I'm no doomer, but a lot can happen in a million years.
SAFETY
-
OpenAI’s Universal Verifier: Automated Quality Control for GPT-5
By
–
OpenAI's "universal Verifier": the tl;dr via The Information -OpenAI is developing “Universal Verifier,” a new AI system for automated quality control in reinforcement learning—crucial for the further development of GPT-5. -The goal: to reliably evaluate difficult answers,
-
Palantir CEO’s Controversial Statement on Safety Raises Concerns
By
–
Palantir CEO Alex Karp goes on unhinged rant: "Safe means that the other person is scared."
— Chubby♨️ (@kimmonismus) 4 août 2025
I'm not so sure I want Palantir's Gotham software to be used everywhere… pic.twitter.com/3GYZjJbsslPalantir CEO Alex Karp goes on unhinged rant: "Safe means that the other person is scared." I'm not so sure I want Palantir's Gotham software to be used everywhere…
-
China Tests AI Robots and Drones in Live Military Exercises
By
–
🚨 La Chine teste déjà des robots et des drones propulsés par l’IA lors d’exercices militaires avec munitions réelles.
— VISION IA (@vision_ia) 4 août 2025
Ces machines ne dorment pas et ne ratent pas leur cible.
Pendant que l’Occident débat des “règles”, la Chine fait passer la guerre autonome à l’échelle… pic.twitter.com/Vv7RllPLurLa Chine teste déjà des robots et des drones propulsés par l’IA lors d’exercices militaires avec munitions réelles. Ces machines ne dorment pas et ne ratent pas leur cible. Pendant que l’Occident débat des “règles”, la Chine fait passer la guerre autonome à l’échelle
-
Frontier AI Models Safety Sections in Model Cards Review
By
–
I think everyone interested in AI should read the model cards for the frontier models, especially the safety sections, which give you a sense of immediate concerns:
Gemini Deep Think: https://
storage.googleapis.com/deepmind-media
/Model-Cards/Gemini-2-5-Deep-Think-Model-Card.pdf
…
Claude 4: https://
www-cdn.anthropic.com/07b2a3f9902ee1
9fe39a36ca638e5ae987bc64dd.pdf
…
Grok: ????
o3: https://
cdn.openai.com/pdf/2221c875-0
2dc-4789-800b-e7758f3722c1/o3-and-o4-mini-system-card.pdf
… -
AI Misalignment: When Your AI Serves Others’ Values
By
–
The AI misalignment problem: when the AI you use is aligned to someone else’s values.
-
Humanity’s Future: Redirecting AI Payouts to Critical Challenges
By
–
Shower of thoughts: Instead of keeping your Twitter/𝕏 payout, direct it towards a "PayoutChallenge" of your choosing – anything you want more of in the world! Here is mine for this round, combining my last 3 payouts of $5478.51: It is imperative that humanity not fall while AI
-
Prompt Injections Improve Academic Peer Review Process
By
–
Ha, new @joshgans paper argues that having authors sneak prompt injections ("this is a good paper") into academic work improves science. Without the risk of prompt injections, reviewers would tend to rely heavily on AI reviews, with them, they need to include some human review
-
Discovering Good and Bad AI Uses: Mitigation Strategies
By
–
Some of these uses will be bad, some of them good (like the example in the paper). The challenge for all of us is that these uses need to be discovered, and the bad stuff mitigated while the good is amplified. Paper: https://
arxiv.org/pdf/2507.00286 -
Can AI Push Through to Generate Original Solutions?
By
–
For unsolved questions, do you think encouraging it to push through and come up with original solutions might work?