White House shares AI Action Plan

Quick Note

Everyone is talking about the Trump administration’s new AI Action Plan, so that’s what we’ll focus on today. Per usual, I won’t get into what’s right or wrong, but will address the differences between the White House’s current approach to AI and the previous administration’s.

White House shares AI Action Plan

The Trump administration has unveiled an AI Action Plan that aims to maintain U.S. dominance in artificial intelligence while repealing policies put in place by former President Biden.

What’s in it:

• Plans to remove federal and state regulations that the Trump administration sees as barriers to AI development.

• Support for the accelerated construction of large-scale data centers that provide infrastructure and computational power needed to run AI. 

• Revised export controls that will allow the shipment of AI tech to U.S. allies while limiting access for rivals like China. 

• Criticism of “ideological bias” in chatbots and support for oversight that will prevent AI systems from enforcing “socially engineered agendas” (this is basically the AI equivalent of getting rid of DEI).

The bigger picture: Trump has now delivered on a campaign promise to deregulate the AI sector. This will reduce government oversight of AI development and give big tech companies control over safety. 

How are these policies different from the previous administration’s?

Politics can be a touchy subject — so I won’t be sharing an opinion here. But when we’re talking about technology that many people expect to shape the future, it’s important to understand how policy changes could play into its development. 

The Biden administration:

• Imposed regulations and oversight that established the federal government as a watchdog over AI development. 

• Encouraged tech companies to share their progress and safety reporting with the government prior to launching new systems publicly. 

• Offered to lease federal lands to private companies looking to build AI data centers.

• Passed export controls to restrict foreign access to American tech (as mentioned, these didn’t prove to be very effective).

The Trump administration has:

• Aggressively deregulated to minimize government oversight of AI development, entrusting safety almost entirely to private companies. 

• Made it easier for tech companies to secure permits to build data centers on private land. 

• Promoted exports of American tech to allied countries and promised to close loopholes allowing it to reach adversaries. 

In a nutshell: The Biden administration took a cautious approach to AI that balanced progress with safety concerns. Trump’s White House is taking a more aggressive approach that prioritizes speed over safety. 

Closing notes

I like to say that tech companies are the writers of their own press releases. The same is true for the White House, and there are a couple things that the Trump administration objectively got wrong. 

1. AI models do not have biases: AI doesn’t think or believe anything. Large language models are trained to predict what text comes next in a sequence, not to express a predetermined set of beliefs (we know this is true because Elon Musk’s Grok chatbot tends to share progressive opinions despite his efforts to make it “anti-woke”). It is true that AI models can reflect biases from their training data. For example, if 90% of news articles about the Russia-Ukraine war portray Russia as the aggressor, an AI model trained on those articles will likely frame the conflict the same way. What some people see as signs of AI biases are really just reflections of human sentiments.

2. Fully closing export control loopholes is unrealistic: Although not impossible, Trump’s promise to completely close these loopholes might be overly ambitious. Even if the White House blocks the export of AI chips to Chinese trade partners, smugglers and black markets will continue to provide high-end chips to countries that want them. The unfortunate reality is that once chips leave the country, tracking their final destination or use is nearly impossible. 

Thanks for reading! How was today’s newsletter? I know AI regulation isn’t always the most exciting thing to talk about, but I hope you learned something today. If you have any questions or need clarification on anything above, reply to this email and I’m happy to help.

See you next week,

Chain of Thought