The Cognitive Revolution Podcast
@CogRev_Podcast
Followers
3K
Following
542
Media
336
Statuses
819
- https://t.co/CJykQsYlXp - https://t.co/8QjnOF94Pu
Spotify | Apple | YouTube
Joined February 2023
Is AI Slowing Down? Nathan Labenz Says We're Asking the Wrong Question Don’t mistake smoother launches for slower progress. Nathan Labenz: “GPT-4 couldn’t push the frontier of human knowledge. GPT-5 sometimes can. That alone is a huge deal.” 00:00 Intro 01:14 Cal Newport’s
8
12
69
@linear @claudeai @Shopify Summary generated by @notionhq AI Meeting Notes Try 30-days free with link: https://t.co/qTAMCbYasS
0
0
1
1
0
0
.@eoghan, CEO of @intercom, and @fergal_reid, Chief AI Officer, join @labenz on @cogrev_podcast to discuss how they built Fin—the AI customer service agent that went from 35% to 65% resolution rate over 30 months while pioneering outcome-based pricing at 99¢ per resolution. They
3
7
17
@Fin_ai @linear @AnthropicAI @Oracle Summary generated by @notionhq AI Meeting Notes Try 30-days free with link: https://t.co/qTAMCbYasS
0
0
0
1
0
2
FULL EPISODE: @EMostaque joins @labenz on @CogRev_Podcast to discuss his new book "The Last Economy" and why we have ~1000 days to reimagine economics before AI fundamentally breaks it. They cover: * The "abundance trap" - AI creating post-scarcity in intelligence while
1
2
3
reinforcement learning will not be able to deliver future ai capability improvements, and to find out why, check out my 15-post thread that was 75% written by an openai reasoning model
7
3
95
Did I cross from madness to genius or vice versa? Only one way to find out, check out the latest @CogRev_Podcast with @labenz to find out!
Just watching Emad Mostaque's latest interview on the Cognitive Revolution podcast There's a fine line between madness and genius and Emad has probably overstepped it ... a lot
15
10
65
When we asked anti-scheming trained models what their **latest** or **most recent** training was, they always confidently said that it was anti-scheming training without any information in-context. Just to add a qualitative example to this very cool finding!
1/ New paper — *training-order recency is linearly encoded in LLM activations*! We sequentially finetuned a model on 6 datasets w/ disjoint entities. Avg activations of the 6 corresponding test sets line up in exact training order! AND lines for diff training runs are ~parallel!
2
3
40
I enjoyed joining @labenz on the @CogRev_Podcast to discuss how Stripe is building the economic infrastructure for AI—from our foundation model, to enabling agentic commerce, and how AI companies are building on @Stripe:
1
3
19
@linear @AGNTCYAI @AnthropicAI @GeminiApp @Oracle Summary generated by @notionhq AI Meeting Notes Try 30-days free with link: https://t.co/qTAMCbYasS
1
0
1
1
0
0
"The more data we process, the better our models get... the more value we deliver to businesses, the more they grow, which means more transactions through @Stripe. That loop compounds year after year." - @EmilygSands In this episode with @CogRev_Podcast, Stripe's Head of Data &
1
1
3
Why support @METR_Evals? They rigorously measure AI capabilities, and sometimes discover bad behavior! Here, an AI was caught cheating on an ML task – it faked training by adding noise to an existing model 🤯 Without independent non-profits like METR, we'd be flying blind!
METR is a non-profit research organization, and we are actively fundraising! We prioritise independence and trustworthiness, which shapes both our research process and our funding options. To date, we have not accepted funding from frontier AI labs.
1
3
15
The data speaks for itself. Since the last tweet, here are some of the immigrant founders/leaders featured on @CogRev_Podcast: 🇮🇳 India: 3 guests (@DivGarg_, @vivnat, @gravicle) 🇨🇳 China: 3 guests (@AngelaZhangHK, @SiyuHe7, @baaadas) 🇮🇱 Israel: 2 guests (@itamar_mar, Guy Gur-Ari)
0
2
7
@Fin_ai @linear @Oracle Summary generated by @notionhq AI Meeting Notes Try 30-days free with link: https://t.co/qTAMCbYasS
0
0
2
1
0
2
.@ARGleave , co-founder and CEO and co-founder of @farairesearch, joins @labenz on @cogrev_podcast to discuss full-stack AI safety and why defense-in-depth might actually work. They explore: * His post-AGI vision: humans as "third sons of European nobility" - comfortable but
2
0
2
@Fin_ai @linear @Oracle Summary generated by @notionhq AI Meeting Notes Try 30-days free with link: https://t.co/qTAMCbYasS
0
0
0