2022 has been an astounding year in AI. I don’t think anyone needs a recap (and I’m not the guy to provide one), so I’ll namedrop Stable Diffusion and GPT as examples of what I’m talking about.
While only the most jaded person could see what’s going on in AI and not be completely blown away, there is debate around exactly what the socioeconomic implications of this new phase of AI will be. On one hand, you have people already predicting that we’re in the end game for white collar workers, such as artists, writers, and software developers. Other folks, perhaps more attuned to the limitations of the AIs we’ve seen, are urging a much more sober outlook. I see massive change on the horizon, but along a different axis than some of the more simplistic predictions.
From perspective of AI skepticism, we have the Chaos Engineering newsletter’s evaluation of the applications of AI to FinTech. Francisco’s point is that the breakthroughs aren’t necessarily applicable to existing business problems in FinTech. His reasoning seems pretty good, and I largely agree.
But I think it would be wrong to conclude that–just because it doesn’t solve existing industry pain points–generative AI isn’t about to have a massive societal impact. The tech is at a point where it will prove good enough to solve things businesses weren’t even thinking of as problems. Breakthroughs in technology make their biggest impacts in ways that are often tangential to existing models–if they make an impact at all. They surprise us.
New technology is rarely exactly what’s needed to revolutionize approaches to existing high value problems. It’s one of the reasons blockchain boosters have been so wrong for so long, for example. While the blockchain is an incredibly impressive technology, it maps to very few business use cases.
Often, it is unpredictable combinations of technologies that change the world, some of which are really boring. For instance, you don’t really get the iPhone smartphone revolution without a couple tedious decades of incremental battery technology and mobile network improvements. But once you had Li-ion batteries capable of powering a full-color display for a full day and affordable data plans, suddenly, entirely new mobile experiences became possible.
ChatGPT is an application of a particular AI tech called large language model (LLM). It is astonishingly good at generating text that sounds like it is reasoning like a smart human, in response to natural discussion prompts. That means it can be effective in situations where it is important to get a human to know or believe something. It’s a breakthrough in human-computer interface technology.
I’m not sure LLMs, in and of themselves, will change the world, beyond being impressive toys. However, if you combine these LLMs with the social reach of social media, the power of cloud computing, and APIs that can cause real world effects, we’re on the verge of seeing autonomous social systems with massive amount of leverage to influence real world systems. Maybe also throw in Bitcoin as a way for autonomous systems to share a rudimentary economy, without permissions or access to banking.
What do the applications of this kind of technological mashup look like? I have no idea. I’m rarely the person to come up with outside-the-box ideas. It just feels to me that the deck is stacked for some wild stuff, the barriers to entry are extremely low, and I see almost no guardrails.
We may not be on the verge artificial general intelligence, but I don’t think you need AGI for a quantum leap in societal impact of AI.
The only thing I know for sure about what happens next is that we’re in for a wild ride.