4/ Notice that the EleutherAI team is hoping to release GPT-NeoX with 150-200 billion parameters (competing with GPT-3 DaVinci) by Aug-Dec 2021. Just around the corner.

3/ @CoreWeave is doing a great thing in supporting EleutherAI by donating high-performance GPU compute time.

I hope to see other companies doing the same.

2/ I’m even more excited by the FAQ of the project:

“Q: Are you serious when you say you are going to train a model comparable to the biggest GPT⁠-⁠3 (175b parameters)?
A: Yes…We expect our final model to be somewhere between 150 and 200b param”

https://t.co/P2LWvL8dpY

1/ What an exceptional effort. I’m very excited about this. Must-read if you are looking to build something with #AI that is not computer vision. >

GPT-3’s free alternative GPT-Neo is something to be excited about https://t.co/Jlmg7ASRuK

@dwellington I never say “chatbots”, in any circumstance.

I have been watching (and using) @chirrapp for quite some time now. It’s a great (and rare) example of how you can still innovate in a market that seems crowded. I’m amazed that all those Twitter clients out there have never done the things that @kossnocorp has implemented.

2/ There’s so much more #AI could do for IT operations and I’m amazed that after years of great progress in ML we still haven’t seen anything. But it’s also true that we are in a winter period for enterprise management. Ideas and startups will eventually come back.

1/ The term “AIOps” really refers to infra/app monitoring supported by ML for things like anomaly detection and root-cause analysis. It’s a real shame because it implies that #AI can only help ITOps in the monitoring space. It’s not true, and a self-imposed artificial constraint.

GitHub/GitLab: code showcase
@Replit: running code showcase

I highly recommend reading the announcement:
Replit Apps https://t.co/BfYJN0excZ

If you have time to read only one blog post this week, read this one. Full of difficult questions worth asking. >

Selling hours https://t.co/qsvLxYQ0EC https://t.co/iGb9TxjSIb