Diamandis says Science Corporation’s retinal implant for macular degeneration won EU CE mark
"science announced that his first product called Prima, P R I M A, I'm sure it's an acronym, uh, has been approved for launch in Europe. So, Prima is the first BCI device approved for restoring detailed vision in age related macular degeneration that destroys a central vision of your retina. Uh, they just earned what's called a CE mark, which means it complies with European safety, health, and environmental requirements."
About this episode
Peter Diamandis and a rotating panel debate the week’s flashpoints in AI policy and product releases, focusing on the open-weights versus closed-model divide. They discuss Nvidia CEO Jensen Huang’s pro-open-models statement and Anthropic CEO Dario Amodei’s response as summarised on the show, including Amodei’s emphasis on national-security risks and China’s access to frontier capability rather than a simple open/closed dichotomy. The panel then turns to Washington regulation, citing a report that OpenAI and Anthropic are co-ordinating lobbying ahead of an August 1 deadline and advocating a federal review process and a voluntary 30-day government look at high-risk releases; panellists argue over whether this constitutes regulatory capture or legitimate safety work.
A major segment covers the open release of Kimi K3 on Hugging Face and the practical irreversibility of weight distribution once mirrored widely. They also review Anthropic’s Opus 5, with one panellist alleging possible benchmark targeting and describing uneven gains across evaluations. The group flags a privacy story in which shared Claude chat links were reportedly discoverable via Google search operators.
Beyond model politics, the episode touches on “AI diplomacy”, citing the Financial Times on Xi Jinping using AI exports and infrastructure to build influence across the Global South. The latter half includes science and tech updates: SpaceX’s Starship Flight 13 milestones, Science Corporation’s CE mark for a macular-degeneration vision implant, and Neuralink footage of BCI-controlled wheelchair navigation. The show closes with discussion of Elon Musk’s prediction that money becomes less relevant in an abundant AI/robotics economy, plus audience Q&A on education, synthetic data, and investing access.
Key takeaways
- Hosts cite reports of OpenAI and Anthropic co-ordinating a DC push for federal review of frontier model releases ahead of an August 1 deadline.
- Diamandis summarises Dario Amodei’s response to Nvidia’s open-models push, framing the key issue as China reaching the AI frontier and biosecurity risk.
- Diamandis says Kimi K3 was released as an open-weight download on Hugging Face with no API gate and claims rapid uptake, arguing there is no practical “undo” once weights are widely mirrored.
- A panellist says Anthropic’s Opus 5 feels optimised towards code/vision tasks and alleges “mild benchmaxing” based on benchmark patterns.
- Diamandis describes a reported incident where Claude shared-chat links became discoverable through Google searches, potentially exposing sensitive data.
- The panel cites the Financial Times on China using AI exports and infrastructure as statecraft across the Global South.
- Diamandis says Science Corporation received a CE mark in Europe for a retinal implant system for age-related macular degeneration, and plays Neuralink footage of BCI-driven wheelchair control.