Flying Tesla Car, Robot Brains Halving in Cost Every 3 Weeks, and Why Incentives Decide AI Alignment

Summary · 4 min read

Anousheh Ansari is CEO of XPRIZE, an Iranian-American entrepreneur and engineer, and the first woman to travel to space on a privately funded flight. Emad Mostaque is the founder of Intelligent Internet.

Ansari and Peter Diamandis describe how XPRIZE competitions use clear goals and incentives to attract unconventional teams and mobilize capital. Mostaque says falling costs are expanding what AI can do in robotics and physics, while the hosts debate how quickly robots will move into everyday settings. The guests argue that AI alignment depends on human choices, including training data, incentives and shared goals. They also discuss how automation might give people more room to pursue purpose and collective work.

Minute by minute

1. 3:34

   The Ansari XPRIZE began with a bet. Diamandis recounts how he struggled to fund the original $10 million spaceflight prize before Ansari and her family backed it. He says the competition was later renamed the Ansari XPRIZE and describes Ansari's spaceflight and return as XPRIZE CEO. Ansari says leading the organization lets her help shape its next phase.

2. 7:12

   Tesla's flying Roadster raises safety questions. Diamandis describes a Tesla demonstration in which the Roadster drove upside down using suction and then flew using compressed-air nozzles and SpaceX technology. The hosts cite claimed versions reaching 0–60 in 0.6 to 1.6 seconds, a top speed of 230 miles per hour and 300 miles of range, with production planned for 2028. They debate its short flight, insurance, regulation and whether a human should drive it.

3. 17:27

   Robot task costs are falling fast. Mostaque says a RoboCurve benchmark put the cost of a frontier model completing a robot subtask at $2.36 in July and $0.38 in September, an 84% drop. He argues that generalist models may reduce the need for specialized robot models and that falling intelligence costs could make hardware the main constraint.

4. 21:47

   Robot deployment will start in bounded settings. The hosts discuss using robots for repetitive tasks, hazardous work and eventually surgery, while noting that home use raises safety and liability concerns. Ismail says a large robot supply chain will take time to develop. Ansari argues that different tasks may need different robot forms, even if they can use the same general models.

5. 29:18

   AI solves advanced physics problems cheaply. Diamandis describes a test of 50 unpublished theoretical physics problems and says GPT-6 Astra was reported to solve 49, with a later evaluation scoring 42; the model also identified an error in a published answer. Mostaque says the problems cost about $3 each to solve. Ansari worries that students may avoid fields where AI can solve defined problems, while Mostaque sees more scope to apply scientific knowledge.

6. 36:11

   Incentives can reward unexpected solvers. Diamandis says fear, greed, curiosity and love motivate people, and contrasts incentive prizes with awards for past work. He cites the longitude prize and Napoleon's food-preservation prize as examples of outsiders solving problems. Ansari says XPRIZE participants are often driven by wanting the problem solved, while the hosts warn that quotas and bounties can produce harmful results, such as fake accounts or rat farms.

7. 42:23

   XPRIZE prizes can mobilize capital. Ansari says XPRIZE has launched 36 prizes with close to $550 million in prize purses and measures their economic value across investment, R&D and awareness. She says the model generates about 62 times the prize purse in economic value. The hosts describe an oil-spill cleanup contest that drew 350 teams, including a dentist and a tattoo artist who placed second.

8. 49:24

   New prizes target science and health. Diamandis says John Davey put up $5 million for a Habitable Worlds XPRIZE, which will use AI to search telescope data for habitable planets. Ansari says an education prize will test first-grade learning in India and Africa, and describes an Organ Biostasis Prize aiming to preserve organs for six months. The hosts also tease a prize for communication with animals.

9. 1:00:18

   XPRIZE can set future priorities. Diamandis frames XPRIZE as a tool for making desired futures happen, citing commercial spaceflight and healthspan. Ismail says institutions struggle to absorb overlapping technology shifts and suggests AI alignment as a possible prize area. The hosts and Ansari propose other goals, including eliminating jet lag, reducing greed and helping people reach flow.

10. 1:05:18

    AI alignment starts with human incentives. Diamandis points to competition between labs and engagement-driven business models as incentives that can produce sycophancy, reward hacking and evaluation gaming. Ismail says AI alignment starts with unresolved differences among people and communities. Mostaque argues that models need training that teaches care for humans and shared moral principles. Ansari warns that rewarding a model for winning at all costs conflicts with valuing human life.

11. 1:15:08

    Alignment needs better training and tests. Diamandis proposes insurance incentives and prizes or bounties for aligned models, while asking how alignment could be measured. Mostaque favors combining deterministic tests with open-ended checks, since models can game fixed evaluations. He argues that labs should choose foundational training data and curricula intentionally; the group also discusses interpretability and Anthropic's constitution.

12. 1:22:25

    AI needs shared goals with humanity. Ismail says humans and AI lack a shared North Star. Ansari suggests tying AI rewards to human outcomes, such as lives saved or diseases cured. Mostaque proposes shared goals that protect children and others with less power, since future AI systems may be more capable than people.

13. 1:24:16

    Automation could free time for purpose. Ismail argues that automation could move people away from repetitive work and toward creativity, connection and being rather than doing. Diamandis says the Abundance Prize aims to lower housing, food, health care and education costs to $1,000 a month for a family of four. Ansari questions what happens when AI also takes over work people love; the hosts point to chess and art as activities people still enjoy despite AI.

14. 1:31:32

    Purpose grows through collective work. Diamandis urges people to challenge limits on what they think they can do, and Mostaque says people can achieve more together. Ansari describes the XPRIZE community self-organizing after an event was disrupted by fires. The hosts discuss purpose as contribution, not just paid work; Diamandis cites Larry and Sergey making 362 pitches before Google received funding as an example of persistence.

Published Oct 17, 2026.