($) AI Executive Order: Likes and Dislikes
Four dislikes and two likes of a laundry list of self-contradicting rules
This is premium member only content. Thank you for being a paid subscriber of Interconnected Premium.
The much-anticipated White House AI executive order (EO) dropped on October 30 – almost exactly 11 months after ChatGPT was first released to the world on November 30, 2022.
The lengthy EO reads like a laundry list of oftentimes self-contradicting rules – typical of executive guidances that have been so heavily lobbied by so many interest groups. From big tech to civil rights organizations, every group wants to have its fingerprint somewhere in the document. And that’s what this document became – an artifact to keep the most number of interest groups happy at the expense of logical coherence, all out of fear of what an 11-month-old “AI toddler” could do to the world.
I’m not oblivious to the potential safety and security concerns that AI could pose and have long posed well before ChatGPT was released. Yet, the technology is so young still that prematurely restricting it would only stunt its growth, limit its benefits, all the while preventing theoretical fears that, as Andrew Ng who started Google Brain and led Baidu’s AI lab argued, amount to nothing more than a hypothetical “it could happen.”
You would never put up so many guardrails around your 11-month-old child. You would want to nurture her, get to know her, let her grow up first. But somehow there is enough consensus to believe we should restrict a 11-month-old AI. Having read through this AI EO in detail, on balance, there is more I don’t like than like. In the spirit of fairness and a balanced perspective, I’ll share my “likes” and “dislikes” of this smorgasbord of rules and guidelines.
Since I’m the type of person who prefers to know the bad news first, let’s start with my four “dislikes”, then discuss my two “likes”. (All screenshots below of specific EO sections come from the original source.)