ChatGPT as Garrulous Guerrilla

Photo by Axel Antas-Bergkvist via Unsplash

The tsunami coming for the creative class’s jobs

Will I be replaced by ChatGPT? No. Will some (most?) of my work be replaced by ChatGPT? Yes.

Such is the quandary for writers, directors, illustrators and other “creatives” in the decade ahead. More than 100 million users are estimated to have either toyed or fell in love with ChatGPT since it was released to the public a few months ago, a breathtaking adoption rate. The initial euphoric hype of the launch has become a bit more demure once users reached the bot’s limitations and encountered its flaws. An “AI hangover”, if you will.

When it comes to factual output from bots, ChatGPT — and generative AI models more widely — suffer from two evident flaws. First is a complete lack of structured thinking or analytical rigor in their output. ChatGPT and other AI models don’t actually know anything but are rather taking inputs and producing outputs using a probabilistic model of what seems right. As Josh Wolfe and I discussed with NYU professor and prominent deep learning AI critic Gary Marcus on the “Securities” podcast two weeks ago, generative AI will quite literally help you kill yourself — and not even know it’s trying to do so (see more in Gary’s piece in Wired, “The Dark Risk of Large Language Models”).

Second, generative AI models lack any capability to be rigorously original. This often gets confused with their ability to be stylishly original; these models can indeed invent worlds and images and paintings the world has never seen before or provide tools for creatives to perform their craft. For instance, Lux-backed RunwayML hosted its AI Film Festival this week at the iconic Metrograph theater in New York City, and the capabilities that the platform can offer to directors and editors to imagine new worlds are plentiful and empowering. All new forms of film are possible as viewers will see in the years ahead.

But that’s about style and creative productivity, and again, not about rigorous originality. AI models can’t develop a logical argument on a topic they haven’t seen before, for the simple reason that they have no actual cognition — probabilistic language models are a far cry from an intelligence that can acquire, process and synthesize knowledge into logical relations.

To detour a bit, it’s been about a decade since I wrote my undergraduate thesis on the history and development of computer science at Stanford, and so I spent time re-reading it and also publishing it to the web last weekend. I spent a year reading archival documents from a dozen or so collections of papers to weave the story together, gaining access to early tenure cases in the field and also classified defense budget information. Rigorous originality here comes in two flavors. One is that the evidence itself is original, in that no human before me has ever even glimpsed the documents and materials in question (besides of course the authors and contemporaneous readers). Second is the analysis itself, a novel argument built upon original facts and proof.

If you had asked ChatGPT about the development of the computer science field at Stanford in the late 1950s and early 1960s, it would have no material in its corpus to build upon, and so it would output only blandishments. The irony of course is that now that I’ve published my thesis on the web, OpenAI and ChatGPT will one day hoover the damn thing up and regurgitate my sentences on future prompts.

There is still an ample amount of original work conducted today. On the analysis side, there’s archival-based historical work, laboratory research, clinical evaluations, enterprising investigative journalism, econometric modeling and more. On the creative side, there’s all forms of art, from novels and sculpture to drawings and film, where a work of art embodies an artist’s perspective, sensibility and taste while entering a dialogue with other artifacts.

The quandary for the creative class is this: how much of the work that the millions of global “creatives” do is original? Here’s an uncomfortable and elitist point of view: very, very little.

Most journalism today is a rehash of press releases, press conferences, and PR-mediated information. Most marketing copy (think the millions of small businesses advertising across media) is a simplistic remix of prior work. Nearly all editorial columns and DC-style think tank policy memos have conclusions that are obvious before they are even written (and have about as much personality as a blank chalkboard). How much of our social media conversation is “insightful” (hint: basically none, especially on that hellscape known as LinkedIn).

That’s the analytical side. On the creative side it gets more ambiguous and open to interpretation, but ask directly: how many artists and auteurs ultimately develop their own taste and a sense of their unique contribution to the creative output of humanity? How many artists in all different media and forms “converse” with their field? Originality is a tough bar to meet.

What’s uncomfortable is that the creative class was supposed to be the last bastion for employment, a zone of safety from the aggressive automation of the modern economy. Factory work can be automated with robotics, but bots could never replace the thinking and originality of the human mind. Go to college rather than trade school, it’s mind over muscles.

What was patently obvious though is that the economy never needed tens of millions of writers, artists and other creatives. There’s never been an audience for all of these people and all of their work, which is why the remuneration and job security for the professionals in these fields has been plunging over the past few decades. While the internet is often blamed for disrupting the economic models of creative and intellectual work, few seem to point out that incomes have plummeted just as the supply of creatives dramatically expanded. That’s simple supply and demand.

There is plenty of room for iconoclasts and trailblazers to write bestselling novels and sold-out plays, to paint brilliant works and direct heart-wrenching films. But 90% — and maybe even 98% — of the creative class’s output can be replaced today with generative AI (and just wait until the field continues to evolve and perfect its quality over the coming years).

As a long-time professional writer, my personal tracker tells me that I have written 1,411 articles, essays, papers and other media for publication (this newsletter is #1,412). Scanning through those pieces, there definitely exists work that is unique, original and non-displaceable by generative AI. Is it a majority of my oeuvre? No. If I had to guess, I’d say somewhere between 1,200 and 1,300 of those 1,411 pieces could be written by AI today. Those articles would probably lose something of my … joie de vivre (and dad jokes) in the process, but the world never really asked for my style in the first place (nor, in general, does it want to pay for it). Worse, AI is “free” or at least far cheaper than a human, and that means that the market demand for substitution is strong.

There are two displacements coming up then for creatives. The first is that many, many professionals will have to brutally come to terms with the fact that their work is unoriginal. Some will migrate to more sophisticated and original work and float above the high tide — but the vast majority won’t, or can’t. Entire categories of creative professionals today will be eliminated.

But the bigger challenge societally is that we are going to have a collapsing number of pathways to creative excellence. Nearly all creative class jobs are designed around apprenticeship paths, even if that specific word is not often used. Actors start with small and practically invisible gigs, journalists typically begin with basic beat coverage before hitting tougher and more enterprising assignments and directors launch with short films before developing their skill to handle feature-length works. Pay is almost always a pittance before reputations are built.

What happens in a world where all of those simpler, AI-replaceable assignments that used to offer experience and skill development to creatives are suddenly expunged? How do you develop as a creative? How do you get compensated for the years required to develop a craft?

Michael Lewis and Elizabeth Kolbert aren’t going to be displaced as writers by AI. But how will the next generation develop the skills that took the two of them decades to acquire and perfect? So much AI commentary speaks about whether jobs will exist or not, but doesn’t consider the evolutionary dynamics of how careers actually upgrade and adapt over the years. New entrants are going to have a tough time making their way financially in this new world.

Generative AI, even in the haze of a bit of an AI hangover the last few weeks, is akin to a tsunami on the creative class: those on high ground are going to survive and stay dry, but the vast majority are much closer to the water and much more at risk of being forcibly dragged out to sea. What will millions of creative 20- and 30-somethings do when their work is casualized and demoted? That’s going to be one of the most … interesting political epochs to watch for in recent memory.

Just Released: Lux LP Letter 2022Q4 (“From Strife to Strive”)

We’ve got a popular drop for this weekend: Lux’s latest quarterly letter from Josh Wolfe made its public debut this past week. From the intro:

Fundamentals have been replaced with fundamentalists, profitless prophets declaring boom or doom amid an information economy where the truth is liminal and the lies limitless. Bubbles are popping and balloons are dropping from the North American heavens, all while tensions seethe across markets, managers and the multitudes. Some analysts have gone so far as to call our present moment a “polycrisis”, of intersecting and intensifying chaos that could destabilize our very foundations. It is Lux’s contention, however, that we aren’t in a polycrisis but rather have a mono-opportunity: a singularly unique moment to transmute today’s conflicts into an opportunity to advance the interestsfor our future, the future of our Limited Partners, and the future of all humanity. And indeed, contention itself has precisely that two-faced meaning: a heated disagreement (“strife”) and an assertion (“strive”). The word “contention” itself is derived from Latin’s contendere meaning to strive, delivering us to the essence of today’s moment — the need to turn conflict into intention, to turn chaos into order, to turn strife into striving.

Our summary view of the present market moment:

Contrary to the current consensus of a soft landing for the economy, we expect market conditions to be very volatile with wild chaotic swings: widespread surprises; large layoffs (what started with white-collar workers will spread to blue); significant LP pullback from being overcalled and overallocated; and GPs being under-reserved and under duress. All of these events will trigger crushing consolidation in the venture industry, and startup founders and fund managers will realize that the estimated $300 billion of “dry powder” will in fact prove quite “wet” due to not-yet-realized and largely spoken-for capital. We expect a flood of recapitalizations and drastic down-rounds, with managers forced to play their hands protectively against punishing pay-to-plays and other highly structured terms.

And the future of the economy:

There are plenty of famous investors and thinkers to listen to for where the world is headed and where one might wisely allocate capital. The teacher better than Buffett or Howard Marks however may well be Xi Jinping. Investor wisdom of “Don’t fight the Fed” may give way to “Heed what Xi hints” […] The new areas sanctioned and subsidized by Xi are the hard sciences and deep tech: biotech, aerospace, defense, space, satellites, and semiconductors—the latter of which has been the subject of a multi-season, Netflix-style show in China that celebrates a hero who returns from the United States to build a national asset in China’s semiconductor industry.

Read the full letter (PDF) on Google Drive or read Josh’s summary thread with screenshots on Twitter.

Lux Recommends

  • I have brought up the growing desire of the South Korean public to acquire nuclear weapons, a movement which has even recently gotten the backing of the country’s president. Kyung-joo Jeon has a reasonable perspective for the Korea Economic Institute of America on the constraints and challenges of building out a nuclear weapons program at this late stage and what it would mean for U.S. extended deterrence.
  • Shaq Vayda recommends an article from a few years back by Jacob C. Kimmel at NewLimit who writes on “Learning representations of life”. "Framed in the proper historical context, the ongoing convergence of computational and life sciences is a reprise of biology’s foundational epistemic tools, rather than the fall-from-grace too often proclaimed within our discipline.”
  • One of the classics of early-2010s internet essays is Tom Scocca’s “On Smarm”, which popularized the term. "Smarm is a kind of performance—an assumption of the forms of seriousness, of virtue, of constructiveness, without the substance. Smarm is concerned with appropriateness and with tone. Smarm disapproves.” Now, Stephen Marche (the author of The Next Civil War, which I reviewed all the way back in Issue 1 of “Securities”) writes a 10-year followup to Scocca’s essay, exploring that moment in time and what the legacy of smarm has meant for the world at large. "The current state of public discourse, if it’s even worthy of that name, is a strange fusion where smarm and snark wrestle and embrace one another in vicious shadowy vacuums. It is less clear than ever which side is winning.”
  • Our scientist-in-residence Sam Arbesman recommends James Somers’s analysis in The New Yorker arguing that it isn’t ChatGPT we should be excited about, but rather OpenAI’s Whisper product. “Nearly perfect speech recognition has become not just an application but a building block for applications. As soon as this happens, things move very fast.”
  • Finally, two articles on Russia’s mercenary overseas army Wagner Group that were interesting, one from the Financial Times on just how little anti-money laundering procedures are apparently followed in London (with Wagner Group’s head proving his St. Petersburg residence by showing a copy of his 83-year-old mother’s utility bill) as well as in Le Monde (in French) on how Wagner is expanding its influence across the Sahel region of Africa. Russia’s goal? Destabilizing Africa will lead to distractions for Europe and therefore less pressure on the Ukraine conflict.

That’s it, folks. Have questions, comments, or ideas? This newsletter is sent from my email, so you can just click reply.