The rapid advancement of artificial intelligence—the discovery of mind-bending use cases and launch of new businesses—has warped my sense of time. Suddenly, leaps in computing that previously took years seem to be unfolding in weeks. When I first wrote about ChatGPT, back in December, I described my excitement about how A.I. could potentially transform art and industry, but also my fears about mistakes and unintended consequences along the way. It was this same concern, after all, that led Sam Altman and Elon Musk, among others, to found OpenAI, the company behind ChatGPT and Dall-E, with a mandate to develop these tools in the safest possible way.
Since then, of course, OpenAI created a for-profit subsidiary (which operates in a decidedly less-than-open fashion), partnered with Microsoft, and began churning out new developments at a breakneck pace. The A.I. arms race is on. In just the last week, Microsoft announced the inclusion of large language models across its full Office suite in the form of an integrated A.I. system called Copilot. You can think of it like the old “Clippy” helper, but with total information awareness of your data in ways that would make an intelligence agency drool. I think the promotional video is important to watch. Just ignore the appeals about reconnecting people to “the soul of work” and focus on the new capabilities and new dependencies being introduced. The job I had for years as a strategy consultant won’t exist in that form again, and neither will many others as presentations build themselves, spreadsheets analyze themselves, and emails compose themselves.
Despite C.E.O. Sundar Pichai declaring Google an “A.I. first” company all the way back in 2016, and despite the fact that the “T” in ChatGPT refers to transformer technology pioneered at Google, the company has seemed to lag these latest developments. It also experienced controversies that may have slowed its public deployment just as competitors were ramping up, from the failed Duplex service that would robocall businesses to book appointments to the dismissal or resignation of A.I. and ethics researchers including Timnit Gebru. But now it looks like the days of slow rolling are over. Last week Google announced the integration of large language models and generative A.I. into its Workspace productivity suite. What felt like a steady, largely internal, consideration of A.I. is now being driven by the competitive pressures of the market.
Perhaps most importantly, last week OpenAI also unveiled GPT-4, an upgrade to the model powering ChatGPT, and announced new partnerships and capabilities. Here are a few eye-popping examples: DuoLingo, the language learning app, added conversational roleplay so you can practice talking in your new language; Be My Eyes, an app that connects sighted volunteers with blind or low vision users added an A.I.-powered virtual volunteer option using the new image-processing capabilities of GPT-4; and GPT-4 can now pass a number of tests (AP exams, the GRE, even the LSAT and the bar) without having seen them before. Oh, and Midjourney, the image synthesizer that generates much of the A.I. art clogging your feeds, launched version 5 last week as well.
It’s not all generative A.I. rainbows and puppies, though. I have a feeling I’m going to be writing and worrying about A.I. for the foreseeable future. One of the foremost challenges, to my mind, is the potential further erosion of individual data control, consent, and agency in an era of A.I. supremacy. The fluency of a predictive email system increases the more it scans your inbox. The same is true for our photo apps, indeed all of our experiences. Each of these A.I. tools become more “useful” to us as they accumulate larger data sets, either by digesting the information we feed them or by essentially consuming the internet, itself. As with facial recognition apps or biometric scanning, we must submit to further digital surveillance to get the most value from what these tools can offer. The flipside of this is that any part of our lives not captured by the system can’t be enhanced, and so the promise of an A.I.-powered future is limited to the types of knowledge, experiences, and sentiments that can be digitally captured and rendered. It should go without saying that, for now at least, there’s much more to human history and experience than can be fed into a computational system.
All these tools will further shift the nature of work, and indeed human existence. Our measures of skill, competence, social belonging, even purpose will derive less from an embodied experience and more from our ability to manipulate technology. A small example is the rise of calculators. I was there when the TI-82 graphing calculator rocked my high school math class and changed how teachers teach and test. Last week I met a kid who’s going to UC Berkeley, and he proudly shared that ChatGPT drafted most of his admissions essays. The underlying nature of these technological changes isn’t new, but the speed of change, scale of impact, and range of applications is. The way we achieve our goals, accomplish our tasks, and realize our will is going to increasingly depend on our ability to properly command a computational system. Essentially we are becoming magicians or wizards, and the quality of our magic will depend on the quality of our imaginations and the quality of the spells we cast. So it bears asking repeatedly, who has the resources to learn spellcasting, and what are the limits and biases of the underlying magic?
I also worry, of course, that government regulators and law enforcement won’t be able to keep up. Impostor scams were the most common fraud reported to the Federal Trade Commission last year, as bad actors seized on the ability of A.I. models to simulate a person’s voice, using only a few seconds of sound. We still don’t have a handle on SWATing, the 2010s-era phenomenon in which people use untraceable internet calls to phone in fake threats to 911, resulting in real people experiencing potentially deadly police raids. Now we’re rushing into an era in which A.I. tools can replicate photos, voice and video with near-perfect accuracy, or where a chatbot accidentally or maliciously invents ostensibly peer-reviewed scientific papers, or government “information.” All of this is suddenly moving at a speed which risks further destabilizing the social order. Our country is already under stress from growing inequality, distrust, partisanship, and rising authoritarianism—and that’s just in the age of social media and smartphones. I fear we’re going to need something else to hold onto for stability in a society experiencing multiple technological revolutions within a single generation.
The Real Woke Mind Virus
Speaking of instability, I’ve been thinking a lot this week about Silicon Valley Bank, the go-to banker for the innovator set, which made some shockingly bad bets on low interest rates at the well-publicized end of the free money era, and about the venture capitalists who pride themselves on efficiency and rationality who freaked out and instigated a social media-driven panic. Those who rant against government oversight and regulation begged, nay, insisted on a bailout to save themselves, their companies, and maybe inadvertently, all of us from a larger bank run. After all, the vulnerability in our “fractional banking” system is that our money isn’t actually held anywhere, but rather is continually lent out and leveraged to generate profits for bank shareholders. (For the best takes and analysis of all this, read my colleague Bill Cohan). The financial system, which we are encouraged to believe is a scientific and mathematical endeavor powered by reason, is ultimately faith-based. We believe banks are sound. We believe currency has value. We believe these things and so they are “true,” not the other way around. And when we stop believing, they stop being true.
What if we believed in something else? I find myself reflecting on a conversation I had recently with the economist Kate Raworth. She’s the pioneer of “doughnut economics” which envisions a new design for our economic system. In this new model, we don’t pursue endless growth. Endless growth in biological systems is a malignant cancer that destroys the host, and endless growth in economic systems has pretty much done the same to our host planet and our ability to live here.
Instead of limitless growth, Raworth, like many indigenous communities have practiced for millennia, advocates for economic activity within bounds: a lower bound below which we don’t let anyone fall, and an outer bound beyond which no one exceeds. The shape of this hypothetical economy is not represented by a line forever going up and to the right. It is circular and regenerative. It’s a donut! I explored this concept with Kate on my How To Citizen podcast, and wish we could seriously consider this model as we face yet another opportunity to question our financial system and the assumptions holding it up. To wit: What if banks weren’t incentivized to take risks with our deposits? Sure, traditional banking would be a lower margin business. And perhaps bankers would return to being compensated more like functionaries than capitalists. The financial world could be a slower, safer place.
But America’s champions of deregulation don’t want to look inward for answers when there’s progressive scapegoats about. One of the more outlandish explanations for SVB’s failure was that it came about due to the bank’s preoccupation with “wokeness,” whatever that means in this context. The bank with an all-white executive team and single Black board member was not undone by a focus on diversity. It was undone by inexcusable miscalculations that overlooked the basics of banking. Not surprisingly, it turns out that the management team was terrible at multiple things.
Of course, these facts haven’t stopped Ron DeSantis and Wall Street Journal columnists from promoting this quack theory. They love to hate letters like D.E.I., E.S.G., and C.R.T. Maybe they love to hate all letters because letters can become words, and words can spread ideas, and ideas don’t always agree with them. This must be a confusing time for Republican thought leaders. Historically, they side with business in all things, at the expense of workers, the planet, and even health. Let businesses do business has been their mantra since Reagan’s day. Yet if a business tries to remain relevant by responding to demographic and cultural change, or by trying to preserve its market opportunities by preserving the natural environment that makes all market opportunities possible, these leaders cry foul. The people who welcome unlimited corporate campaign contributions as “speech” whine that businesses are getting too involved in politics and are reacting by limiting their speech.
I’ve been speaking to and informally advising businesses on diversity and inclusion for years, but I was curious to get a deeper understanding of how companies are responding to this G.O.P. backlash. In speaking with my friend Julie Ann Crommett, a 14-year D.E.I. leader with a focus on storytelling and the media industry, I asked what she was seeing among her clients. She described three types of responses. According to her, most are continuing the work, just more quietly. Some are putting their heads in the sand, hoping the backlash is temporary (or “transitory” like the inflation the Fed hoped to wish away). And then there’s a third category who are laying off D.E.I. teams but hiring outside firms to continue the work, sometimes by the same people they fired. I kind of love this last group. They are ultimately paying their workers of color more as consultants than as full time employees. Win-win?
I don’t think the D.E.I. backlash and anti-woke whining emanating from Republican state leaders is about free speech. Just like I don’t think the extreme abortion bans are about protecting life. Both are about control. In the case of anti-D.E.I. rhetoric, it’s about cultural control. We never told the truth in this country about its founding and who did the building. Only in the past few decades have we begun to approach telling the truth, and that tiny taste is proving too much for some. Either they truly fear their position being threatened by that truth, or they see profit in peddling fear to others, or both. Regardless, they know our memories and attention spans are short, and if they can miseducate another generation or two, they can shift the narrative for the long term and maintain that control.
I oscillate between finding all of this amusing and terrifying. In their fervor to invent victim narratives for themselves, the people who think being “woke” is a problem presumably prefer teachers, businesses, and all of us to be asleep instead. It’s certainly more convenient for Republicans to frame the collapse of SVB as the cost of promoting social or racial equality, rather than for Americans to awaken to the reality that the bank’s former C.E.O. actively lobbied Congress to evade enhanced capital requirements under Dodd-Frank. Did nobody inside the SVB boardroom think to raise questions about taking on so much interest-rate risk, with so little reward? Perhaps Silicon Valley, like Wall Street, could use more diversity of thought after all.