Lynette Bye

Commentary by Lynette Bye

From the commentary rail — every link leaves the site for the original piece.

• 2 October 2026 · Epoch AIHundreds of millions of AI agents are coming. Is there work for them?The AI compute buildout could support more AI agents than the US has workers
• 28 September 2026 · Epoch AIAI is getting cheaper faster than any other transformative technologyThe price is falling 13x year over year
• 24 September 2026 · Epoch AIHow far behind Nvidia is Huawei?Huawei will substantially improve its AI compute solutions by 2030, but export controls constrain its most important scaling levers, making it unlikely to catch up with Nvidia.
• 27 August 2026 · Epoch AIAn update on AI’s most important numberHow quickly are Anthropic and OpenAI growing their revenue?
• 14 July 2026 · TransformerA data bottleneck could slow the superintelligence raceAnd that could be a good thing
• 18 March 2026 · TransformerNo, alignment isn’t solvedProgress on ensuring models are in step with humans has calmed nerves. But some of the biggest problems are far from solved, and many more lie just over the horizon
• 14 January 2026 · TransformerWhy no one can agree on what AI will do to jobsWill AI repeat history — or break it?
• 9 December 2025 · TransformerWhy AI reading science fiction could be a problemThe theory that we’re accidentally teaching AI to turn against us
• 28 October 2025 · TransformerWhere have the really big AI models gone?Why the race to scale up pretraining isn’t over
• 29 September 2025 · TransformerWhen AI starts writing itselfWhy automating AI R&D could be the most dangerous milestone yet
• 21 May 2025 · TransformerMisaligned AI is no longer just theoryA host of new evidence shows that misalignment is possible — but it's unclear whether harm will follow
• 15 May 2025 · TransformerWe need to know what’s happening with AIWithout transparency, society is flying blind toward potentially catastrophic AI capabilities
• 20 March 2025 · TransformerThe quest to build better defenses for AI risks'Societal resilience' measures might offer some protection to the proliferation of dangerous AI capabilities
• 27 February 2025 · TransformerAI coding tools are quietly reshaping software developmentThey’re making an economic impact, despite not being very good yet
• 13 February 2025 · TransformerDecentralized training isn't a policy nightmare — yetGovernments should still be able to keep track of who's training frontier models — though a shift to reinforcement learning could make that harder
• 22 January 2025 · TransformerThe way we evaluate AI model safety might be about to breakAs systems become more capable, researchers think we need a new type of safety evaluation
• 20 November 2024 · TransformerSynthetic data is more useful than you thinkFears of ‘model collapse’ are overstated — at least for now
• 7 November 2024 · TransformerWhy AI companies are eyeing the Middle EastAnd what the US government is doing about it
• 21 October 2024 · TransformerIs OpenAI being fair to its non-profit?Experts fear OpenAI’s conversion to a for-profit might not adequately compensate its charitable arm
• 30 July 2024 · TransformerAI companies are falling short on their promises to the White HouseA year on from making voluntary commitments to the White House, adherence is patchy