The Death of Proof

I noted three weeks ago that I am not terribly concerned about the power of artificially intelligent platforms to create new and interesting stories, artwork, music, and other … products. Or not yet. And I don’t think they will soon get human-scale intelligence, which involves understanding, reasoning, direction, intention, and agency. But that does not mean I am not concerned.

Right now, these mindless machines can create—at lightning speed and on command—any set of words, any graphic image, and/or any sound or piece of music, all from digitized samples. And while I don’t fear what the machines themselves will want to do, I am concerned about what they will be able to do in the hands of humans who do have intention and agency.

In our documented world, proof of anything beyond our own fallible human memory is a form of information: what somebody wrote, what they said in proximity to a microphone, what they were seen and photographed doing. And increasingly, that information is in digital form (bits and bytes in various file formats) rather than analog recordings (printed words on paper, grooves on discs or magnetic pulses tape, flecks of silver nitrate in film stock). If my Facebook friends can publish an antique photograph of farmhands standing around a horse that’s twenty feet high, or a family shepherded by a gigantic figure with the head of a goat and huge dangling hands, all in grainy black-and-white images as if from a century and more ago, then what picture would you be inclined to disbelieve? How about a note with a perfect handwriting match to a person who is making an actionable threat of violence? How about a picture with perfect shading and perspective showing a Supreme Court justice engaged in a sexual act with a six-year-old?

Aside from written text and recorded words and images, the only other proofs we have of personal identity are the parameters of someone’s facial features as fed to recognition software (easily manipulated), the whorls of their fingerprints and x-rays and impressions of their teeth (easily recreated), and the coding of their DNA, either in the sixteen-or-so short segments reported to the FBI’s Combined DNA Index System (CODIS) database or in fragments recreated from a person’s whole genome. Any of these digitized proofs can now be convincingly created and, with the right—or wrong—intention and agency, inserted into the appropriate reference databases. We’ve all seen that movie. And artificial intelligence, if it’s turned to firewall hacking and penetration, can speed up the process of insertion.

My mother used to say, “Believe only half of what you see and nothing of what you hear.” With the power of artificially intelligent platforms, make that “nothing and nothing.”

In the wrong hands—and boy, these days, do we have a bunch of hands pushing their own agendas—the speed and power of computers to make fakes that will subvert our recording and retrieval systems and fool human experts launches the death of proof. If you didn’t see it happen right in front of you or hear it spoken in your presence, you can’t be sure it happened. Or rather, you can’t be sure it didn’t happen. And if you testify and challenge the digital proofs, who’s going to believe your fallible human memory anyway?

That way lies the end of civil society and the rule of law. That way lies “living proof” of whatever someone who doesn’t like or trust you wants to present as “truth.” That way lies madness.