Top NN (neural net) engineers are getting paid $5-20M annual compensation packages. I've now heard this from 4 different extremely reputable sources but I don't see people talking about it publicly yet so here goes...
— JJ (@JosephJacks_) April 13, 2023
What are the implications of this insanity ?!?!?!? ... 🧵
The rest of the tweet stream:
1) New AI infra startups building and scaling their own models are going to continue to raise $100M to $1B within their first year ... dozens of them! And this will sustain the market prices for these scarce NN magicians.
2) GPU costs will continue to be high, supply scarce, large clusters more celebrated than small ones, and more custom silicon startups building TPUs.
3) Startups embracing open source models and fully open source AI infra will outperform those who don't by a large margin.
4) More fundamental work that promises MANY orders-of-magnitude more efficient and scalable NN architectures than the Transformer will get serious funding ... and we will see replacements for the "Transformer religion" before the end of 2023.
5) The few startups that figure out how to make fundamental progress without raising $50M+ will have the best shot of being the NVIDIAs, Googles and Apples of the next decades... Not the ones that raised billions.
6) NN courses like @avapamini and @xanamini who run @MITDeepLearning will become REQUIRED education for anyone coming into the entire technology industry as an engineer making meaningful contributions of any kind. MIT DL has 50K~ students a year... how about 5M+ a year?
7) in 5 years or so, these prices for top NN talent will come down a lot... maybe by 10X, but even then, they will still be the highest paid across the entire technology industry or ANY profession for that matter.
No comments:
Post a Comment