Tech
A new Berkeley human-rights assessment finds that large language models can help people reach legal, educational and journalistic resources they otherwise lack. The same convenience can also weaken judgement when assistance becomes substitution.
In focus
Tech article
Samsung is becoming one of OpenAI’s biggest enterprise customers and an important supplier to its AI infrastructure ambitions. Reports that the two companies are already co-developing next-generation processors, however, went further than OpenAI was prepared to confirm two days later.
Tech article
A former DeepMind communications staffer says the lab once prohibited public discussion of AI-caused human extinction, even while researchers worried about alignment internally. The allegation cannot be independently verified from public records, but DeepMind’s own publications show a striking change: from cautious talk about generic AI risks to an explicit programme for studying existential risk.
Tech article
Public concern is giving AI-safety advocates a larger political constituency. The complication is that voters worried about jobs, surveillance, copyright or data centres may support the same guardrails for very different reasons.