Not long ago, the Salvador Dali Museum announced a fun AI project. They fed all of Dali's written and recorded words, including audio and video from every source they could find, into a big "large language model" program and created a cool device: A lobster-shaped telephone that allows you to have a conversation with AI Dali.
Visitors to the museum can pick up the phone and ask anything they want. The voice of Dali then answers their question and proceeds to have an AI-generated conversation with them. The voice was trained on recordings of Dali, so the result both sounds like his voice and the content is drawn from his thoughts.
But some people get the description of this wrong. And as we see more projects like this with famous dead people, we need to be careful about how we describe it.
This is not Salvador Dali. The answers one might get are not the answers Dali would give. They are in his "voice" and use phrases he might use. But the answers are not what Dali would say if he were here today. How do we know that? Read on.
You may have seen that the voice of Al Michaels will be used for the AI bot that will be commenting on the Summer Olympics. The NBC network has been very careful to walk the line. They try to be super clear that this is the voice of Al Michaels, but the words of generative AI. Going forward, as artificial intelligence and video/audio generation get better, we will see this kind of thing everywhere.
Note: You should never believe that having a conversation with the Dali phone is just like having a conversation with Salvador Dali. And watching "AI Al" Michaels will not be the same as having the real Al Michaels in the broadcast booth.
The big difference is the human spark of creativity. An AI program might sound exactly like Salvador Dali, but it will always lack the one quality that leads one to talk to Dali: His genius. Dali became famous by using his artwork and commentary to show the world a different way of looking at itself. His artwork was his form of social commentary.
Dali passed away in 1989. Imagine all the commentary he might have about the Internet, smart phones, streaming media, digital artworks, pandemics, politics, . . . and generative AI creating artwork inspired by his creations. The best guess you can make about his commentary should be, "I have no idea what he'd say." He might love or hate where society has gone. He might approve or disapprove. But his actual response will be new, original, and unpredictable.
The one area where AI excels is being predictable. So, you might be able to work hard enough to get an AI model to figure out how Dali might have reacted in the 1970's or 1950's. But it could never create the unique, inciteful, and unpredictable commentary that only the human Salvador Dali could come up with!
Keep this in mind when people are tempted to say, "It's just like talking to the real . . .."
-- -- --
For more information on the project, see this articles:
:-)
No comments:
Post a Comment
Feedback Welcome
Please note, however, that spam will be deleted, as will abusive posts.
Disagreements welcome!