When personal computers went mainstream in the 1980’s, a euphoria around what they could do was met by contrarians pointing out their limitations. Sure, they could do analytical tasks better than humans, but not intuitive, strategic tasks, such as playing chess. And then computers were built to become chess champions. Yes, they could reason faster than humans but not coordinate movement tasks. And then computers were built into deft manufacturing robots and human prosthetics. Sure, they could do what humans programmed them to do, but they couldn’t outlearn their programming. And then AI shredded that assumption.

As ChatGPT burst onto the scene, a similar euphoria erupted around what it could do, again followed by contrarians pointing out limitations. Sure, it could learn and regurgitate case law faster than a human, but it also made up cases. Then AI was improved to show sources. Yes, AI can write content better than many people and faster than all, but it’s wreaking havoc with publishing and educational practices. Then AI was developed to detect AI-developed content. And so on. We have decades of experience telling us that the edges we suppose to machine intelligence are but the starting point for the next X-prize. And so it is now as we witness the explosive use of AI in good hands and bad—arguably even its own hands—with command of knowledge and networked effects we cannot even imagine.