Should I set up a personal AI agent to help me with daily tasks?

—Ask for help

Generally speaking, I think relying on automation of any kind in daily life is dangerous if taken to extremes, and even when used in moderation, can create a sense of alienation, especially in terms of personal interactions. An AI agent that organizes my to-do list and collects online links for further reading? marvelous. An AI agent that could automatically send my parents a quick life update every week? It's terrible.

However, the strongest argument for not incorporating more generative AI tools into daily work remains the environmental impact these models continue to have during training and output generation. With all of this in mind, I dug into the archives of Wired magazine, which was published during the glorious dawn of chaos we call the Internet, to find more historical context for your question. After some searching, I am sure that you may already be using AI agents on a daily basis.

The idea of ​​artificial intelligence agents, or God forbid “agent AI,” is the current buzzword for every technology leader trying to hype their latest investments. But the concept of automated assistants dedicated to completing software tasks is not a new idea. Much of the talk about "software agents" in the 1990s mirrors the current conversation in Silicon Valley, with tech company leaders now promising the emergence of legions of artificial intelligence-powered generative agents trained to do online chores on our behalf.

“One of the problems I see is that people question who is responsible for an agent’s actions,” wrote MIT professor Pattie Maes in a Wired interview originally published in 1995. "Especially things like agents taking up too much time on machines or buying things you don't want on your behalf. Agents raise a lot of interesting questions, but I don't believe we could survive without them."

In early January, I called Mays to find out how her views on AI agents have changed over the years. She's as optimistic as ever about the potential of personal automation, but believes "extremely naive" engineers haven't spent enough time grappling with the complexities of human-computer interaction. In fact, she said, their recklessness could trigger another AI winter.

“The way these systems are built right now is optimized from a technology perspective, from an engineering perspective,” she said. "However, they are simply not optimized for human design problems." She is concerned that despite improvements in the underlying models, AI agents can still be easily tricked or resort to biased assumptions. False confidence can lead users to trust answers generated by AI tools when they shouldn’t.

To better understand other potential pitfalls of personal AI agents, let’s break this nebulous term into two distinct categories: those that feed you and those that represent you.

A feeding agent is an algorithm that contains data about your habits and tastes and can search through vast amounts of information to find content relevant to you. Sounds familiar, right? Any social media recommendation engine that fills my timeline with customized posts, or a non-stop ad tracker that shows me those gummy mushrooms on Instagram for the thousandth time, can be considered a personal AI agent. As another example from an interview in the 1990s, Mays mentioned a newsgathering agent that was fine-tuned to bring back the articles she wanted. This sounds like my Google News landing page.