Generative AI: Not So Much a Tool But Something Quite Different
August 24, 2023
Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.
Thirty years ago I had an opportunity to do a somewhat peculiar job. I had written for a publisher in the UK a version of a report my team and I prepared about Japanese investments in its Fifth Generation Computer Revolution or some such government effort. A wealthy person who owned a medium-sized financial firm asked me if I would comment on a book called The Meaning of the Microcosm. “Sure,” I said.
This tiny, cute technology creature has just crawled from the ocean, and it is looking for lunch. Who knew that it could morph into a much larger and more disruptive beast? Thanks, MidJourney. No review committee for me this morning.
What I described was technology’s Darwinian behavior. I am not sure I was breaking new ground, but it seemed safe for me to point to how a technology survived. Therefore, I argued in a private report to this wealthy fellow, that if betting on a winner would make one rich. I tossed in an idea that I have thought about for many years; specifically, as technologies battle to “survive,” the technologies evolve and mutate. The angle I have commented about for many years is simple: Predicting how a technology mutates is a tricky business. Mutations can be tough to spot or just pop up. Change just says, “Hello, I am here.”
I thought about this “book commentary project” when I read “How ChatGPT Turned Generative AI into an Anything Tool.” The article makes a number of interesting observations. Here’s one I noted:
But perhaps inadvertently, these same changes let the successors to GPT3, like GPT3.5 and GPT4, be used as powerful, general-purpose information-processing tools—tools that aren’t dependent on the knowledge the AI model was originally trained on or the applications the model was trained for. This requires using the AI models in a completely different way—programming instead of chatting, new data instead of training. But it’s opening the way for AI to become general purpose rather than specialized, more of an “anything tool.”
I am not sure that “anything tool” is a phrase with traction, but it captures the idea of a technology that began as a sea creature, morphing, and then crawling out of the ocean looking for something to eat. The current hungry technology is smart software. Many people see the potential of combining repetitive processes with smart software in order to combine functions, reduce costs, or create alternatives to traditional methods of accomplishing a task. A good example is the use college students are making of the “writing” ability of free or low cost services like ChatGPT or You.com.
But more is coming. As I recall, in my discussion of the microcosm book, I made the point that Mr. Gilder’s point that small-scale systems and processes can have profound effects on larger systems and society as a whole. But a technology “innovation” like generative AI is simultaneously “small” and “large”. Perspective and point of view are important in software. Plus, the innovations of the transformer and the larger applications of generative AI to college essays illustrate the scaling impact.
What makes AI interesting for me at this time is that genetic / Darwinian change is occurring across the scale spectrum. On one hand, developers are working to create big applications; for instance, SaaS solutions that serve millions of users. On the other hand, shifting from large language models to smaller, more efficient methods of getting smart aim to reduce costs and speed the functioning of the plumbing.
The cited essay in Arstechnica is on the right track. However, the examples chosen are, it seems to me, ignoring the surprises the iterations of the technology will deliver. Is this good or bad? I have no opinion. What is important than wild and crazy ideas about control and regulation strike me as bureaucratic time wasting. It was millions a years ago to get out of the way of the hungry creature from the ocean of ones and zeros and try to figure out how to make catch the creature and have dinner, turn its body parts into jewelry which can be sold online, or processing the beastie into a heat-and-serve meal at Trader Joe’s.
My point is that the generative innovations do not comprise a “tool.” We’re looking at something different, semi-intelligent, and evolving with speed. Will it be let’s have lunch or one is lunch?
Stephen E Arnold, August 24, 2023
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
