Will AIs Have Free Speech?

What if it's the world's most persuasive AI?

Hi Everyone, and welcome to my new home for writing. I’ve written some posts here over the last year, but as I phase out of InsideAI, this will be my primary outlet. The focus of this substack is Investing In AI but, I use the term “investing” broadly. While the focus will be on financial investing in AI companies, it’s also important to cover investing time and effort and attention and other resources, and so there will be posts covering those things as well.

Watching the Twitter drama, in particular the debate over whether or not Trump should be allowed on Twitter, raises a lot of interesting questions about free speech. Today I want to ask the question - should an AI have free speech? What if it’s dangerous? What if it’s the world’s most persuasive AI, able to personalize arguments at scale to be convincing to many types of people, maybe even most people.

In 2014 I remember visiting one of the customers of my startup at the time. We sold into I.T. departments and this was the head of I.T. at a major growth stage silicon valley company. He had spent his life in political technology and so I asked him why he moved on.

He told me he had worked on many campaigns, but his last one was the Obama 2012 campaign and, at this point the technology had gotten so good that (I’m paraphrasing here from memory) “it could tell you how to change your message street by street, house by house, to tell people exactly the thing that would resonate most with them.” He quit because he felt politics shouldn’t be that way.

Someday someone will compile a good data set of arguments, and use it to build a chatbot that tries to convince people of various points of view. That will lead to a bigger data set and that will lead eventually to a very convincing chatbot on a host of topics. Let’s say one political party national committee, or one consumer company, or one entity that has an agenda other than the broad welfare of humanity owns this bot. Would you allow it to have “free speech” and be open on the web, on Twitter, or wherever else it wants to be?

We often think of free speech as something for humans in democracies. What about bots as they get closer to human status? What was the original purpose of free speech, and does a new AI use case still fit? These are complicated questions without easy answers. For now, I just want to stimulate discussion and general thinking about these issues, but hope to address them here in coming months in more detail, and possibly a framework for thinking about them.

Thanks for reading.

Rob