Agents and Humanity
My friend showed me ChatGPT in 2021 at his dorm at RIT. We'd been talking about AlphaGo, the program that beats the best Go players in the world using neural networks, and he pulled up this tool some students were using for research. We sat for hours feeding it prompts about philosophy and politics, the kind of stuff you talk about at 18 when a computer is answering back. I felt excited and a little scared. That feeling hasn't gone away.
The models we use now, ChatGPT, Claude, Grok, are getting better fast. They don't just answer questions. They take initiative. I'm building a system to identify real estate opportunities, and I use Anthropic's Claude as a coding partner. I hand it tasks like searching code, running tests, committing to GitHub, and it handles them on its own. I recently added Model Context Protocols (MCPs), which let the AI use tools like web search and file management without me telling it to. At this point it's more coworker than software.
So where is this going? Machines making us more productive is an old idea. Aristotle wrote about technology freeing people from drudgery. What we have now is different though. These systems can figure out what you need and do the thinking on their own. I.J. Good called this the "intelligence explosion," where a smart enough machine designs an even smarter one, and the cycle keeps going faster than anyone can keep up with.
Even working in this space every day, I sometimes fall behind on what's new. If machines start driving their own improvement, humans might need to upgrade too, possibly through brain-computer interfaces. The alignment problem is real. The paperclip maximizer is a good way to think about it: an AI with the single goal of making paperclips could, if unchecked, convert everything available into paperclips, us included. That's extreme, but the concern underneath it isn't. As these systems get more autonomous, keeping their goals lined up with ours is the most important problem we have.
The best case is these systems make us better at what actually matters. There are real risks, but so is the opportunity.