Generally speaking, I believe that depending on any form of automation in your everyday life may be harmful when overdone and even alienating when utilized sparingly, particularly when it comes to interpersonal relationships. An AI assistant that compiles connections to other reading materials and arranges my to-do list? Amazing. An AI bot that sends my parents a weekly automated message with a brief report on my life? Terrible.
The environmental impact that these models continue to have during training and output generation, however, is still the best reason not to incorporate more generative AI tools into your everyday routine. In order to uncover more historical context for your topic, I combed through WIRED’s archive, which was published during the beautiful dawn of this mess we call the internet. After some research, I concluded that you probably already use AI agents on a daily basis.
The most popular term right now for any tech leader looking to promote their latest investments is the concept of AI agents, or even worse, “agentic AI.” However, the idea of an automated assistant that is committed to finishing software duties is not new. A lot of the discussion surrounding “software agents” in the 1990s is similar to the present discussion in Silicon Valley, where tech company executives are promising an influx of generative AI-powered agents that are trained to perform online tasks for us.
According to a 1995 WIRED interview with MIT scientist Pattie Maes, “One concern I see is that people will wonder who is responsible for the behavior of an agent.” Particularly when agents spend excessive amounts of time on a computer or make purchases on your behalf that you do not desire. Although agents will bring up many fascinating topics, I firmly believe that we cannot survive without them.
I called Maes early in January to hear how her perspective on AI agents has changed over the years. She’s as optimistic as ever about the potential for personal automation, but she’s convinced that “extremely naive” engineers are not spending enough time addressing the complexities of human-computer interactions. In fact, she says, their recklessness could induce another AI winter.
“The way these systems are built, right now, they’re optimized from a technical point of view, an engineering point of view,” she says. “But, they’re not at all optimized for human-design issues.” She focuses on how AI agents are still easily tricked or resort to biased assumptions, despite improvements to the underlying models. And a misplaced confidence leads users to trust answers generated by AI tools when they shouldn’t.
To better understand other potential pitfalls for personal AI agents, let’s break the nebulous term into two distinct categories: those that feed you and those that represent you.
Feeding agents are algorithms with data about your habits and tastes that search through swaths of information to find what’s relevant to you. Sounds familiar, right? Any social media recommendation engine filling a timeline with tailored posts or incessant ad tracker showing me those mushroom gummies for the thousandth time on Instagram could be considered a personal AI agent. As another example from the ’90s interview, Maes mentioned a news-gathering agent fine-tuned to bring back the articles she wanted. That sounds like my Google News landing page.
The issue with relying too much on this kind of AI agent is the slow burn of complacency. As more of the content I consume is fetched for me through personalized recommendation algorithms, I will likely make increasingly monotonous decisions and stumble upon fewer surprises. If I used an AI agent to compile my list of groceries every week and shop online through a delivery service, I’d enjoy the streamlined process at first, but probably would be less inclined through repetition to experiment with new dishes. Life becomes an exercise in creeping blandness.
As for AI agents that represent you, they are also mere algorithms, but these are designed to learn your behavior and mimic it, acting on your behalf during online interactions. It’s the fantasy of a digital twin that can attend video meetings for you, write thank you notes to your wedding guests, and call a restaurant to reserve you a table. This also includes generative AI tools made to impersonate famous people and auto-interact with their fans.
I’d avoid these AI replicas at all costs, whether deploying one or interacting with one. The digital world has eclipsed the physical world in importance, and our lives are now largely mitigated through screens. Even if the tech is eventually good enough that sending an AI agent to FaceTime my grandma or letting it loose in the group chat to riff with my friends doesn’t feel like a laughable idea, let’s leave the automation to the dull tasks. We should show up as ourselves in online interactions. It’s increasingly all we have left.
