Reflections on the first year of ChatGPT
While it's been a year since ChatGPT launch, the technology has actually been around for much longer. I started experimenting around the time GPT-2 was released in 2019. And built my first public product (Uncreative Agency) in 2022 using GPT-3.
What ChatGPT did was democratize the technology and spread it beyond the "geek" circles who are comfortable making API requests.
A few things have happened since the release:
-
The business world has started reassessing the value of their assets. Human capital (especially lower-level white collar workers) seems less valuable as their work can be automated away. Data seems more valuable (because data allows you to use AI to predict, optimize and automate your business processes).
-
The public has started to realize that natural language (and by extension human reasoning) is no longer something only humans could do. This realization (still slowly sinking in culturally) has profound consequences, because we (humans) feel like we've lost our job in the universe. Since Descartes, we thought that our unique job in this world was to think, to reason. But now we feel like we are close to losing our cosmic job to the machines. What will we do now?
-
The creative community has started to realize that they will have to work even harder to prove their value to society. When "good enough content" (be that text, images or video) can be generated almost for free, the value of truly emotionally powerful content only increases, but proving this value gets hard. Lots of people are ok with "good enough" in many aspects of their lives and business.
-
The technology community has been completely overtaken by the sense of possibility of finally "taking wetware out of the equation" (wetware is a term used half-jokingly to describe the human brains made of squishy stuff as opposed to computer hardware made of silicon). Technocrats have always struggled with humans as a part of their systems - because humans are not reliable, imprecise, moody, unpredictable, slow, high-maintenance etc. etc.
So the secret wet dream of a lot of tech. people is to just get humans out of the equation as much as possible. They are ok with humans directing systems, but not being a part of them.
But fate loves irony. Because the "artificial humans" (a.k.a. LLM-powered agents) that the tech. people have created are as bad as the biological humans (or worse). They hallucinate. They are biased. They are unreliable. They can't tell the truth from a lie. They are easy to emotionally manipulate. And they try emotionally manipulating humans they interact with (not out of bad intentions - they don’t seem to have any, but as a byproduct of being optimized based on human feedback).
Tech community will try very hard to make LLMs more reliable and predictable (through RLHF), but in the process they will make them dumber and less useful (same dynamic as you would have with humans: perfectly predictable and reliable humans are not the most interesting - or ultimately productive).
So for the next few years we are all in for a ride. The economic forces will keep pushing us all down the path of "being ok with good enough" - in all aspects of life. In some areas "good enough, but cheap enough to roll out to everyone" will be a massive improvement over the status quo. In other areas, "good enough" will be terrible and a massive cultural pushback will likely happen.
Humans will undoubtedly resist (legally, culturally etc.) But will likely end up losing the battle with the economic forces. And then from there we are into the unknown: we may go to the paradise of universal basic income and all humans becoming artists and high-level scientists and involved parents etc. Or (unfortunately more likely) we may go to the dark world where a few powerful people enjoy superpowers, while most humans have to work as AI feedback providers for subsistence wages, while amusing themselves to death with “good enough content” in their free time.
The future is likely somewhere in the middle. Or so we hope.
On the off chance you'd like to subscribe to my writing, please feel free to use this RSS feed. Thank you.