Defending Reality From AI

Criticisms of AI often focus on concerns like the intellectual theft behind AI models, the environmental impact of AI data centers, and the soullessness of AI-generated content. But Ted Gioia is concerned about something more existential: the threat AI poses to our shared sense of reality.

It is now possible to alter reality and every kind of historical record — and perhaps irrevocably. The technology for creating fake audio, video, and text has improved enormously in just the last few months. We will soon reach — or may have already reached — a tipping point where it’s impossible to tell the difference between truth and deception.

[…]

At the current rate of technological advance, all reliable ways of validating truth will soon be gone. My best guess is that we have another 12 months to enjoy some degree of confidence in our shared sense of reality.

But what happens when it’s gone?

Gioia’s article might seem a bit alarmist, even doomsday-ish. We only have 12 months before reality comes crumbling down? Really?!

But then you read about AI-induced psychosis and deceptive chatbots. Or see just how easy it is to create realistic videos with AI and how easily people fall for AI-generated content — or use it to promote lies and falsehoods.

Suddenly, Gioia’s concerns seem less like a Black Mirror episode, and more disturbingly plausible and even apocalyptic. His conclusion?

We have always lived in a world of disputes, but never on this new level of total skepticism. Consider a football game: I think the ref made a bad call, and you disagree — but at least we both believe that a game is actually happening.

Not anymore.

We once disagreed on how we interpreted events. Now we can’t even agree on the existence of events.

Near the end of his article, Gioia lists some potential solutions. But one in particular caught my eye: “reality notaries,” i.e., “custodians of reality” whose job is “validating the actuality of events and media.” That sounds pretty high-falutin’, like some black-robed gatekeeper or mountain-dwelling sage whose knowledge one seeks only with great trepidation.

Of course, in our cynical age, how would we even go about determining who is sufficiently objective and knowledgeable for such an august role? But Gioia’s broader point is that maintaining and defending a communal understanding of truth and reality requires us to actually be in community with others.

Traditional institutions and organizations like churches, nonprofits, neighborhood associations, sports leagues, and book clubs can certainly help build community. Physical media like books, CDs, and films allow us to commune with ideas presented by other humans from throughout history. (Such media is, as Gioia puts it, “immune to the schemes of bots and technocrats.”) And though I know I’m often on a pro-blog soapbox, I do believe personal blogs and websites can also play a role here.

Our ultimate goal should be interacting directly with real human ideas put forth by real human beings in real human contexts, and not settling for some approximation of human thought as presented through an AI-mediated experience or hallucinated by a chatbot. However, that’s far messier and time-consuming than submitting prompts to ChatGPT, and possesses its own problems with deception, misinformation, etc. It also requires us to be humble, quick to hear and slow to speak, open-minded, and curious — qualities that modern media and technology often encourage us to suppress and ignore.

We’ve obviously struggled to agree on basic facts — historical, scientific, or otherwise — even before AI emerged onto the scene. And certainly, there have always been those who embrace and promote falsehoods for power and profit. But there’s no denying that AI amplifies the problem to an exponential degree, and absent any significant regulation or oversight, nothing will change that simply because there’s too much money at stake.

Gioia concludes his piece with this thought: “In the meantime, the rest of us will do what we can. And plan ways to build back the shared faith [that] the technocracy has destroyed. The sooner we start on this, the better.” I’d never be so presumptuous as to call myself a “custodian of reality” — though it does have a nice ring to it — but I will promise you that everything you’ll ever read here on Opus is 100% human-generated, as per my AI policy.

You might not agree with everything you read on Opus — in fact, I hope you don’t because where’s the fun in that? — but you can rest assured that everything you do read comes from a limited, flawed, and 100% human perspective, and not some machine’s idea of that.