“Dangerous” is a bit extreme here lol I’m just saying that it sounds rad.
The article did say that even with the cost of a launch, the energy footprint is 1/10 of what it would be on earth. While I agree that it’s a fluff piece, it’s also not at all crazy to believe that NVidia could afford to harden a single H100 and send it to orbit. They wouldn’t even have to use it for anything for the article to be “true” lol.
This is also hyperbole, obviously we don’t have the electrical capacity of tomorrow available today…that’s why everyone is building more of it lol. It would be true due to population growth, electric vehicles, and outdated infrastructure, even without AI data centers.
All of this being said, I am sure it’s a hype piece meant to pump up the stocks before years end. But again, NVidia can absolutely afford to send a single GPU to orbit just for funzies, and it’s not at all “dangerous” to believe it, nor is it dangerous to be excited for the space race/to dream of future possibilities
Sure, based on something which, in the opinion of a knowledgeable person above, is extremely unrealistic… Sounds a lot like people who are perfectly fine with Elon constantly hyping FSD as long as ‘line goes up’. To the rest of us, that’s fraud.
You and I both assume he is actually knowledgeable, we (as people) are constantly assuming things. What if I told you I also work in space industry and have been working with satellite data on a daily basis for over a decade, and that I think it is entirely possible for NVidia to launch a single GPU to orbit in 2026? What would happen then?
Haha brother I would pull a Luigi if I ever met Elon lmao fully agree FSD is fraud. NVidia investing in a start up meant to research the feasibility of computers in orbit, and releasing an article about said start up launching a single GPU into orbit is not fraud.
For some it’s an assumption, but there are folks here in the forum who know what spacecraft I’ve built, am currently flying, and the one I am currently building.
One GPU in space does not a space data center make.
A point we agree on.
Edit: likewise I won’t take the topic further off course.
There is a lot good (like writing code under supervision) and a lot bad about AI.
Some people think it is going to wake up one day and spontaneously get its own goals. Like in Minecraft, the objective of the game is to get materials to get to the End dimension to defeat the Ender Dragon. And humans decide, nah, Imma make a massive RISC CPU out of redstone and pistons.
It could pick something equally bizarre relative to the goals we try to give it, like turning the world into paperclips. If it’s as powerful as Sam Altman is hoping, we won’t be able to stop it.
But I don’t think that will happen. I made a video explaining why.
The issue I have is the intent of the owners/operators of any LLM or GAI - It cost money to make and maintain they want that money back, they aren’t interested in Altruism
I was going to suggest that if AI could provide the solution for cheap abundant energy for everyone there would be no business advantage in sharing that - but even that’s a bogus argument. We have the solutions for cheap energy, eliminating world hunger, childhood mortality and poverty right now. Have had for decades probably - but we don’t. So AI isn’t going to change that.
Well, I used Claude to work out a github zip system for the fluidnc configs. Like you guys said, It wasn’t perfect but got me close enough to figure it out eventually with a few more prompts. It was odd in that I asked for some specifics but it still made up its own names on things.
I guess, the only way to stay relevant is use everything at your disposal.
I don’t think that’s staying relevant as much as keeping the learning curve shallow. If you wait until it’s perfect the new kids will be light years ahead.
AI has made me redundant already. I had a call from one of my daughters (in her late 40’s) last week. She and her husband had just finished polishing updates to their resumes. Normally they would make a draft outline and shoot it to me to fill it in using “grown-up” words. . They were quite excited to have had the machine do the job, and in her words, “maybe not better but much faster than DadGPT”!
I am a sceptic of AI. With that said, I taught myself coding for Arduino. I designed and built a solar tracking system for my solar panels and it worked pretty well.
I wanted to redesign it so out of curiousity I used CoPilot to write the code. Was it perfect? No. But it did approach if differently than I did and I used it to learn a new way of organizing my code.
As long as there is an air gap between it and the ability to actually build something I am fine…
. They were quite excited to have had the machine do the job, and in her words, “maybe not better but much faster than DadGPT”! 