We are in a midst of perhaps the greatest computing power buildup ever. Sure, the vast majority of it is going towards the AI products, but at some point, between two LLM trainings, someone will decide to use at least a fraction of it for other purposes. Hey, let’s simulate…
55
63
801
Replies
@tunguz
It’s a lot like the space race I imagine. My dad told me growing up that he always supports us moving forward with space research because he saw first hand the products of it back on earth as it happened. Feels like the same thing to me
2
0
65
@HunterALanier
Indeed, a very apt analogy. With one major difference that AI in itself will be immensely and directly valuable.
1
0
56
@tunguz
Feels unlikely. We're not much better at predicting weather than we were 50 years ago - and it's got nothing to do with model scale.
3
0
7
2
0
28
@tunguz
i'm stoked for the weather one, outdoor event planning in most places on Earth is way too hard right now more than several weeks out unless you just plan to account for rain.
Humanity needs more events with millions of people outside, and so we need to be able to host them when…
3
0
8
@tunguz
Yes, actually having the capacity means that you can put a price on it. Even a day of the year worth of idle time will be enough to do many great things!
0
0
1
@tunguz
@ccerrato147
Some ideas
The Hyperspace Story
🔥 Mission
We want to make AI free and abundant, both as in beer (lower costs) and as in speech (freedom of expression and use). Free from rate limits, model limits, and free from biases. We also want to make it smarter, and we believe that is only possible…
24
33
194
0
0
0
@tunguz
When you dig into the details,
many world changing techs were developed accidentally
while pursuing other avenues of research.
The advances currently being made in material physics in order to develop ever greater compute
are sure to have many other impacts as well.
0
0
2
@tunguz
Nvidia is already using its AI computing to create a digital twin of the earth that will be used to predict weather and climate change.
0
0
0
@tunguz
In terms of predicting chaotic systems like weather patterns, neural networks are probably the best use of the computing power. We can only approximate weather as there are too many variables, so a universal function approximater is the ideal fit.
0
0
1
