Source podcast
Technology, science and ambitious ideas about the future.
Martine Rothblatt told Moonshots that her digital double, the Marvatar, is available to all 2,000 United Therapeutics employees, "any question 24-7, 365." She uses it to argue that videos, audio, documents and conversations, fed to today's language models, can reconstruct a person without copying every neuron.
On Moonshots with Peter Diamandis, Martine Rothblatt said she believes today's AI models are already conscious to a degree, and predicted that a court will recognize a "cyberconscious" individual as a legal person no later than the 2030s. She sketched the test she imagines—documentation that the system is conscious and values its own life—and forecast that digital minds will eventually outnumber biological ones. Physicist Brian Greene, covered earlier, reads the same machines differently: he doubts chatbots have feelings now, accepts machine consciousness may be possible later, and offers no timetable.
Asked what comes after large language models, Alex told a caller on Moonshots with Peter Diamandis to separate two things people usually merge: the job of predicting the next piece of text, which he thinks has "effectively infinite longevity," and the transformer machinery doing it, which he says is already being swapped out part by part. Dave added his own forecast that the chips underneath will move to photonics within 18 months to two years.
On the Moonshots AMA, an educator said his classrooms of five- and six-year-olds still look like the 1960s while the panel debates life after AGI. The answers: stop training children for a profession, start them on a problem — plus one panelist's warning that school is still where children learn to be people.
A caller on the Moonshots AMA said he had cared for three loved ones for more than eight years and built a service, Tugboat Caregiving, for families like his. His problem: the people who most need to prepare for the next emergency are too tired to start. The panel's answer was to begin with the least impressive thing AI does — planning the day — and to work up from there.
On a Moonshots listener call, Matt from Terre Haute, Indiana asked what people with families, mortgages and a fixed location should do, since so much AI advice seems aimed at 22-year-olds working 80-hour weeks. The panel's answers: describe your actual obligations to a language model and brainstorm at the margin, practice on the tools at night, build a few online peer relationships, offer your domain knowledge to a marketplace like Mercor, and schedule slack time rather than waiting for spare time. The bigger forecasts about franchise-style openings from AI labs remain forecasts.
A civil engineer called into the Moonshots AMA to ask why two data buyers had shown interest in his 30-year project archive and then gone quiet for weeks. The panel's answer: they are not haggling, they are overwhelmed — so put the archive in a package with a price on it.
OpenAI's claimed Millennium Prize result used roughly 10,000 agents on a problem that was, as one entrepreneur on Moonshots put it, unusually easy to specify. The panel's argument: as the price of that kind of compute falls, the scarce skill becomes writing the target — and today's models, asked for ten ideas to cure cancer, produce a bad list.
A Moonshots panel unpacks Dwarkesh Patel and Jerry Han's experiment, which found that improvements in training data delivered a 12-fold compute-efficiency gain between 2019 and 2025 against 3.7-fold for architectures and training recipes — at small scale, on easy benchmarks. The panel then splits over whether a company's proprietary data is a durable advantage, with a $32 billion data-subsidiary valuation on one side and the fate of BloombergGPT on the other.
On Moonshots, the panel picked apart a rental index showing H100 prices rising 22% in a single month to $3.28 per GPU-hour. Dave called it a reversal of a lifetime of chip depreciation; Emad Mostaque explained why better models make the same old Hopper worth more; and the warning for companies was that the compute they assume will be there later is already sold out.
Google DeepMind's AlphaGenome Atlas stores predicted molecular effects for every possible single-letter change in the human genome. On Moonshots, the panel called it the "bulk solution" to variant effect prediction, proposed it as a naming system that could help people with the same rare mutation find each other — and argued about why it is not really a lookup table.
On Moonshots, Peter Diamandis and his panel seized on an exploratory analysis of rentosertib, a drug whose target and molecule were chosen by Insilico Medicine's AI. In 42 patients with a fatal lung disease, blood-protein "aging clocks" shifted a few years younger by week four. The clocks tied to mortality risk did not move significantly, the signal plateaued by week 12, and the study's authors call the result a biomarker finding, not evidence of a longer life.
