Just this morning, one of my AI agents suggested I send flowers to someone as a thank-you. It was a good idea, and it offered to help me order them and add a personalized note. Somewhere between appreciating the suggestion and clicking through, I had a second thought. This thing knows a lot about me, a whole lot.
A deluge of AI products is out there. In the last few weeks, since the AI companies have started offering personal agents, we have Muse, Dots, Duet, and at least 4 new models. And yes, I have signed up for most of them. Curiosity and a weird AI FOMO have taken me over. Over time, I’ve connected more and more of them to my Google account, to my computer, letting them read portions of my email and see what I’m doing. I’ve realized it has gone too far. It started with Google AI Social network based on my Gmail, one that I just erased and already forgot its name. Then Muse; I signed up, and I find it really useful. Then Claude Duet came out. Then Dots (the only one I don’t have). Each comes with its own offers and its own connectors, and each one puts its claws a little deeper into our digital lives.
Why does this matter so much? In a conversation yesterday, I started thinking about the fact that roughly half of our lives is now digital. This is where we get our information. This is where we communicate and get our news. For the most part, these are the same devices we use for entertainment. When we connect agents to our digital lives, we are not connecting them to a side corner. We are connecting them to half of who we are, a very private half. We've already relinquished a lot of privacy, but with AI we are giving it access to the full personal stack.
Part of my problem is occupational. I’m deeply entrenched in the AI world, and I feel almost compelled to try every new product and see what works. I experiment and try to squeeze more productivity for my life.
But I’m realizing that what I really want is one product that works really well and does most of what I want an AI to do. And I want to limit what it knows about me. I know it is a contradiction in terms; a personal agent can be useful only if it knows a lot about me. Am I willing to pay the price? The second challenge is that I have gotten rid of most of my social media, and what I use I try to use sparingly, so I don’t want a new steady stream of intriguing features that turn out to be one more way to capitalize on my engagement.
My advice to myself and for anyone else navigating this world of AI and agents. With any product that you want to try, start with:
Does it give back more time than it takes out of your day, your week, your evening?
If it takes more time than it gives back, it’s the wrong product. We should not repeat our mistake with social media that capitalized on our time and engagement. Long conversations or what seem to be helpful suggestions that make you browse and go down rabbit holes online may seem delightful at first, but I quickly found them consuming my time.
Does it feel invasive?
Do your agent's suggestions feel helpful, or do they feel invasive? Imagine asking AI about food for pre-diabetes; the next thing it starts encouraging you to exercise more or sell you nutritional supplements? At first, it feels helpful, but I found it quickly becomes disturbingly personal.
Does it do most of what you need it to do?
If it does most of what you need just let it go. You do not change your browser nor buy new socks every week, so resist the urge to switch. If your agent leaves a gap, maybe you need a second one; in my case, I am thinking of a personal agent for my personal life and a work agent.
Separating work and Personal agents
The need for separating work and personal agents is especially true if work pays for the agent- we probably should not use a job-related agent for our personal lives, ethically but also from a privacy perspective, since your work will have access to anything personal you ask. If you have time to experiment, try that second product for a different use. Then stop there; today's frontier agents are all excellent.
Next
As for me, I am starting to disconnect my extra agents. Yes, it is easy to connect another agent (Silicon Valley’s magic is reducing friction), but the cost to privacy, engagement, and sanity may be too high.
At the end of the day, I do not want all of these companies connected to all of my data in multiple ways. That sense of invasion of privacy, of being taken over, is something I’m balking at. One good agent is plenty.
No comments:
Post a Comment