Imagine asking an all powerful, painfully literal genie to make you a peanut butter and jelly sandwich*. You watch as the genie smears the peanut butter and jelly all over the table and then throws the bread on top.
You then have to direct the genie: “No, Genie, you have to combine the ingredients.” So the genie, says “Okay boss!” then smushes the bread and the peanut butter and the jelly together into a ball in its big strong hand.
So you have to guide the genie again: “No, Genie, the ingredients have to be layered and flat, not mushed in a ball.” And so on, until finally, the genie is able to make the perfect peanut butter and jelly sandwich. This story may make you hungry — and frustrated since lunch needn’t be so difficult! — it also points to opportunities and limitations of computers. They can run tests in simulation infinitely, but theses tests and iterations are only useful when you ask the right questions. Despite these limitations, computers are amazingly powerful when partnered with human’s ability to reframe questions, describe outcomes and imagine alternatives.
From asking if it’s possible to reduce road deaths by creating a car that drives itself, to whether stratospheric balloons can be navigated to deliver internet access, the teams at X have been asking creative questions of the A.I. “genie” for years. We’re now exploring a range of new problems in which this approach could be helpful; from improving food production, to speeding up biological research and testing. I recently joined Derek Thompson for The Atlantic’s “Crazy/Genius” podcast to discuss how this powerful but painfully literal genie can rapidly increase the rate of innovation and some of the ways A.I. paired with human creativity could help us unlock new solutions to really hard problems. You can listen to our discussion here.