I see, it looks like he's the one behind ThreeJS. Well, he had to make ThreeJS before he could make this, and that must have been a learning experience, right? :D
If the people in the Netherlands can do it - where it's flat, windy, and rainy most of the time - then most people in the US can also do it. If it's too hot, go in the mornings.
Does the Netherlands have huge roads where everyone travels at 50mph+ in huge pickup trucks and SUVs with high hoods and drivers looking at their phones while driving?
Because that is what the majority of Americans deal with. Not to mention a significant majority deal with colder and or hotter and more humid weather. Obviously, it can be done, but the Netherlands is not the standard for low quality bicycling environments.
Jon Stewart took exception: "This is how we're doing this? 2 am? Mar-a-lago basement? no lighting? You don't even have one of those influencer halo things? You just go down in the basement? and this is what we're wearing? Blazer, no tie, shirt unbuttoned? Looking more like the father of the bride settling up with the caterer? ... and not to nitpick, but baseball hat? ... " and it goes on.
And they likely used hair style that requires absurd amount of maintennance. If there were women in the group, they likely used expenasive and painful body modifications on top of that. Republicans have own style, they are big on looks, after all.
I mean MAGA styke plastic surgeries there: botox, lip augmentation, jaw contouring, microneedling, facials, chemical peels and so on.
Body modification like that is a status symbol amoung republican women. Both high level women around Trump and in normal level republican areas with serious plastic surgeries industries.
How about if you run this loop (one year from now) on this kind of hardware but with something like Claude/Kimi K2. How about that? Because that's where it'll go.
Why so negative lol. The speed and very reduced power use of this thing are nothing to be sneezed at. I mean, hardware accelerated LLMs are a huge step forward. But yeah, this is a proof of concept, basically. I wouldn't be surprised if the size factor and the power use go down even more, and that we'll start seeing stuff like this in all kinds of hardware. It's an enabler.
You don't know. You just have marketing materials, not independent analysis. Maybe it actually takes 2 years to design and manufacture the hardware, so anything that comes out will be badly out of date. Wouldn't be the first time someone lied. A good demo backed by millions of dollars should not allow such doubts.
Did you not see the chatbot they posted online (https://chatjimmy.ai/)? That thing is near instantaneous, it's all the proof you need that this is real.
And if the hardware is real and functional, as you can independently verify by chatting with that thing, how much more effort would it be to etch more recent models?
The real question is of course: what about LARGER models? I'm assuming you can apply some of the existing LLM inference parallelization techniques and split the workload over multiple cards. Some of the 32B models are plenty powerful.
