Trying to Save the World Is No Excuse

PodcastMoney StuffSilvia KillingsworthSep 18

The discussion examines legal and competitive constraints on AI development, including the tension between safety concerns, commercial deployment, and the race to build more advanced models. It also considers pension funding risks and a proposed pension-like 401(k), alongside questions about public trading signals, options hedging, and liquidity limits in private-credit funds.

AI Safety Artificial Intelligence Retirement Planning Private Credit Trading Strategies Market Liquidity

Preview · 13 more points in the app

Read the Full Summary

Key Takeaways

Anthropic’s CEO argues that frontier AI development should slow because it could be dangerous, but the speaker says labs face strong pressure to keep advancing: customers want the most capable models, leaderboard position matters, and labs compete for researchers. The speaker also says that if the four leading labs agreed to slow rollouts, the agreement could raise antitrust concerns, making coordinated restraint legally fraught as well as competitively difficult.

The speaker describes an AI race in which training frontier models consumes substantial data, chips, and spending, even though existing models already have commercial uses and wider deployment requires training and support. Anthropic and OpenAI are partnering with private equity firms to help companies deploy AI, while competitive pressure continues to push labs toward training their next models.

The speaker says AI-risk messaging can both signal a company’s capabilities and reflect genuine concern among major-lab employees, while some developers who worry about the risks still want to build AI before Google does. That tension helps explain why safety concerns do not necessarily lead developers to leave the AI race.

GEO-POLITICS

China Crackdown Suit — China’s crackdown restricted lucrative cross-border trades by Chinese brokerages, sending their stocks lower. Susquehanna alleges suspiciously timed put purchases caused losses, but a judge denied its request to freeze the accounts, finding the case could proceed without doing so.

“In May of this year, China announced a crackdown on basically cross border trading from its brokerages…”—

AI Antitrust Risk — The speaker says that if the four leading AI labs agreed to slow product rollouts, their agreement could resemble a classic antitrust conspiracy. That risk could constrain a voluntary effort to slow deployment.

“But if you got, like, the four competitors in a room and they agreed to, like, slow the rollout of their products, that looks like a very classic antitrust conspiracy…”—

TECHNOLOGY

Anthropic’s Slowdown — Anthropic’s CEO published an essay arguing that development of frontier AI models should slow because continued development could be dangerous.

“The CEO of, of Anthropic published an essay over the weekend called pacing the frontier…”—

Safety And Competition — The speaker says AI developers who fear the technology’s risks may still want to build it themselves so Google does not build it first. Safety concerns can therefore coexist with participation in the AI race.

“But they all... Dario Amade especially, their main thing was, like, we're so concerned with how AI is going to potentially ruin everything…”—

AI Risk Messaging — The speaker says AI-risk messaging can also serve as marketing by signaling how powerful a company’s AI is. At the same time, the speaker says most people at major AI labs are genuinely concerned that AI could cause harm.

“Maybe. That's him. Different. It's But do I do I think that most people at the big AI labs are genuinely worried about AI killing us? Oh, yes…”—

ECONOMY

Pension Funding Risk — Defined benefit plans require companies to set aside and invest money for promised future benefits. Poor investment results can leave pensioners without those benefits. Matching future payments with Treasury bonds can be expensive, while stocks may not grow enough to cover the obligations.

“And how it meets those obligations is like sort of in the first instance ops of the company. And so it like sets aside some money and invests it…”—

401(k) Risk — A 401(k) balance can fall with the market without creating the same company-level pension funding gap or risk of pension bankruptcy. The account may still be inadequate to support a retiree’s desired lifestyle.

“At the stocks go down, they go down. But there's no, like, funding gap. Right? Right. You're never like... I mean, there is. Like Sure. There is. Right?…”—

MARKETS

Public Trading Signals — The speaker argues that trading on publicly visible trades or patterns is not insider trading simply because the activity looks suspicious. He distinguishes public market information from confidential information provided by a company insider.

“Like, if you see looking at, like, public stock tapes, if you if you just see a public trade and you're like, that must be insider trading, and then you…”—

Options Hedging — Options market makers commonly hedge put sales by shorting stock, though they can also offset exposure by buying puts. Susquehanna alleged that customers made about $70 million on the puts it sold. The transcript notes that its hedges matter to its net position and the damages calculation.

“Options market makers, like, just naturally immediately hedge. Not always…”—

Susquehanna Options Lawsuit

Susquehanna says suspicious put-option trades around China's brokerage crackdown cost it money, but the court declined to freeze accounts without evidence identifying which traders had insider information.

Options Market-Maker Hedging

Market makers routinely hedge or offset options positions, so Susquehanna's claimed losses on puts may not reflect its net exposure or the full picture of its trading.

Anthropic's Frontier Pause

The hosts discuss Dario Amodei's call to pace frontier AI development and conclude that AI leaders can genuinely fear the technology's risks while still competing to build it.

AI Deployment Versus Training

The hosts argue that existing AI models still have substantial commercial potential, but labs continue spending heavily on new models because competitive pressure rewards frontier progress.

AI Competition And Antitrust

Labs may have business reasons to slow model development, yet customer and talent competition make unilateral restraint difficult, while coordination could trigger antitrust scrutiny.

AI Founders' Written Record

The hosts reflect on how human-authored memos shaped the AI race and wonder how those foundational decisions may change as AI increasingly helps produce ideas and writing.

Defined-Benefit Pension Risks

Traditional pensions promised retirement income but were difficult to carry between jobs and exposed workers to funding and investment risks that defined-contribution accounts avoid.

401(k)s And Wall Street

Although 401(k)s broadened Wall Street's retail customer base, low-cost index funds have kept fees low, while pension funds remain valuable clients for private-market investments.

Private Credit Liquidity Limits

Retirement savers may have long horizons, but retail investors still expect access to their money, as private-credit fund redemption limits have made clear.

BlackRock's Pension-Like 401(k)

BlackRock's proposed approach would preserve 401(k) portability while emphasizing professionally managed assets and projected monthly retirement income, potentially using private investments and annuities.