4 charts that show how fast AI is learning. These charts from Time help explain the overlapping reasons why AI is improving so radically: New models are being trained with fast-improving algorithms and with exponentially increasing computing power and data. As a result, computers are leapfrogging humans in more tasks every year—and they’re progressing from “novice” to “expert” level very quickly.

HBR: How generative AI changes productivity. From Harvard Business Review, here’s a podcast (with transcript) exploring how AI is transforming creativity and innovation. One theme: As AI tackles more repetitive and lower-level work, humans will be able to reorient around the tasks they do best. As Karim Lakhani says, “If my accountants have the superpower [of AI assistance], what more could they be doing? What additional, higher-level stuff could they be doing?”

• Related: Ethan Mollick, a professor at Wharton, describes how he uses AI to “[do] the boring stuff fast.”

The right (and wrong) way to use AI in brainstorming. Who’s better at generating ideas: humans or AIs? For this working paper, the authors asked students in a product design course to face off against ChatGPT. The result: ChatGPT generated ideas 40x faster than humans, and the AI’s ideas rated better, on average, than the humans’—but the AI’s outputs were also more variable in quality. That is, ChatGPT’s ideas were more likely to be really good or really bad.                                   
                        

• Implications: You know the cliché about how there are no bad ideas in brainstorming? When you’re just throwing out ideas, it doesn’t matter how bad your worst idea is; you just need your best ideas to be as good as possible.
  
  The “right” way to use AI to brainstorm, then, is to ask it for lots of ideas—ideally by prompting it with good examples—and then to choose its best outputs. (The “wrong” way would be to ask AI for just one idea or, worse still, to avoid AI simply because some of its outputs are terrible.)