Is generative AI codifying average?
There’s no such thing as the average person, or so the wisdom goes.
The logic says that if you were to create a profile of the average man or woman through a variety of factors—height, weight, income, weight, tolerance for Taylor Swift, etc.—you wouldn’t be able to find the real-live version of that person. (So, the next time your friend says they just wanna be average, let them know that they’re chasing one of the least attainable goals of all time.)
If the average human ideal doesn’t exist in flesh, might it exist digitally? This idea has stuck with me since I heard Dennis Yi Tenen make the following point about generative AI—more specifically, large language models (LLMs)—on episode 265 of Douglas Ruskhoff’s Team Human podcast1:
In a way, you’re having a conversation with an average . . . Imagine having a conversation with a thousand—or a hundred thousand—people, and I’m going to kind of average out the answer.
AI is math. A lot of math done really fast. But it’s math. While LLMs appear to be capable of thinking, they’re in fact just guessing with math. When answering a prompt, LLMs try to predict the best answer based on the most probable outcomes based on its training data.
So it appears that the developers of generative AI and LLMs have made average more accessible and more affordable, more quickly. And companies investing heavily into incorporating this technology into their everyday business may very well be investing a lot of time and money, and taking a lot of risk, for average.
Average is not smart. Average doesn’t stand out. So, average is bad business. Might that same money be better spent on something that makes the business special and more competitive?
With the help of LLMs, we’re one step closer to codifying average. In a matter of seconds after prompting, we can see what the average answer looks like for anything we’re curious about. If you need help just getting by, then average may be fine. But innovation and insight don’t emerge from average. Any Seth Godin2 fan knows that average is death for a business. Average means you can be easily swapped for another business.
Maybe average is fine for certain tasks that people are using LLMs for. But businesses should be sure that using generative AI helps them add real value elsewhere. Or, when all these businesses are using the same generative AI from the same small handful of vendors, they’ll most likely sound like every business in their niche.
With the help of generative AI, it’s becoming easier to bring average to the masses. And if that’s all the AI community is doing, then how long until the bubble bursts and the industry falls back down to a healthier average in terms of valuation?
Jake LaCaze is embarrassed to admit he’s a middle-aged man who finds himself bouncing to Olivia Rodrigo tunes. ‘vampire’ is a banger, as the kids these days say. But it’s also quite human.