fooker

26 ms·

511.
by fooker 9mo ago
I have burned through more than my salary in AI API calls and nobody seems to care!
512.
by fooker 9mo ago
> from a tornado to a casual encounter with an elephant A sims style game with this technology will be pretty nice!
513.
by fooker 9mo ago
For the current state of AI, the harness is unfortunately part of the secret sauce.
514.
by fooker 9mo ago
It would be tragically ironic if this post is AI generated.
515.
by fooker 9mo ago
Yeah, these benchmarks are bogus. Every new model overfits to the latest overhyped benchmark. Someone should take this to a logical extreme and train a tiny model that scores better on a specific benchmark.
516.
by fooker 9mo ago
I don’t doubt the end goal. My point is that it won’t be a ‘layer’ like it is now and the technology will be completely different from what we see as agents today.
517.
by fooker 9mo ago
I’m saying we won’t call it agents and it will involve substantially different technology compared to what we mean by agents today. Of course AI will keep improving and more automation is a given.
518.
by fooker 9mo ago
> it doesn't mean "it can't work" I don’t have an answer for you, but take some applause from me for spelling this out :) It’s very difficult for most people to intuitively understand that what they could not figure o
519.
by fooker 9mo ago
Good example.
520.
by fooker 9mo ago
No it’ll be some idea we have not developed or named yet. The current ‘agent’ ecosystem is just hacks on top of hacks.
521.
by fooker 9mo ago
> I suspect ten years from now, people will look back at 2024-2025 as the moment Apple had a clear shot at owning the agent layer and chose not to take it Ten years from now, there will be no ‘agent layer’. This is like predicting Micros
522.
by fooker 9mo ago
You are missing one important thing here. You do not radiate all the heat away from a GPU, a modern GPU can run pretty hot. Also look up how this is getting better for the next generation of GPUs. Maybe repeat your calculation with updated
523.
by fooker 9mo ago
> Instead you put your confidence in Elon No, I put confidence my ability to do a web search, pretty rare skill nowadays ;) You'll see that none of these are Elon/spacex, hopefully? https://medium.com/@cognidown
524.
by fooker 9mo ago
> some hype merchants Excluding Spacex: Nvidia, Google, China, European Commission, Blue Origin And this being HN, a YC funded company has put a single GPU rack in space and demonstrated training a reasonable sized model on it. But yeah,
525.
by fooker 9mo ago
> Right now only upsides ... You are missing some pretty important upsides. Lower latency is a major one. And not having to buy land and water to power/cool it. Both are fairly limited as far as resources go, and gets exponentially
526.
by fooker 9mo ago
The bottleneck is deploying solar physically, not making the cells. We have increased the manufacturing of pretty much every piece of technology you see in front you by 200x at some point in history. Often in a matter of years.
527.
by fooker 9mo ago
Yep, definitely being scammed by not dismissing things outside my area of expertise out of hand. I wish I had your confidence about everything!
528.
by fooker 9mo ago
Maybe we are talking about different things here? I don't doubt spacex can fail at this. I also don't doubt we are fairly close to making this plausible. > plenty of them who have spent careers building data centers Famously, p
529.
by fooker 9mo ago
> Fighting thermodynamics is a losing race The great thing about your argument is that it can be used in any circumstance! Cooling car batteries, nope can't possibly work! Thermodynamics! Refrigerator, are you crazy? You're fig
530.
by fooker 9mo ago
> Can people please try and use their fucking brains for a second? Have you considered that people smarter than you think it is plausible?
531.
by fooker 9mo ago
It doesn't work like that, but it can. Latency would be fine for inference, this is low earth orbit, that is about 25ms optimistically. Well within what we expect from our current crop of non local LLMs.
532.
by fooker 9mo ago
Yes, Rubin reportedly can deal with running significantly hotter. That makes radiating a much more practical approach to cooling it.
533.
by fooker 9mo ago
LLMs specifically are fine with random bits flipped for the results to be 'creative'.
534.
by fooker 9mo ago
It will cost less to put it in low earth orbit than it will be to purchase land for it at any reasonable location.
535.
by fooker 9mo ago
beltalowda!
536.
by fooker 9mo ago
There are several other companies that have announced efforts to try data centers in space.
537.
by fooker 9mo ago
Great, so you seem to agree the technology exists for this and it is a matter of deploying more of it?
538.
by fooker 9mo ago
Sure, speed matters. Developing new technology happens to matter more. I'm sure investors are going to do their own analysis on this and reach their own conclusions, you should try betting against it.
539.
by fooker 9mo ago
> Figuring out how to radiate a lot of waste heat into a vacuum is fighting physics. Radiators should work pretty well, and large solar panels can do double duty as radiators. Also, curiously, newer GPUs are developed to require signific
540.
by fooker 9mo ago
>Tomorrow? Eventually :) Markets are forward looking, and not really bound to 'tomorrow'.

More ›