldn't trust it bl
144.
by api 1mo ago
Chinese scale is one way RISC-V could win. If China suddenly starts reaching <5nm process nodes at scale and uses RISC-V, they'd flood the market with cheap high performance RISC-V chips and probably start using them in their domest
145.
by api 1mo ago
I’m giving my kidneys a participation trophy.
146.
by api 1mo ago
If China leapfrogs EUV, this could happen.
147.
by api 1mo ago
Linux is decent for its core use cases, but it is far from a solid pro-grade OS in a lot of areas... and in the areas it did get there, it took a long time to get there.
148.
by api 1mo ago
Eventually CPUs and GPUs converge: huge numbers of CPUs with wide vector units.
149.
by api 1mo ago
5% of that might get built. Maybe. We are nearing the top of a giant DC build out bubble. My guess is that there’s someone making money for getting DCs permitted. Like the paperwork done. 474 GW of capacity is silly woo woo land. I’m not su
150.
by api 1mo ago
X86 is the best argument that you can build a fast efficient RISC-V chip... because the X86 instruction set is a much bigger mess. It just blows my mind sometimes when designers don't learn insanely obvious lessons from the past, basic

More ›