2025-04-02

AI Experts Q&A: How Humanoid Robots Will Impact Every Industry | EP #160

Live audience Q&A from the Abundance Summit with Peter Diamandis moderating Figure CEO Brett Adcock and investor Vinod Khosla. Questions range across fusion energy siting, Figure's humanoid robot roadmap and chassis strategy, UBI and job displacement, humanoids in space and agriculture, robot cybersecurity, and startup capital strategy. Khosla closes with predictions on near-free healthcare expertise, AI model explainability, and AI multiplying the world's effective scientific workforce 10-100x within five years.

▶ Watch on YouTube

Topics

Fusion energy roadmap and siting bottleneck Energy ▶ 1:34

Khosla names Commonwealth Fusion, Helion, and Realta Fusion as high-credibility efforts and sees a path to 5,000 fusion plants by 2050, but says regulatory/siting delays now outpace the technology development itself.

Figure's humanoid scaling and chassis strategy Robotics ▶ 3:44

Adcock explains Figure's 100,000-robot shipping milestone and the decision to keep one general-purpose hardware platform rather than customize per customer, to hit a ~$20,000 cost target.

UBI and AI-driven job displacement Economy ▶ 9:35

Khosla argues AI will bring abundance and GDP growth alongside rising income disparity, and that unmodified capitalism will need policy changes, decided country by country.

Humanoids in space exploration Space ▶ 12:39

Adcock and Khosla discuss robots as a natural fit for hostile space environments; Diamandis recounts asking Sam Altman about self-replicating AI probes leaving Earth.

Robots in agriculture Robotics ▶ 15:12

Panel discusses farm robots starting indoors before moving outdoors into variable weather, and the gap between demonstrating a capability and real adoption.

Diamandis's personal health stack Longevity ▶ 18:11

Sponsored segment on Diamandis's use of Fountain (full-body diagnostic uploads), Viome (metatranscriptomics-based nutrition), and OneSkin (peptide skincare).

Figure's military policy and robot cybersecurity Robotics ▶ 21:38

Adcock states Figure has a no-military-use mandate since 2022, and that its robots run fully embedded on local GPU hardware without needing network connectivity, limiting hacking risk.

Startup capital strategy Economy ▶ 26:09

Khosla says staged fundraising produces better outcomes than large upfront raises, because scarcity forces companies to test assumptions rather than get complacent.

AI investment market structure Economy ▶ 28:01

Khosla expects most AI investments to lose money (~80%) but net gains to exceed losses due to exponential winner outcomes; sees a diversified market with several winners per category, not winner-take-all.

Robotic surgery and healthcare automation timeline Health ▶ 30:52

Khosla predicts diagnostic/consultative healthcare expertise will be effectively free before 2030, while invasive procedures and unsupervised robot surgery remain 10-20 years out due to FDA regulation.

AI safety and model explainability AI ▶ 31:35

Khosla acknowledges AI safety concerns raised by Yoshua Bengio and Geoffrey Hinton, but expects major explainability breakthroughs within two to three years.

AI-driven science acceleration AI ▶ 32:56

Khosla predicts AI acting as autonomous chemists, physicists, and biologists will multiply the world's effective scientific output 10-100x within five years, running hypothesis and testing without human intervention.

Predictions made

open Vinod Khosla: There is an easy, achievable path to building 5,000 fusion power plants by 2050.

· due: 2050 · ▶ watch

“I see a easy path to building 5,000 fusion power plants by 2050, maybe more.”

Your call:

open Vinod Khosla: Within the next five years, robots will be demonstrated working capably in farm fields.

· due: 2030 · ▶ watch

“In the next five years we will demonstrate robots can work in the farms.”

Your call:

open Vinod Khosla: Diagnostic and consultative healthcare expertise will become effectively free, well before 2030.

· due: 2030 · ▶ watch

“Healthcare expertise will be free in the next couple of years, like well before 2030.”

Your call:

open Vinod Khosla: Unsupervised robot surgery is more than 10 years away, likely 15-20 years, depending on FDA regulatory pace.

· due: 2040-2045 · ▶ watch

“I expect we're more than 10 years away from a unsupervised robot surgery. Whether we are 15 or 20 years away depends on regulatory.”

Your call:

open Vinod Khosla: Within two to three years, AI models will no longer be considered an unexplainable 'black box' due to major advances in explainability.

· due: 2027-2028 · ▶ watch

“I suspect in two to three years you'll start to hear about this is no longer a black box.”

Your call:

open Vinod Khosla: Within five years, AI acting as autonomous chemists, physicists, and biologists will multiply the effective number of working scientists by 10 to 100 times.

· due: 2030 · ▶ watch

“Within five years we'll multiply that by 10 or 100x why because we will have AI chemists, AI PhD chemists, AI physics biologist.”

Your call:

open Brett Adcock: Figure's CEO anticipates humanoid robots will be present in homes within the next 2 to 3 years.

· due: 2027-2028 · ▶ watch

“Brett Adcock, the CEO of Figure, anticipates they'll have robots in our home in the next 2 to 3 years.”

Your call:

Numbers that matter

Figure's magic number is 100,000 robots shipped Adcock says Figure deliberately picked a minimal number of partner companies to reach 100,000 robots shipped, already hit with its first two groups.
Figure's target hardware cost is ~$20,000 Adcock ties the single general-purpose chassis strategy directly to hitting this cost level.
Commonwealth Fusion is building a 500 megawatt reactor Cited by Khosla while discussing differently sized fusion reactors and grid transmission needs.
Realta Fusion is building 50 megawatt reactors, small enough to fit in a substation footprint Khosla argues smaller reactors sited locally could reduce the need for new transmission lines.
A Fountain diagnostic upload digitizes about 200 gigabytes of personal health data Diamandis describes his annual full-body diagnostic scan used to catch disease at inception.
About 80% of AI investments are expected to lose money Khosla's estimate of the AI investment landscape, offset by exponential returns from winners.
About 95% of deep-tech startups (robotics, EVs, satellites, AVs) go bankrupt Adcock's figure for how rarely capital-intensive deep-tech companies succeed.

Worth digging into

🕳️ Khosla's claim that less-funded startups make more absolute profit

A counterintuitive capital-efficiency thesis that directly contradicts SoftBank-style mega-funding strategy, with real implications for how moonshot ventures should be advised.

🕳️ Sam Altman calling Diamandis's 15-year self-replicating probe estimate 'too conservative'

Implies OpenAI leadership privately expects self-replicating space probes sooner than public roadmaps suggest.

🕳️ Khosla's 2016 AI dystopia-vs-utopia paper and newer 25-page UBI piece

Lays out his full framework for job displacement, income disparity, and policy choices across the 2020s-2040s, referenced but not detailed on stage.

🕳️ Figure's fully offline, embedded-weight robot architecture

Adcock claims the Helix kitchen robots run entirely on local GPU with no network connection, a notable answer to public fear of hacked humanoid robots.