When will AI disrupt everything?

The disruption will come but it won't be as quick as many think

Having been disrupted by AI, white collar knowledge workers wait on the breadline.

Everyone says that AI is going to disrupt everything. If you listen to the venture capitalists, who have never seen a technology that won’t change the world, the AI disruption is imminent. Prepare, we are told, for the breadline: you have been obviated and obsolesced. All jobs will now be done by infinitely scalable, autonomously intelligent agents, and there is no more need for human labor. Shortly after the release of ChatGPT in November 2022, I saw venture capitalists on Twitter breathlessly exclaim that no company would hire any more people within a year. While I’ll grant that this kind of extreme reaction to a genuinely revolutionary and powerful technology was an outlier, its less extreme cousin, which is that an AI-led disruption is imminent, was, and remains, pervasive.

Here’s the thing—I think that an AI-led disruption is inevitable. I think it will be pervasive and systemic. Further, I think that AI will restructure vast swathes of the economy, culture, and society. Where I disagree with many venture capitalists and other technology-adjacent people is on the timeframe of this disruption. I think the disruption will take longer to occur, and be more of a continuous process of adaptation, than many others seem to think. Most revolutions do not operate like a light switch. They’re a slow burn. Ignore the smoldering and you’ll eventually get scorched. Anticipate the eventual conflagration, and you’ll be well-positioned to take advantage of the inevitable.

Perhaps an anecdote will explain my reasoning. A while back, when Jon Stewart still hosted The Daily Show, I attended one of its tapings. I had never seen a TV show recorded live before, and I learned a lot about how the show I watched every day was produced. What has stuck with me over the years is simply the sheer number of people it takes to produce an episode of a TV show. You don’t just have the talent. You have makeup and wardrobe people. You have people who schedule guests. You have people who write dialogue. You have people who man cameras. You have people who operate lighting rigs. You have people who design and dress the sets. You have people who organize all of these people. You have people in the background who just manage all of the people managing other people.

Movies, too, are like this. Except that, since movie budgets tend to be much larger than TV show budgets, you have even more people. Movies and TV have a lot of costs embedded in their operations.

AI will eventually change a lot of this.

Mike Gioia

has a great post about these embedded costs:

While technology is lowering the cost of filmmaking for 99% of people, it’s enabling Hollywood to set new spending records every year. The avertage cost of producing an hour of television is about $3-5 million. More ambitious shows like the recent The Lord of the Rings series manage to spend $80 million per hour.

Film productions are money pits. As soon as you feed the pit, you can’t stop. Production wants more lights. Directors need more explosions. Unions require more assurances. Ironically, technology has only helped companies excavate deeper, more mysterious money pits. Where does the money go? It’s increasingly hard to say.

Jeff Bezos was famous for saying, of competitors, your margin is my opportunity. What he meant by this is that companies which have large margins, tend to have large margins because they are able to charge a higher price for their goods or service than they would if they had competitors. The opportunity, then, resided in underpricing competitors. If you’ve grown fat and lazy because you have no real competition, then all it takes is a more nimble and more efficient company to disrupt you and kill your margins.

Similarly, movie and TV production has been able to spend gobs of money simply because traditional movie and TV production has had no real competition to force costs to decline. But, as Mike Gioia notes later in his piece, generative AI promises to disrupt traditional movie and TV production:

Hollywood’s business is exporting 16:9 pixellated rectangles with sound. No matter the project, the final delivery is a digital file of color and sound. Producing these 16:9 rectangles requires a lot of expensive manipulation of physical matter– from closing city streets to rigging lights to directing crowds of people. Computers have increasingly been composing pixels in that rectangle. And the technology has reached an important frontier with generative AI. These AI models, trained on trillions of pixels across billions of rectangles, tease the ability to create the entire rectangle. For an industry revolving around the production and exporting of rectangles, that’s a pretty big deal. 

Over the last 18 months, artificial Intelligence has shown an astonishing capability to make high quality images and audio. Tools like Stable Diffusion, PikaLabs, and Eleven Labs can instantly create synthetic visuals and audio. These tools are very accessible to consumers. They’re readily available, relatively cheap, and don’t require specialized skill to operate– just writing commands in natural English and a good deal of patience. While the early results are good, they aren’t quite there yet. But there is most certainly a there there.

Already these tools are taunting Hollywood’s monopoly on production value, intellectual property, and celebrity. Tom Cruise can appear in anything now. Wes Anderson’s aesthetic is being applied to the Fellowship of the Ring. And dads are making Pixar movies based on their 7 year-old’s concepts. Well-funded lawsuits are doubtlessly piling up, but the cat’s out of the bag. There’s a big ecosystem of open-sourced models, Discord servers, and subreddits that are several steps ahead.

Of course the lawsuits are predictable. It’s much easier for Hollywood studios to call their lawyers, than to build new business models which leverage this new technology. It’s hard to get up the gumption to do actual work when the fetching lawyers down the street beckon you with their sweet, sweet bills. The law is a clumsy cudgel used by incumbents to maintain their status. As the studios feverishly shell out millions to lawyers charged with protecting the status quo, the disruptors will continue to improve their algorithms, and creative people will continue to operate, much more cheaply, outside the confines of Hollywood. Eventually, technology will enable people to make customized Hollywood-caliber movies via a series of AI prompts. Sure, most of the movies that people make will be crap, but some will be works of art. This is, by the way, the same thing we see today: most Hollywood movies are pretty awful and some are sublime.

Disruption is not a sudden, discontinuous change. Rather, it is a continuous process, an accretion of success layered atop success. Netflix didn’t disrupt Blockbuster simply by coming into existence. It disrupted Blockbuster over a series of years because it provided customers with access to a much larger supply of movies than Blockbuster ever could. Blockbuster couldn’t respond because its entire existence, its raison d’être, was predicated on maintaining the status quo, and living off the cash flows that the status quo generated. Hollywood movie studios (and Netflix!) today are in the same position as Blockbuster was decades ago. In fact, Mike Gioia includes the following chart in his post:

YouTube is disrupting the disruptor. Same as it ever was: technology tends to disrupt incumbents. AI will be no different. In fact, there is a reasonable argument to be made that AI’s disruptive effects will be more systemic, and take place over a shorter time frame, than previous technological revolutions. AI will force change across many different industries. Rumors are flying around, for example, that Google will lay off thousands of ad sales people in 2024 as it re-orients its sales efforts around AI. Way back in 2018 (well before ChatGPT was released to the public), the Brookings Institution released a report which noted how AI is disrupting the financial markets:

Investments in financial AI in the United States tripled between 2013 and 2014 to a total of $12.2 billion. According to observers in that sector, “Decisions about loans are now being made by software that can take into account a variety of finely parsed data about a borrower, rather than just a credit score and a background check.” In addition, there are so-called robo-advisers that “create personalized investment portfolios, obviating the need for stockbrokers and financial advisers.” These advances are designed to take the emotion out of investing and undertake decisions based on analytical considerations, and make these choices in a matter of minutes.

A prominent example of this is taking place in stock exchanges, where high-frequency trading by machines has replaced much of human decisionmaking. People submit buy and sell orders, and computers match them in the blink of an eye without human intervention. Machines can spot trading inefficiencies or market differentials on a very small scale and execute trades that make money according to investor instructions. Powered in some places by advanced computing, these tools have much greater capacities for storing information because of their emphasis not on a zero or a one, but on “quantum bits” that can store multiple values in each location. That dramatically increases storage capacity and decreases processing times.

Fraud detection represents another way AI is helpful in financial systems. It sometimes is difficult to discern fraudulent activities in large organizations, but AI can identify abnormalities, outliers, or deviant cases requiring additional investigation. That helps managers find problems early in the cycle, before they reach dangerous levels.

It’s easy to read all of this, and conclude that AI will disrupt everything, everywhere, all at once. And, while I think that the timeframe for disruption will be shorter than for previous technological revolutions, I also think it’s a mistake to conclude that the disruption will be instantaneous. AI technology is so revolutionary, excitable venture capitalists claimed, that we would all be on the breadlines. This has been a sci-fi trope for decades: AI obviates labor. And yet, the reality isn’t so simple. I commented on

Matt Beane

‘s excellent post:

When ChatGPT was first released in November 2022 I saw assorted VCs and other 'thought leaders' across the technology industry assert that it was going to put virtually all white collar employees out of a job within a year. None of these people, of course, had any experience working in mainstream trad corps, which is to say, companies akin to the airlines to which you repeatedly refer in this post. LLMs are extraordinarily powerful technology, and they will no doubt completely restructure companies' operations, but as you point out, that will take significantly longer than many pundits predict.

I want to unpack this statement a bit, because it gets at the heart of AI and disruption. Matt’s post is long, and it’s somewhat theoretical but it provides an excellent framework for understanding why disruption takes a while to unfold. The context of his piece is companies restructuring their operations to avail themselves of the benefits of generative AI technologies. Why, he asks, are organizations so slow to adopt new technology, when it is evident that the new technology is so much better than legacy tech?

It’s a trite statement to say that companies are complex things, and complex things are slow to change. It’s also accurate. Coca-Cola and other large traditional non-tech corps will re-orient themselves and restructure their operations to be AI-first, but it just takes time. Law firms, such as the UK’s Allen & Overy, have jumped on the AI bandwagon, but the AI tech they’re using is meant to help its clients with low value legal work like contract negotiations. There may be a future point where AI disrupts much of the practice of law, but we’re not yet at that point, and institutional inertia may well militate against it happening in the timeframe expected by the AI bulls.

I’ll conclude with this: the best way to understand how AI will disrupt everything is to understand that the disruption will occur at different speeds in different organizations. Further, while improvements to the underlying technology are rapid, and seem likely to continue to be rapid, don’t confuse improvements in technology with improvements in application. Just because a technology rapidly becomes more powerful or useful doesn’t mean that we also know how to apply the technology. Learning how to integrate new technologies into extant workflows requires time, negotiation, experimentation, and consideration.