It’ll be years before we can say, with certainty, whether generative A.I. is a revolution or a fad.
No one doubts that A.I. (more generally) is revolutionary and indeed it’s already revolutionized our lives in ways most people aren’t even aware of. But as it relates specifically to the generative A.I. applications that took the world by storm late last year and in early 2023, I still see elevated “fad risk,” if you will.
My experience with DeFi last year left me cautious about grandiose claims for new technology. This year’s ChatGPT headlines read quite a bit like headlines touting Web3 in late 2021 and early 2022. Now, as then, every other headline suggests another business model is being disrupted, another company is jumping on the bandwagon, that mainstream adoption is proceeding at lightning speed and that the technology is making inroads at an alarming rate.
Almost doubtlessly, there’s more truth to such headlines as they relate to generative A.I. compared to those which grew up around the Web3 fairy tale, but it’s hard to say with certainty. We tend to forget how enthralled we were with the last craze once it passes, often to avoid coming to terms with our own penchant for irrationality.
Remember: Folks like Matt Levine (allegedly among the smartest, most cynical market observers around) and David Rubenstein, were still countenancing Sam Bankman-Fried just months before FTX’s implosion. Bankman-Fried effectively admitted, to Levine, that he (Bankman-Fried) was in “the Ponzi business,” as Levine famously put it. And who can forget the painstakingly rendered, biographical, novella-length “spotlight” piece Sequoia commissioned from a private historian to extoll Bankman-Fried’s many visionary virtues?
That’s not to suggest generative A.I. is any kind of legal fraud (although there are plenty of rumblings about legal action tied to A.I. output, given that much of it is just a mashup of existing writing and visuals). Rather, the point is that when a fad is in full swing, it’s sometimes hard to identify it as such, even when common sense is screaming at us.
With that in mind, I wanted to briefly highlight three other notables from the evolving A.I. craze that caught my eye this week (“other” because what most grabbed my attention from a “pure markets” side, was Chegg’s rough day, and I already covered that):
- On Monday, in an interview with Bloomberg, IBM indicated it’s in the process of pausing hiring for what could be as many as 7,800 positions seen as replaceable by A.I. Those positions include back-office roles and — in an accidental pun — human resources. “I could easily see 30% of that getting replaced by A.I. and automation over a five-year period,” CEO Arvind Krishna said.
- On Tuesday, Citadel Securities CEO Peng Zhao, speaking at the Milken conference, tried to put a positive spin on A.I., using a familiar line of reasoning. “Being liberated from the repetitive, mental laboring aspect of what we have to do as humans, allows humans to focus on the highest-value portion of planning,” he said. So, when A.I. replaces you at work, you should rejoice because you’ll be free to become a flâneur, only without the money.
- Most notably, the “godfather” of A.I., Geoffrey Hinton, left Google and delivered a stark warning while speaking to The New York Times for what amounted to an exit interview: “It is hard to see how you can prevent the bad actors from using [generative A.I.] for bad things. Maybe what is going on in these systems is actually a lot better than what is going on in the brain.”
I’d encourage readers to remember that humans, even the smartest among us, are predisposed to being swept away by emotion, fear, excitement and hype. That’s part of what it means to be human, and somewhat ironically in this context, these human emotions may be causing us to exaggerate the potential of this burgeoning technology.
I should reiterate: I’m not questioning A.I.’s potential. I’m questioning 2023’s ChatGPT mania. Have a look at the figure below.
That’s a newsflow indicator from SocGen, and as the bank’s Manish Kabra noted, the launch of ChatGPT is solely responsible for the “exponential” increase.
The A.I. discussion isn’t new. We’ve been predicting our own demise at the hands of A.I. we create for decades, both in serious academic literature and in Hollywood screenplays.
What’s new in 2023 is that we all got a look at a highly advanced chat bot. Whether that’s cause for the ensuing panicked delirium can, I think, be legitimately questioned.
But who am I, right? “Look at how it was five years ago and how it is now,” Hinton told the Times, of A.I.’s rapid evolution. “Take the difference and propagate it forwards. That’s scary.”
Read more on A.I. in the latest monthly letter: Reinventing Wheels
Indeed scary. I did laugh at point 2, last line. Thanks!
The part of AI that scares me the most is the cybersecurity aspects. With deep learning systems, it has been shown that, not only can biases be introduced by the ordering of training data but completely undetectable backdoors can be placed into the systems by unscrupulous third-party trainers. Even systems like ChatGPT have security vulnerabilities that researchers are just now starting to map out. Companies that jump onto this band-wagon without weighing these risks could create major liabilities for themselves.
In some ways the nasty unintended consequences of AI are a comfort to me, at least. These bad outcomes that may result in the loss of customers, unforeseen liabilities and lawsuits should soon be able to naturally regulate the spread of this opaque disruptor. My hope is that in its annual rush for political donations, Congress will not make the same mistake it made with the internet by being too anxious to cash in without putting any limits on its expansion and the unwarranted protection of its users at our expense.