Tag
Briefings and ideas about Transformers from AI podcast discussions and the research they cite, with links to the original sources.
Prosecutors allege an OpenAI engineer bugged an executive conference room. On Moonshots, Chamath Palihapitiya said labs should assume constant leaks, and Alexander Wissner-Gross predicted a Cold War over superintelligence.
Biohub's head of science, Alex Rives, says virtual cell models fall short above all because biology can't yet measure what matters. He expects an information-level cell model within about five years.
Dyna Robotics chief scientist Jason Ma argues that robots should predict future camera video, not only their next moves. He says the bet could fail if touch or sound makes rival models far more data-efficient.
Goodfire researcher Eric Bigelow resampled a GPT-3.5 model's reasoning at every token and saw answers collapse at single tokens, some arbitrary. He argues the "decision" happens in sampling; newer reasoning models often fork more smoothly.
Biohub's Sal Candido and Google DeepMind's Pushmeet Kohli argued on a panel that scaling laws in biology must be found, not assumed, and that the problem should decide whether to build better models or gather new data.
Positron CTO Thomas Sohmers said AI agents are multiplying demand for memory. He claimed Positron's first-generation product sustained 93% of theoretical memory bandwidth, against an average of 30–40% for NVIDIA GPUs when decoding with transformer models. Its next chip, Asimov, is planned for 2027.
Reflection AI's open-weight Beam is built for efficiency, by the company's own estimate. On Moonshots, Alexander Wissner-Gross said US open labs still lack a business model that would pay for frontier capability.
TypeSafe says its Jev model returns typed choices or scores in 70–500 milliseconds in its own tests. On Moonshots, panelists said such models could handle many small organizational decisions and noted open-source versions exist.
MIT researcher Alex Zhang argues that TypeSafe's Jev, which returns fast typed decisions instead of prose, matters as a test of new language-model designs, not as a mere classifier. He says the open question is how it was trained.
Gladstone AI's Edouard Harris says friendly Chinese rhetoric must be weighed against Beijing's track record. A Chinese lab or regulator slowing down after its own AI incident would be a much harder signal to fake.
Alexander Wissner-Gross argued AI models are a weak form of mind uploading. The closest matching study, from 2022, found GPT-2 partly predicted 101 listeners' brain scans, but only correlationally, reaching at most 32% of the noise ceiling.
Robinhood CEO Vlad Tenev argued AI research is easier to automate than consumer apps because its experiments have ready-made evals, while app changes must wait for enough users to show statistically significant gains.