Sep 22, 2026 · 1h 26m
Moonshots panel tests abundance against AI’s hardest risks
Ask the Mates Anything Round #2 | MOONSHOTS AMA #293
The discussion connects AI’s promise of abundance to the institutional, economic, security, and human risks that could determine whether it arrives safely.
- 1AI alignment requires progressive real-world testing, not simply isolating powerful systems or slowing development.
- 2Automation may create enormous value, but ownership and entrepreneurial leverage will determine who captures the gains.
- 3Abundance depends on rebuilding education, governance, healthcare, monetary systems, and other institutions for continuous AI-driven change.
Don't miss
The panel turns its abundance thesis into an institutional challenge, arguing that education, governance, healthcare, law, and monetary systems must be rebuilt for AI-driven change.
The brief
Peter Diamandis and the Moonshots panel frame AI, robotics, biotechnology, and space technology as tools for expanding human capability—but insist that alignment, sovereignty, and social disruption remain unresolved.
The sharpest disagreement concerns AI safety: progress should continue, but powerful systems may need progressive release and real-world interaction rather than either reckless deployment or laboratory isolation.
Across education, work, and entrepreneurship, the panel argues that AI productivity will reward ownership and adaptable builders, while legacy institutions risk capturing gains or failing to keep pace.
The conversation widens from lunar manufacturing and robotic food infrastructure to AI-native customer interfaces, cybersecurity liability, and language learning built around authentic human interaction.
The standout conclusion is institutional: abundance is not automatic, and realizing it requires rebuilding education, governance, healthcare, law, and monetary systems around rapid feedback loops.
What was said on this episode
46 statements · 33 positive · 9 negative · 2 mixed · 2 neutral
Humanity will overcome major AI-related risks within ten years.
“Humanity will put those risks behind us definitely within 10 years.”
Listen at 0:43
Internet-trained foundation models are already a weak form of human mind uploading.
“large language models, foundation models trained off of human behavior on the internet, are already a weak form of human mind uploading.”
Listen at 3:50
Only other AI systems will ultimately be able to police AI.
“nothing's going to ever be able to police AI other than other AI.”
Listen at 5:55
AI alignment is relatively straightforward if internal model states are observable.
“I don't think it's as hard a problem as people think if you can see into the brain.”
Listen at 6:29
Keeping powerful AI capabilities isolated in labs is highly unsafe.
“one of the worst possible outcomes for AI safety is to have strong and new capabilities bottled up inside the labs”
Listen at 6:37
Many AI education projects can be launched immediately without external funding.
“you can just go and do it right now without funding.”
Listen at 9:08
AI systems will become increasingly energy efficient.
“AI is going to get more and more energy efficient.”
Listen at 12:42
Ireland is an excellent test case for experimenting with AI-era governance.
“Ireland is an incredible test case”
Listen at 17:32
AI enables experimentation with thousands of governance models.
“it is very possible to experiment with thousands and thousands of different ways to manage and govern in the age of AI.”
Listen at 18:33
Lunar industrialization should develop a self-contained native industrial ecology.
“what I would most like to see from anyone wanting to help disassemble the Moon is a native industrial ecology”
Listen at 21:11
Self-replicating lunar machine shops can use local resources and energy to reproduce.
“a machine shop on the Moon that is able to make copies of itself using only native resources and solar or other energy that are native to the Moon.”
Listen at 21:27
Self-replicating machines are the right path for off-world manufacturing.
“self-replicating anything is the right path.”
Listen at 22:03
Software opportunities are declining while hardware opportunities will persist for years.
“Software is cooked, but hardware will go for many, many years.”
Listen at 23:05
Drone delivery will arrive imminently.
“Drone delivery is imminent”
Listen at 24:27
Human lifespan will double within five to ten years.
“we're going to double the human lifespan in the next 5 to 10 years.”
Listen at 27:27
All disease will be cured within ten years.
“curing all disease in the next 10 years.”
Listen at 27:33
Most AI-generated economic value will accrue to owners rather than payroll.
“almost all value is going to accrete to capital gains through ownership and equity and not to payroll.”
Listen at 30:45
Generative-AI films will succeed primarily through strong storytelling.
“anything that's going to be successful is a great story.”
Listen at 31:50
AI filmmakers should release many microdramas and iterate from audience response.
“Don't release one movie, release 1,000 microdramas”
Listen at 33:43
Traditional university curricula are becoming obsolete.
“the curriculum is clearly going away.”
Listen at 36:48
Major research universities should be restructured as efficient public-benefit corporations.
“I'd love to vivisect them, turn them into the for-profit public benefit corporation at best operations they actually are”
Listen at 37:31
The traditional university model is breaking down.
“The traditional model of a university breaks.”
Listen at 38:37
Managing AI agents offers a large but temporary opportunity.
“there's a massive opportunity in management of agents that is wide open for some period of time, but certainly right now.”
Listen at 40:37
Human-management principles immediately apply to managing AI agents.
“everything I learned about management of people immediately applies to management of agents”
Listen at 40:44
AI can productively orchestrate human activity for at most ten years.
“AI can usefully and productively orchestrate human activities. Call it a few years at most, 10 years maximum.”
Listen at 45:35
Human-machine merger will become necessary for human inputs to remain economically relevant.
“after that, human-machine merger in order for human inputs to remain economically relevant.”
Listen at 45:45
Language-learning businesses could scale to hundreds of billions by becoming life-coaching platforms.
“that business model scales to many hundreds of billions of dollars when it moves from language to life plan.”
Listen at 49:32
AI-enabled cyberattackers will arrive within one to five months.
“the attackers are coming, you know, 1 month, 3 months, 5 months from now”
Listen at 51:34
AI models will discover a flood of software vulnerabilities.
“there's just going to be a flood of vulnerabilities”
Listen at 52:28
Consumer-service websites should provide AI-forward XML interfaces.
“all of those websites need to have XML interfaces that are AI, you know, AI forward.”
Listen at 54:46
Reversing gene expression could reverse some inherited retinal disease effects.
“that should reverse, right? This is the exact work that David Sinclair is doing.”
Listen at 57:34
Future brain-computer interfaces may enable ultraviolet and infrared vision.
“in the BCI path in the future, you'll not only be able to see in visual spectrum but ultraviolet, infrared.”
Listen at 58:18
Superhuman vision will be available within a few years.
“I would count in a few years on having superhuman vision.”
Listen at 58:42
Fine-tuning is especially suitable for creating culturally specific AI companions.
“This is a poster child for fine-tuning.”
Listen at 1:02:44
The GLM model offers flexibility for fine-tuning AI companions.
“the new GLM model, which has fewer parameters, but it's still a very long context window, which gives you more flexibility.”
Listen at 1:03:32
Most future AI applications will produce radically positive outcomes.
“the vast, vast majority will be radically positive.”
Listen at 1:06:40
The probability of a dramatically better AI future exceeds 90% after ten years.
“on the time scale of greater than 10 years, EZoomT is greater than 90%.”
Listen at 1:07:06
The probability of a highly positive AI future is 99.9% if near-term risks are overcome.
“I'm 99.9% P-bloom.”
Listen at 1:07:45
AI can reduce manufacturing operating costs by 10–30% through administrative automation.
“we could drop 10%, 20%, 30% more to the bottom line with stuff that AI can do right now.”
Listen at 1:10:59
Mainstream media will remain structurally broken because of financial pressures.
“the media is always going to be broken from here forward because they're starved for money.”
Listen at 1:14:57
Narrowcast media is the antidote to mainstream media dysfunction.
“the antidote to mainstream media is narrowcasted media”
Listen at 1:16:05
Europe is currently constrained by internal and external factors.
“Europe has been hobbled both due to external factors and internal factors.”
Listen at 1:17:00
European AI builders should solve energy and data-center constraints or relocate.
“either solve the energy plus data center crunch together with the, the concomitant policy issues, or just move”
Listen at 1:18:33
Multi-party computation is unlikely to be the main constraint on international peace.
“I don't think that's the limiting factor for, say, international peace.”
Listen at 1:22:26
Resolving Taiwan and domesticating supply chains would promote international peace.
“Solve the Taiwan issue and re-domesticate all supply chains everywhere to every country”
Listen at 1:22:51
Eliminating dependence on global trade would produce a more peaceful world.
“obliterate— as perverse as this sounds— obliterate, cook, incinerate the need for global trade in products and services.”
Listen at 1:23:37
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Peter H. Diamandis
The Diamond Age
David Sinclair
European UnionBooks & mentions
Listen to the full episode and explore every guest, topic, and moment on PodLume.