• The explosive OpenAI controversy:
- On Friday, Sam Altman was fired by the OpenAI board from his role as CEO over a surprise Google Meet call due to his “lack of transparency”, triggering Greg Brockman to quit (X / Twitter post)
-The move has been described as a coup, caused by safety concerns over the speed of development and commercialization, even Microsoft CEO Satya Nadella wasn’t informed of it and he was “furious” (Bloomberg)
-After a whole weekend of rampant speculation on Twitter, it was announced Monday morning that ex-Twitch CEO Emmet Shear takes over the OpenAI interim CEO chair, while Altman and Brockman are going to spearhead a new AI research department at Microsoft (X / Twitter post, New York Times)
-Ilya Sutskever finally released a statement too, saying “I deeply regret my participation in the board's actions.” (X / Twitter post)
-Almost all OpenAI employees have signed a letter calling for the board to resign, or they’ll leave to work with Sam at Microsoft (X / Twitter post)
-Whether Sam returns as OpenAI CEO or moves on, one thing is sure: he’s working with Microsoft, says Satya Nadella (Bloomberg)
• This situation brought to light the division between AI safetyists (AKA effective altruists) who would like to slow down AI research and their newfound, somewhat cult-like opposition of “effective accelerationists” (e/acc) – if you want to catch up on what e/acc is, here are some materials including a post from the anonymous proponent of the philosophy, Beff Jezos. - Beff’s Newsletter, Business Insider
• Perhaps in opposition to AI safetyism, Meta has disbanded their Responsible AI team – although it might just be a structural change and not a removal, as a Meta representative said team members will support other groups in responsible AI development. - The Verge
• One of the most fun new AI features is “Make Real” in the whiteboard tool tldraw, which turns your whiteboard creations into actual UI elements – only issue is that to use it, you need to provide your OpenAI developer key. - X (Twitter)
• On a similar note, I think real-time AI image generation and manipulation is a gamechanger, plus it’s a lot of fun to play around with – you can try it out with Dynamic from AI company fal. - fal.ai
• How far can model optimization go? A good example is Insanely Fast Whisper, based on OpenAIs Whisper speech recognition model – with the right method, it takes the time to transcribe 2,5 hours of audio from 30 minutes down to less than 2 minutes. - GitHub
• The potential misuse of AI in creating biological weapons has recently been discussed by US politicians and AI industry leaders – but biosecurity experts are skeptical and say that this risk is overestimated due to the difficulty of developing such technology. - Semafor
