LLMS

→ View original post on X — @jiqizhixin

How Well Can ChatGPT Psychoanalyze Your MBTI Type?

By

@jiqizhixin

–

05 May 2025 22h12

What’s your MBTI type—from ChatGPT's perspective? Try this prompt and find out how well it knows you: “Based on our past conversations, what MBTI personality type would you assign to me—and why?” Let the LLM psychoanalyze you

→ Voir le post original sur X — @vision_ia

OpenAI to Release Open-Weight Model Built on Democratic Values

By

@vision_ia

–

05 May 2025 19h21

Directeur Produit d’OpenAI, Kevin Weil :

Nous nous préparons à publier prochainement un modèle à poids ouverts, construit sur des valeurs démocratiques.

Ce ne sera pas notre modèle de pointe : il sera volontairement une génération en retard, afin d’éviter de favoriser… pic.twitter.com/eUkWbMq3F2
— VISION IA (@vision_ia) 5 mai 2025

Directeur Produit d’OpenAI, Kevin Weil : Nous nous préparons à publier prochainement un modèle à poids ouverts, construit sur des valeurs démocratiques. Ce ne sera pas notre modèle de pointe : il sera volontairement une génération en retard, afin d’éviter de favoriser

o3 Demonstrates Strong Understanding of Jurassic World Timeline

By

–

05 May 2025 19h14

Will note that o3 has a robust enough sequence of events from the Jurassic World film to construct and critique a timeline of character actions in a particular scene, and while it’s clearly riffing it understands much more detail than I would have expected.

Next Token Models Hallucination Research Findings

By

–

05 May 2025 18h44

“Sometimes next token models hallucinate” one of the more clarifying research results in the last few years.

LLM Design Choices Outperform Traditional Sci-Fi Set Design

By

–

05 May 2025 18h11

… “transition corridor.” (Probably not what I would have called it, but it’s a two stage airlock design, because LLM makes better choices than movie set designers out of the box.)

LLM World Models and Failure Case Defense Mechanisms

By

–

05 May 2025 18h10

Sometimes people say that the LLMs make obvious errors because they don’t have a world model, but one is welcome to one’s guess as to what percentage of population has sufficiently detailed world model to discuss layers of defense against failure cases in a remotely operated…

AI-Generated Creative Content Quality and Economic Valuation

By

–

05 May 2025 17h38

Result is not Shakespeare but it’s subjectively similar to what you’d expect a Google SRE with a flair for fanfiction to crank out in 2-4 hours, and that’s a pretty expensive person to task with on demand flash fiction.