🔥Telegram İfşa

The wave that first hit protein folding scientists in 2019 is now coming for mathematicians

Props to Demis for being to date the highest integrity AGI lab CEO. Ironic when you know the founding history of the competitors 🥲

1. The new model seems like a big deal. The parallel thinking innovation seems novel and promising. 2. The tweet convey immaculate wizardry vibes. Simply perfect.

AI security is becoming a thing

Naming idea: Claude 4.05

I noticed this trend where the worst the product is, the more Altman is hyping up the thing ahead of time and faking the "I'm scared". For GPT-4 he was actually downplaying the jump and highlighting the weaknesses. For o1 he didn't discuss the risks. For 5 (minor jump) he did

OpenAI: *Hold my beer*

TIL Gemini Gems were the equivalent of Claude and ChatGPT Projects. Another win for Google comms team!

AI mass persuasion is one of the most underrated areas of concern Episode 1 was Blake Lemoine. Episode 2 is below.

Yes! Congrats xAI team for maintaining a fairly truth seeking chatbot! I'd be really excited if xAI managed to keep improving on their core mission. As the informational footprint of these rising sapients grows, we'll increasingly need it.

Finally, someone did the obvious thing! It's framed for open source models but close source providers should also take notes.

Sig should receive a referral fee at this point.

Real-world misalignment 👀?

I've entertained that theory for a few years and have been confused since then why people expected confidently so much GDP growth. Basically prices if goods should crash so fast that the question of "how do you count inflation" will become the first order parameter of whether and

I have this theory that we are in a period of increasing marginal utility of capabilities. GPT-2 to GPT-3 jump was a bigger jump than 3 to 4, which was bigger than 4 to 5. But the utility jumps have been increasing. My core thesis for why is that most use cases are bottlenecked

Good thread, as usual