2026-06-19
OpenAI researchers show that reinforcement learning on desired behavioral traits like truthfulness and corrigibility works across domains. Training on health data also improved deception detection, and the model scored better on 44 out of 53 benchmarks. The approach differs from Anthropic's constitution-based method.<br /> The article OpenAI researchers show small doses of "benefic [...]
2026-06-19
Aliens: Fireteam Elite 2 game director Chris Cross has explained why the team at Cold Iron Studios decided to make the sequel four-player co-op. [...]
2026-06-19
A US investor wants to buy into China’s most successful AI-video company. The timing could hardly be more awkward. General Atlantic is in talks to lead the first outside funding round for Kling AI, the video arm of Chinese tech group Kuaishou, Bloomberg reported. Kling is seeking more than $2bn at a valuation of $18bn, […]<br /> This story continues at The Next Web [...]
2026-06-19
Isaiah Granet spent three weeks at Y Combinator getting turned down by 180 investors. Their reason was always the same: phone calls would not exist in a year. He has now raised more than $100mn to prove them wrong. His startup, Bland, has closed a $50mn Series C led by Dell Technologies Capital, Fortune reported. […]<br /> This story continues at The Next Web [...]
2026-06-19
Until last week, about 200 organisations could use a preview of Anthropic’s most capable model, Mythos. They got in through a programme called Glasswing, after the system flagged thousands of software vulnerabilities. Then, on the evening of 12 June, the US Commerce Department ordered Anthropic to cut off every foreign national, anywhere. The company could […]<br /> This story continues [...]