RT @AlphaSignalAI: Google just released MetNet-2, a deep learning model that can predict rain up to 12 hours in advance.

Published in Nat…

“[This] effectively completes our v1 Copilot story,” GitHub CEO Thomas Dohmke told me.
Dohmke expects that soon, Copilot will be able to generate 80% of a developer’s code. Today, that number is about 46% across programming languages — and 61% for Java.

https://t.co/ICAz35ZRl5 https://t.co/8CfP7JMyVk

These statistics are not surprising. I’ll repeat what I said before:
https://t.co/NQrnzpHZqo

For a glimpse of Copilot v2, you should check the VScode extension “Copilot Labs”.

We know 🙂 https://t.co/cWknP2vkGI

RT @janleike: With the InstructGPT paper we found that our models generalized to follow instructions in non-English even though we almost e…

To put this in context, the EU has a total population of 447m people (2021, but very little difference in 2022).

286m were aged 15-64 years old in 2021.

So, every month, Meta reaches almost 64% of all teenagers+adults in the EU. A massive capability to influence people. https://t.co/Xmzh3Lmc4X

See, this is the problem with hiring managers. Always looking for junior resources with not enough years of experience. https://t.co/LGuc9s0qBU

#GenerativeAI is the final nail in the coffin for DSLR. Once the technology will reach a level of maturity and integration with the hardware (I’m looking at you, @Apple) to do on-device what @runwayml is doing on the desktop, in real-time, there will be very few people wanting it https://t.co/hUSFOkd7Xo

Go on. Say the word “compression” in every imaginable context for 3 weeks in a row now. 🙄

Once I’ll have access to Bing + #GTP4, it will be interesting to do some comparisons. In the meanwhile, enjoy #AI’s perception of London’s startup scene.