Given the AI arms race with China, it’s critical to establish clear rules of engagement for AI firms in the US. In that context, I agree with preempting the states on AI regulation…but only if the federal government can act swiftly and effectively.

Teaming up with my husband Boris in a “Feldman Interviews Feldman” format, I’ve made a video podcast episode entitled How I used AI in Writing a Book about AI. My most important takeaway: AI serves best as human augmentation, not as human replacement.

Working with Claude, Grok, and other AI programs intensely over the last several months, I have explored where they currently shine and where they struggle.

Under federal law, compounders can help patients adjust doses or ingredients because with medicine, “one size doesn’t fit all.”. But as I mentioned to @SELFmagazine, today’s question is whether those laws still apply when the fit problem is wrong price, not wrong size.

Also, we should consider who benefits from government intervention. We can’t win the AI war by handicapping our top players or dividing capacities that give the US an edge. More at:

Is AI concentration within the Big 5 bad? I’m no fan of industry concentration, but the cure may be worse than the disease. With rapid innovation, the chokepoints limiting entry may change.

Fortunately, anti-money laundering regimes provide a model for how to identify tacit collusion. For details, read the paper (no paywall) in CLA’s Competition Journal: .

Courts will have to develop doctrines that identify the evidence needed to prove that a pricing algorithm is being used for tacit collusion.

The line between efficiency and collusion can be tough to draw. Government shouldn’t tell firms, “Given your industry structure, you can’t ever use pricing algorithms.” But firms shouldn't be able to say, “the algorithm made me do it” as an excuse for price fixing.

In pharma, competition is hampered by highly concentrated intermediaries (PBMs) and an industry in which the consumer is not the buyer. Concerns are heightened by the fact that the intermediaries can pool vast amounts of non-public information.

Algorithmic collusion concerns are heightened when a market is structured in a way that already hampers competition, such as the U.S. pharmaceutical market.

Anticompetitive “Hub-and-spoke” agreements allow competitors to collude without directly interacting. AI algorithms can facilitate this kind of collusion in subtle and hard-to-spot ways.

Today’s conspiracies could look different from drawings of smoky rooms full of well-fed men in suits chomping on cigars and plotting to keep prices high.

Finally, avoiding disasters generally requires enormous cooperation between the public and private sectors. In the end, there is much we can learn from Y2k about distant dangers.

Second, envisioning a disaster doesn’t necessarily mean it will happen. To the extent there are concerns about future AI disasters, we can face them and determine how to avoid the disaster.

First, today is the time to think about where technology will lead tomorrow. As we are seeing from the downsides of social media, tomorrow is a little late.

Today, we are again faced with a transformative technical advancement. And we can find much wisdom in the lessons of rounding the millennial corner.

But after an extraordinary scramble of reprogramming, involving partnerships between the public and private sectors, the dawn of the millennium passed with barely a hiccup.

People thought that elevators would stop running. Flights would be grounded. Data would disappear. ATM machines wouldn’t work, potentially leading to a run on banks. Security and communications would fail, and general mayhem would ensue.

Unfortunately, programmers didn’t focus on what would happen when the 1900s became the 2000s. As the year 2000 loomed, experts predicted that when the clock struck 12:00 AM on New Year’s Day, computers would go haywire—befuddled by a date that literally did not compute.