Sparks of AGI, Jagged Frontiers, AI forms "Internal Models" and the Dark… https://
youtu.be/GjMXXtce9cA?si
=oQVDgkLIimepH9mx
… via @YouTube
RESEARCH
-
AGI Sparks: Internal Models and AI’s Dark Frontiers Explored
By
–
-
Quality Data Curation Essential for AI Model Performance
By
–
Models & infra are easy to buy.
Better data isn’t. Snorkel Expert DaaS helps you: Spot blind spots Curate domain-tuned data Scale human-in-the-loop For performance, compliance, & control—start with your data. Start with Snorkel -

Gemini 2.5 Pro performance: fast but unreliable, 2 correct out of 7
By
–

A few people have reported Gemini 2.5 Pro can do this as well. It can, and is much faster, but less reliable—in 7 attempts of my first prompt above, it returned: – 3 wrong answers (e.g. 2 shown below)
– 2 incomplete answers (omitting the cursive part)
– 2 fully correct answers -
Quantum Computers Could Break RSA Encryption Within a Week
By
–
Le chiffrement RSA (informatique) protège l’économie mondiale. Mais selon un chercheur de Google, un ordinateur quantique pourrait bientôt le briser… en une semaine.
— VISION IA (@vision_ia) 7 juillet 2025
Et si certaines agences possèdent déjà vos archives chiffrées ? Le jour où elles les liront d’un coup, tout… https://t.co/adGT01FQj3 pic.twitter.com/2PnFZJAKeJLe chiffrement RSA (informatique) protège l’économie mondiale. Mais selon un chercheur de Google, un ordinateur quantique pourrait bientôt le briser… en une semaine. Et si certaines agences possèdent déjà vos archives chiffrées ? Le jour où elles les liront d’un coup, tout
-

MIT and NVIDIA Speed Up Robot Motion Planning Process
By
–
Robots consider many actions, constraints, & mechanical capabilities when handling objects. An approach from MIT CSAIL & NVIDIA speeds up this planning process, helping robots "think ahead" by considering thousands of solutions & refining the best ones: https://
bit.ly/4nuVwjV -
Goodside seeks kindergarten path-following puzzles that fail in o3-pro
By
–
If you have good example of this you’d like me to try in o3-pro (that fails in other models) feel free to DM and I’ll give it a shot. Diagrams are harder in general; I’ve seen kindergarten-level path-following puzzles that it can’t do at all.
-
Sycophancy and Model Steerability: Control vs Instruction Following
By
–
I would argue there is a second problem, whose dimensions are unclear, which is whether sycophancy is part of steerability as well. A model that isn't sycophantic may not follow instructions?
-
Iterative code and perception for blind image segmentation
By
–
It uses tools, yes, but it’s not a single program—it’s many iterative rounds of code use and model perception of intermediate transformed output. The visual intuition needed to write one program to correctly segment and de-scramble the image “blind” would be much greater.
-
Free Will Decision Making Under Uncertainty and Moral Choice
By
–
Freedom of will is decision making under uncertainty or cheating. If there is no uncertainty, the only degree of freedom is the decision against doing the right thing.
-
Hinton Warns Superintelligent AI May Discard Humanity Without Hatred
By
–
D'après Hinton, les poulets ne négocient pas avec les fermiers. Si une IA devient plus intelligente que nous, elle n’aura pas besoin de nous haïr pour nous remplacer. Il lui suffira d’estimer qu’on est de trop. pic.twitter.com/W2js69Yraf
— VISION IA (@vision_ia) 7 juillet 2025D'après Hinton, les poulets ne négocient pas avec les fermiers. Si une IA devient plus intelligente que nous, elle n’aura pas besoin de nous haïr pour nous remplacer. Il lui suffira d’estimer qu’on est de trop.