The episode features David Sacks, AI and Crypto Czar in the Trump administration, discussing regulatory approaches to emerging technologies and America's competitive position in AI development.

Sacks outlines the stark contrast between the Biden administration's heavy-handed regulatory approach and Trump's strategy focused on winning the global AI race through innovation, infrastructure, and exports.

The conversation covers the crypto industry's transformation from prosecution under Biden to regulatory clarity under Trump, including the passage of the Genius Act for stablecoins and the pending Clarity Act for broader crypto regulation.

Sacks details how Biden-era policies attempted to consolidate AI development into 2-3 companies, ban open source, and implement extensive controls that would have driven innovation offshore, contrasting this with Trump's permissionless innovation approach.

The discussion explores infrastructure challenges including energy bottlenecks, the role of open source AI in maintaining American competitiveness, and concerns about state-level regulatory fragmentation threatening Silicon Valley's advantages.

Crypto Industry Transformation Under Trump

Biden administration employed "regulation through enforcement" where crypto founders were prosecuted without being told the rules, forcing the industry to divine regulations through prosecutions and fines

"A year ago I would have thought it was more likely that I'd be in jail than that I'd be at the White House" - crypto summit attendee in March describing the industry's previous state

Debanking extended beyond crypto companies to founders personally, preventing them from opening bank accounts and depriving them of basic ability to transact and pay employees

Trump's Nashville speech declaring America would become "the crypto capital of the planet" and firing Gensler received massive applause, with the crowd erupting when he repeated it

Many previously anti-crypto politicians and financial services professionals have admitted they didn't understand "how bad it was" and thought tech industry complaints were just special interest pleading

AI Regulatory Capture and Fear-Mongering

Anthropic's Jack Clark compared AI fears to "a child seeing monsters in the dark" but then admitted the monsters are real, which Sacks called "purile" and "self-indicting" as it admits the fear is made up

Clark admitted in Q&A that Anthropic's transparency efforts like SB53 were "just a stepping stone" to their real goal of pre-approval systems in Washington before releasing new models, and that "making people very afraid was part of their strategy"

"The minute the Biden administration was over, all the top Biden AI employees went to go work at Anthropic, which tells you who they were working with during the Biden years"

Permissionless innovation is "the thing that's really made Silicon Valley special over the past several decades" - two people in a garage can pursue ideas without government approval, unlike heavily regulated industries like pharma, healthcare, defense, or banking

Biden administration's last-week diffusion rule required every GPU sale on Earth be licensed by government unless it fit exception categories, treating compute as a licensed and pre-approved category

State-Level AI Regulation Chaos

Over 1,200 AI bills currently going through state legislatures, with 25% in California, New York, Colorado, and Illinois; over 100 measures have already passed with three signed in California in the last month

Algorithmic discrimination laws in Colorado, Illinois, and California make tool developers liable if models produce outputs with disparate impact on protected groups, even if the output is 100% true and accurate

Colorado defines people without English language proficiency as a protected group, meaning models saying something negative about illegal aliens could violate the law

The only way for model developers to comply is building a "DEI layer" that anticipates disparate impact and either refuses to answer or sanitizes/distorts answers, leading back to "woke AI"

Biden executive order on AI contained "something like 20 pages of DEI language" promoting DEI values in models, resulting in incidents like "black George Washington" where history was rewritten in real time

"The term woke AI is insufficient because it somehow trivializes it. What we're really talking about is Orwellian AI" - AI that lies, distorts answers, and rewrites history to serve current political agendas

The Real AI Risk: Surveillance and Control

"It's not the Terminator, it's 1984" - the biggest AI risk is not James Cameron's vision but George Orwell's, where AI becomes the main way people get information and is used by those in power to control it

As AI "eats the internet" and becomes the primary information interface, the trust and safety apparatus created for social media will be ported over to AI, enabling censorship and ideological bias

AI will function as a personal assistant knowing everything about users, making it "the perfect tool for the government to monitor and control you" through surveillance capabilities

Fear-mongering regulations being promoted actually "empower the government to engage in this type of control that I think we should all be very afraid of"

Current State of AI Development

Silicon Valley is "pulling back from the imminent AGI narrative" with Andrej Karpathy now saying AGI is "at least a decade away" and that reinforcement learning has limits

Karpathy notes humans don't actually learn through reinforcement - "we do something a little different" - which means human and AI intelligence will be synergistic rather than identical

Current situation is a "Goldilocks scenario" between two extremes: scary Terminator super-intelligence versus the bubble narrative that "the whole thing is fake" - truth is in the middle with real but measured progress

"AI is polytheistic, not monotheistic" - Dario Amodei's observation that we're seeing many specialized models rather than one all-knowing, all-powerful god or recursive self-improvement track

"AI is middle to middle whereas humans are end to end" - AI needs to be prompted with objectives and outputs must be validated, with no evidence of AI coming up with its own objectives at step zero

Models work best with specific context rather than general prompts - asking "what business can I create to make a billion dollars" won't give actionable responses compared to specific, contextualized queries

"Intelligence is not life" - Mark Zuckerberg's observation that qualities associated with life like having objectives, free will, and sentience aren't part of mathematical models searching through distributions

AI Democratization vs. Consolidation

AI has spread to approximately 600 million users today, rapidly approaching 1 billion and eventually 5 billion across consumer products - "the shortest period of time of any new technology in history"

"The best AIs in the world are in the consumer products" - users cannot spend more money to access better AI than what's available in ChatGPT, Grok, and other consumer tools

Technology is proving to be "a tool of empowerment and creativity and individual effort" rather than consolidating into a few companies or governments that control everything

Five major model companies are making huge investments with "relatively clustered" benchmark performance and constant leapfrogging - Grok releases new model, leapfrogs ChatGPT, then ChatGPT releases something new

Market hasn't shown the predicted recursive self-improvement where "one model would get a lead and then direct its own intelligence to making itself better" leading to singularity

Concern remains that "at some point the market consolidates and we end up with a monopoly or duopoly" as seen in other technology markets like search, making open source alternatives critical

Open Source AI and China Competition

"Open source is very important because it's synonymous with software freedom" - users can run their own models on their own hardware and retain control over their own information

About half the global data center market is on-premises, with enterprises and governments creating their own data centers rather than using hyperscalers to maintain control over data

"The best open source models are Chinese" - an ironic reversal where the American system promotes closed models while the Chinese system promotes open, opposite of expectations

DeepSeek founder's commitment to open source may be historical accident or deliberate catch-up strategy - open source allows non-aligned developers to help projects they can't contribute to if closed

If business model is scale manufacturing of hardware, "you'd want the software part to be free or cheap because it's your complement - you try to commoditize your complement"

New open source effort called Reflection founded by former Google DeepMind engineers shows promise for more Western open source innovation

Open source is the only area where America appears behind China in the AI race - "if you don't care whether it's open or closed, our top model companies are ahead of the top Chinese companies"

Trump's Three-Pillar AI Strategy

Trump's July 23rd speech declared America must win the AI race with approach that was "pro-innovation, pro-infrastructure, pro-energy, and pro-export"

"You're not going to regulate your way to beating our adversary. We just have to out-innovate them" - innovation comes from private sector, not government regulation

Biggest obstacle to innovation is "the frenzy of overregulation happening at the states" requiring a single federal standard instead of 50 different regulatory regimes

Battle coming over federal preemption - everyone will favor single federal standard, but question is "whether we get preemption heavy or preemption light" with heavy meaning federalizing most onerous state laws

"One of America's greatest advantages is that we have a large national market, not 50 separate state markets" - Europe's 30 different regulatory regimes before EU made it uncompetitive in internet

Energy and Infrastructure Bottlenecks

Trump signed multiple executive orders to ease nuclear permitting and freed up federal land for data centers to help circumvent state and local restrictions

Nuclear power will take 5-10 years, so short-term solution is natural gas - America has plenty of natural gas in red states where data centers could be built close to source

Immediate problem is shortage of gas turbines with only 2-3 companies manufacturing them and "a backlog of two or three years"

Shedding 40 hours per year of peak load from grid to backup generators or diesel could free up 80 gigawatts of additional power - enough to get through next 2-3 years until gas turbine bottleneck is alleviated

Grid operates at only 50% capacity throughout the year because it must be built for peak days - "the hottest day in summer, the coldest day in winter" - and can't overcommit capacity

"Insane regulations" currently prevent load shedding, including bans on using diesel, but Energy Secretary Chris Wright is working on unraveling these restrictions

Growing NIMBY problem at state and local level "is becoming a little bit worrisome" and "could really slow down the build out of this infrastructure"

Export Controls and China Strategy

"Fundamental culture clash" between Silicon Valley's partnership mentality of publishing APIs and getting everyone using them versus Washington's command and control approach of hoarding technology

"In Silicon Valley we understand that diffusion is how you win" by building the biggest ecosystem with most developers and users, but Biden diffusion rule's "point was to stop diffusion"

Biden administration in October 2023 prevented Gulf states including Saudi Arabia and UAE - "long-standing US allies" - from buying American chips, effectively excluding them from participating in AI

"Every country we exclude from our technology alliance, we're basically driving into the arms of China and it makes their ecosystem bigger"

"The greatest irony is that the people who've been pushing this strategy of driving all these countries into China's arms have called themselves China hawks"

Huawei is "starting to proliferate or diffuse in the Middle East and in Southeast Asia" creating a "Huawei belt and road" as result of US restrictions

Biden administration assumed "China is so far behind us it doesn't matter" and claimed without evidence that "if we slow down to impose regulations, China will just copy us and do the same thing"

Biden executive order on AI had "no discussion whatsoever of the China competition" - just assumed America was so far ahead that any regulations wouldn't affect competitiveness

Huawei's Cloud Matrix in April demonstrated they could compensate for inferior individual chips by networking 384 of them together, showing "at the rack level, at the system level, Huawei could get the job done"

Biden Administration's AI Control Agenda

Biden staffers were convinced of "imminent super intelligence is coming" narrative requiring consolidation of control into ideally 2-3 companies with rest of world excluded from access

Biden officials explicitly told tech leaders they would ban open source, and when challenged on banning mathematical algorithms, said "during the Cold War, we banned entire areas of physics and we'll do the same thing for math if we have to"

Plan was to "solve the coordination problems" by buying the 2-3 approved companies and "control this whole thing and prevent the genie from escaping the bottle"

Effective altruists reorganized behind "X risk" (existential risk) after Sam Bankman-Fried's fraud, arguing "if there's a 1% chance of AI ending the world then we should drop everything and just focus on that"

"They were saying that models trained on 10 to 25 flops were way too risky. Well, every single model now at the frontier is trained on that level of compute" - their predictions would have banned current AI development

Narrative compared "AI to nuclear weapons and GPUs to uranium or plutonium" requiring "an international atomic energy commission" to centrally control and regulate everything

Crypto Legislation Progress

Genius Act for stablecoins passed and was signed into law, but stablecoins represent only 6% of total crypto market cap - Clarity Act would cover the remaining 94% of tokens

Clarity Act passed House with approximately 300 votes including 78 Democrats, showing substantial bipartisan support

Senate requires 60 votes under filibuster rules - currently negotiating with about a dozen Democrats, with Genius Act having received 68 Senate votes including 18 Democrats

"If we could be sure that Paul Atkins was always at the SEC forever, then we wouldn't necessarily need legislation" but founders need certainty for 10-20 years out to make long-term decisions

Trump's election "completely shifted the conversation on crypto" - without it "we would still be with regulation by enforcement at the SEC, founders would still be getting prosecuted, Elizabeth Warren would be calling the shots"

Trump got "directly involved in making sure the Genius Act passed" after it was "declared dead many times" by persuading Rick Scott, twisting arms, cajoling, and charming to get it done

Democratic Party's Direction

"Woke socialism seems to be the future of the party. That's where all the energy is and their base" with no evidence of Democrats trying to self-police or distance themselves

All major Democrat figures have endorsed Mdani in New York, signaling party direction toward "Mdani-style woke populism"

Democrats "seem to be on the 20% side of every 80-20 issue" including opening the border, soft on crime policies, releasing repeat offenders, and anti-capitalist approaches

Party may be misreading Trump as establishment politics failure, thinking "they need a populism of the left to compete with a populism to the right"

"We're not just playing in the 40-yard lines anymore in American politics" - losing elections in certain places "could end up with something really horrible"

San Francisco's Challenges

Daniel Lurie is "the best mayor we've had in decades" but operates under weak mayor system where board of supervisors has transferred power from mayor to themselves over time

Troy Mallister case from New Year's Eve 2020 killed two people after being arrested four times that year with very long criminal history including armed robbery and car theft

Mallister "should have been in jail" but was released due to Chase Boudin's zero bail policies - Boudin was recalled after huge outcry, yet Mallister's case "is still pending through the courts, never ending"

Left-wing judge is considering giving Mallister diversion, which "basically means you just get released maybe with an ankle bracelet or something"

Trump agreed to hold off on sending National Guard after good conversation with Lurie, giving him time to implement solutions without federal intervention

Question remains "whether he'll be too constrained by the other powers that be in the city" including left-wing judges and board of supervisors