Last week, both Microsoft and Google gave demos of their new artificial intelligence–powered search assistants. Microsoft’s Bing Chat sits inside its Bing search engine and Edge web browser, while Google’s Bard chatbot will do its thing on the same page where Google’s standard search results appear. Microsoft seems to have the early lead after Google’s launch presentation—reportedly rushed out in an effort to keep up with Microsoft-funded OpenAI’s ChatGPT tool—included errors produced by the chatbot. That said, users report that Bing’s still invite-only AI search can also produce gibberish responses, referred to as “hallucinations.”

Leaving aside for now the mess of ethical and legal questions the current crop of AI platforms raises, the most common criticism is that they just aren’t very good. Generative audiovisual tools like Dall-E 2, Midjourney, Runway or Riffusion create media that’s full of nonhuman tells. Text-based AI and chatbots spit out paragraphs that either aren’t particularly interesting or, worse, are full of errors. It all makes the idea that generative AI could replace human creativity still feel pretty laughable.

And yet—we have a tendency to judge the utility of a new technology on how well it deals with the problems of older ones. For instance, in 2007, The New York Times covered the iPhone’s launch by rhapsodizing about how easy it was to make phone calls and organize your voicemail: “One button press reveals your waiting messages, listed like e-mail. There’s no dialing in, no password.”