Ask the Mates anything | MOONSHOTS AMA #289

Salim argued that Australia can leapfrog established AI approaches rather than assume it is behind, with mindset potentially a greater barrier than infrastructure. He recommended reserving 10–20% of work time for unscheduled experimentation, trying tools and meeting people instead of waiting for spare time. Another speaker cautioned that adoption cannot be forced: people should be ready when colleagues recognize what AI makes possible, though some companies may take decades to change. Personalized AI could filter and summarize information according to an individual’s needs and knowledge.

The speakers described brain-computer interfaces as a possible route to direct interaction with AI, alongside full-dive VR and, in Alex’s view, possible uploading of humans and non-human animals over the next few years. Peter said he invests in and advises three BCI companies and is meeting with a fourth company’s CEO. On consciousness, speakers considered both emergence from complexity and the idea that people receive or localize consciousness; Alex said there is not enough knowledge to determine whether AI could be conscious or to define the question rigorously, but expected neuroscience to give researchers more confidence within five years. He also argued that reasoning models should not be dismissed as mere prediction machines: reinforcement fine-tuning on reasoning challenges does not rule out novelty or insight.

Alex saw overregulation, rather than underregulation, as the main current risk in US federal AI policy, warning that it could hamper competition with China and progress on problems such as medicine. The public-sector question was framed as identifying and mitigating risks, not creating regulation. One speaker advised risk specialists to build credibility in a defined area, red-team scenarios continuously with a team, and publish named analyses of likely threats, noting that governments often act only after a crisis. Speakers also warned that AI agents could disrupt a power grid or break a bank, while suggesting that defensive co-scaling could reduce software vulnerabilities as AI rewrites code originally written by humans.

Dave advised hardware founders to demonstrate that they can build, raise enough funding to reach a prototype or market beta, and assemble a strong team. He expects VCs to become more open to hardware and sees it as a major investment focus over the next five years. For construction drones, he suggested either building a syndicate of VCs and industry partners or joining an established Southern California rockets or drones company for six to nine months before spinning out; founders should view themselves as intelligence companies whose products contain physical components. A civil engineer with 30 years of project records, from problem definition through construction and outcomes, said Mercor and Scale had expressed interest but no deal had emerged after weeks of discussion. Dave recommended packaging the data and due diligence for rapid review, proposing a price, and meeting acquisition decision-makers in person; a company that owns the data may be easier to buy than a project requiring prolonged negotiations.

Education and care were recurring concerns. Patrick questioned whether schooling should focus more on children’s interests and purpose, observing that classrooms for five- and six-year-olds look much as they did in the 1960s. William argued that AI could help a child learn in an hour what might take a classroom a full day, and proposed starting with a problem that excites children before teaching the skills needed to address it. TKS.world was described as a program for ages 13–18 focused on purpose, mindset, AI tools, and entrepreneurship. For exhausted caregivers, a daily log followed by asking an LLM how to make the day more efficient or joyful was suggested as a starting point before considering robotics. Rod McDonald hopes AI can improve disability services, including communication for people with severe motor disabilities and support for people with brain injuries; Alex also pointed to BCIs as a potential way to address motor disorders.

Peter said Fountain Life plans to franchise its operating system and AI systems. It has five centers built, two more under construction in the US, and a partnership to place centers in Estates resorts; he said a franchise open call for interest could be six to nine months away and that he had discussed a center in Greece. Dave said iLabs may offer franchise and data-collection opportunities with ready-made playbooks, estimating roughly 10 to 100 such opportunities this year. Elsewhere, Scott described his Louisiana business as AI-native and focused on data-center cooling bottlenecks, including CDUs and TCSs; he said its work had appeared in four Open Compute Project papers.

The speakers portrayed the current upheaval as roughly 20 simultaneous “Gutenberg moments,” including AI and life extension, forcing institutions built for scarcity to adapt to abundance. They estimated that uncertainty began about 5–10 years ago and could last another 10–20 years, with the aim of minimizing the trough; they acknowledged rough periods, possibly including civil unrest, while remaining optimistic about humanity’s long-term prospects. They compared present AI fears with past anxieties about books and nuclear annihilation, and said superintelligence may not be humanity’s last invention if it helps create what comes next. On competing claims about AI and physics, Griff D proposed comparing models given identical information frozen at a chosen time against later events; a respondent disputed that human-derived priors are a serious obstacle, citing prediction benchmarks and algorithms such as ACSI.