Mike Pound

But for the distilled model you don’t need a data centre. I agree 5m is a huge amount for many, but for organisations it’s a much more practical amount than the previously quoted costs for the big models.

What DeepSeek has just done is approach the performance of a flagship model from OpenAI using modest hardware and a very smart training approach. It screws with the idea that (a) you need a data centre to compete, (b) bigger is always better, and (c) that closed source AI is going to be viable

Thanks Josh! Not at all, I’m happy my videos helped. :)

“I remain unconvinced your solution has merit.”

Have transferred the turkey to the fridge. Hoping to avoid another year like the one where I forgot to defrost it and was outside with a bucket of water at 4am.

Academics finishing the semester: FINALLY I can start working on my work!

Yeah I’ll probably be able to whip it up on the sofa while watching tv?

Oops

Ben Eater I create tutorial-style videos about electronics, computer architecture, networking, and various other technical subjects.

Stuff is very busy at the moment, so why is my brain telling me to order the bits for an 8 bit breadboard computer? I do not have time, brain!

eater.net/8bit/

Created an account! Now to learn how to do things on this site..