AI agents will soon surpass us

I was talking with my daughter the other day. We made a bet five years ago that I lost. I bet her that by this week non-fungible tokens (NFTs) would be a major part of gaming. I was obviously wrong. She made the point that I don’t know anything about the gaming community, and she’s clearly right about that. 

She did admit that I was right about one thing — that agentic coding would become the norm. (I wish I had bet her about that instead, but I didn’t see that coming at all five years ago…) Like many young people, she was skeptical about AI in general, and didn’t like the idea of using it to write code. Leveraging LLMs to build games is frowned upon, but her mind was changed when she read about a company that used AI to solve a gaming problem that a Unity engineer told them was unsolvable. Apparently the gaming community considered this very controversial.

It would seem that I truly don’t understand the gaming community. While they appear to be concerned mainly about the “harm” that AI is doing, I am far more interested in the fact that AI solved a problem that was considered not solvable. I think we are all going to need to get used to that. 

Not unsolvable

AI has been doing this kind of thing for a while now. Mathematicians have been feeling it strongly, as AI has been solving problems that have stumped us humans. For example, this past May, an OpenAI internal model “disproved a central conjecture in discrete geometry,” the Erdős unit distance problem. Now, I’m no mathematician, but “disproving a central conjecture” seems like a pretty big deal. The model’s result wasn’t a complete answer, but it opened a door that humans quickly walked through to a place no one had conceived of yet. 

Fábio Caversan, the CTO of the Stefanini Group, made a similar point to me recently. “Get used to the fact that whatever you code today will be outdated in six months,” he said.

And it occurred to me that we are about to realize that not only is AI better at coding than we are, soon it will be orders of magnitude better. As in, “do things we haven’t thought of yet” better. If Caversan is right, AI code generation will soon be overtaken by something even more functional — and it will produce things that we haven’t imagined yet. 

Open doors

Right now we are expending a lot of effort to review AI-written code to make sure that it is up to snuff. But perhaps we need to start preparing for a world where the code that agents write is so different that we can’t even understand what they are doing. We probably aren’t far from the code not mattering — not because the agents write such good code, but because the agents write code that we haven’t seen before and can’t currently conceive of.

If the code becomes something that we humans can’t or don’t need to understand, then how do we trust it? It would seem that the job of “software developer” changes almost to that of a QA engineer — validating instead of producing. If the agent produces something that runs and does exactly what we ask of it in a manner we don’t care about, doesn’t the job become proving that what it does is correct and what we want? As with the Erdős problem, we probably shouldn’t take the agent’s word for everything, but simply walk through that open door. 

We humans need to realize that when we say something “can’t be done,” what we really mean is “no human has done it yet.” 

This article has been indexed from InfoWorld

Read the original article:

AI agents will soon surpass us