Tech Medic Logo

Every New Tool Changes Us

This is my first blog post, something I've been meaning to start for years. It seemed right to begin with the biggest change I've seen in all my time doing this work.

About 2,400 years ago, in one of Plato's dialogues, Socrates tells a story about an Egyptian king who was offered a new invention: writing. The king warned that it "will produce forgetfulness in the minds" of the people who learned it, and that they would seem wise without being wise. He had a point. And the only reason we know the story is that Plato wrote it down. Every big new tool since has drawn some version of the king's worry, from the calculator to the smartphone.

I've spent nearly 25 years helping people take on new tools, and I've come to believe that every tool that does some of our thinking gives us something and takes something back. AI is the biggest of those tools yet, and the fastest. By AI I mean programs like ChatGPT, or the written answers that now show up at the top of a Google search: you type a question in ordinary words, and you get a written reply.

Our November newsletter covered the practical side. This is the longer version, with more room to wonder about what AI is doing to us and what we still get to choose.

What Our Tools Do to Our Minds

Most of us already sense the trade. In a 2025 Pew survey, more than three times as many Americans said AI will make us worse at thinking creatively as said it will make us better. About half of Americans now say they're more concerned than excited about AI, up from 37% in 2021.

The research so far points the same way. Drivers who leaned on GPS for years had a harder time finding their way without it. In a small MIT study, people who used a chatbot to write essays had the most trouble quoting their own essays afterward, and in a survey of people who use AI at work, those who trusted it more reported thinking less critically. The MIT team's brain recordings also showed the least coordination between brain areas in the chatbot group while they wrote. Much of this research is new and small, so I hold it loosely.

Here's the hopeful part. In a study at Middlebury College published this year, students who had AI write their essays lost that boost a week later when they wrote without it, while students who used AI to explain ideas held on to more of what they learned. How they used it made all the difference.

So my advice is to choose what you hand over. For a skill you want to keep, use AI as a tutor, not a ghostwriter: ask it to explain, and ask it to check its own work. And try things yourself first, then compare. You'll learn more, and you'll notice when it's wrong. For a decision, it can be a strategic partner that raises questions you hadn't thought of. And for work you'd never try yourself, let it do the work: that's where it gives the most and takes the least.

It's worth saying plainly how much good these tools can do. AI can read a menu aloud for someone with low vision, translate a conversation on the spot, or help someone who struggles with writing get their thoughts onto the page.

The Electricity and Water Behind the Answers

Writing never needed a power plant. Every answer an AI gives comes from a data center, a building full of computers that run hot, draw a great deal of electricity and often use water to stay cool.

One question uses very little, about what an LED bulb uses in a minute or two. But there are billions of questions a day, and a building boom behind them. Data centers already use about 4% of this country's electricity, and a federal lab expects that to reach somewhere between 7% and 12% by 2028. Worldwide, the International Energy Agency expects data centers' electricity use to roughly double by 2030.

Water depends on what you count. Google says a typical question uses about five drops, counting only the water that cools its own data centers. A university team that also counted the water used at power plants estimated a bit more than a tablespoon per answer, for an older AI model at an average U.S. data center. Across the country, data centers used about 17 billion gallons of water directly in 2023, and the power plants supplying them used many times more.

The companies building fastest see it in their own numbers. Google's data centers used 27% more electricity in 2024, and Microsoft says its emissions rose 25% in the year ending June 2025, mainly from expanding its data centers. Google's total emissions rose 11% in 2024, mostly, it says, from making servers and building data centers. Here in New England, the group that runs our power grid says data centers haven't been a driving factor in wholesale power costs so far, since not many have been built here. Only two big new power users of any kind are even in formal study here.

There's another side. The International Energy Agency thinks AI could help cut emissions over the next decade if it's put to good use, though nothing guarantees it.

How we use it matters too. Long conversations, big attachments and "deep thinking" modes take much more computing than a quick question. Using a quicker mode for simple questions, and starting a fresh chat for a new topic, saves energy and often time. Managing that well is one of the things I hope to teach in a class next year.

None of this means you shouldn't ask an AI a question. It means the bigger choices about building and powering data centers, for our electric bills, our water and our air, are worth paying attention to.

Can AI Protect Us from AI?

Since the beginning, a big part of our work at Tech Medic has been setting good software against bad: antivirus, spam filters, updates. The lesson is that the tools help enormously and never catch everything, which is why good protection comes in layers.

Defensive AI is already one of those layers. In a 2025 contest run by DARPA, the Defense Department's research agency, AI systems found 54 of 63 flaws hidden in test software, fixed most of them, and turned up 18 real ones nobody had planted.

But the other side has the same tools. The FBI logged more than 22,000 complaints that mentioned AI in 2025, with almost $900 million in reported losses, and says many victims don't realize how much AI may be involved. The biggest share was investment scams, where AI-made videos and voices of celebrities and executives push fake opportunities: more than $632 million in reported losses.

In May 2026, Google reported catching a new kind of attack it believes was built with AI. This is one reason we keep asking clients to install updates promptly: they close the holes that attacks like this look for.

Scammers already use cloned voices to mimic a loved one in distress: the FBI counted more than $5 million in reported losses to these "grandparent" scams in 2025. When a fake can sound like someone you love, the safest test is something you can check yourself. The FBI suggests a secret word or phrase that your family agrees on ahead of time. Make it something made up rather than a family fact like a pet's name, which can often be found online, and agree on it in person or by phone, never in a text or a post. If you get a call that sounds exactly like a family member in trouble and asking for money, ask for the word. If they can't give it, or anything feels off, hang up and call them back on the number you already have. And when a call claims to be from your bank or another organization, hang up, look up their number yourself, and call it directly. Nobody can say how many scams these habits stop, but they're free, and a copied voice can't give a word it never heard.

The FBI also suggests keeping your photos and voice off public posts where you can: make social media accounts private, and limit followers to people you know. And never send money, gift cards or cryptocurrency to someone you've met only online or over the phone.

So here's my answer: defensive AI is a needed layer, never the whole wall. And any AI is only as good as whoever controls it.

Work, Learning and Who We Are

If you have children or grandchildren just starting out, the changes to work may touch them first. Stanford researchers found that young people aged 22 to 25, in the jobs where AI can do much of the work, have about a fifth fewer jobs than you'd expect, mostly because fewer are being hired. Those first jobs are where many of us learned the ropes. The researchers are careful to say it's early, and they see no wave of job losses across the economy.

Work is only part of who we are. At any age, AI can be a patient tutor that never minds a fourth question. It can explain a confusing letter or walk you through a new phone setting, which for many people means doing more on their own. Independence grows when AI helps us do things ourselves, and shrinks when we can't do them without it.

It can quiz you on a few phrases before a trip, help you dig into a family history, or explain a medical term in plain words before you talk it over with your doctor. Some people use it to brainstorm a toast or plan a garden, then make it their own.

For a small business, AI can be an extra pair of hands for work there was never time or money to hire for, like drafting a proposal or summarizing a long report. I think some people will shift into more of an executive mindset, managing the work of their AI helpers rather than doing every step themselves. Small-business owners already know that job: deciding what to hand off, and checking what comes back. Tech Medic is a small business too, and my own work has shifted that way. AI helps us research and troubleshoot faster, while people still make the decisions and keep the relationships. More and more, what I bring is judgment and trust.

A fair warning from my own use: getting AI to take repetitive work off your plate takes more time up front than you'd expect. You teach it how you like things done, then test, adjust and test again before you can trust it. It gets quicker as you build skill and it gets to know you and your systems, but for a job you need done right and soon, the old way is often still faster.

For a lot of us, part of who we are is what we know and what we can do. When a machine can do some of that, it can feel unsettling. I don't think the answer is to race it at everything. It's to focus on what people bring: paying attention, caring, deciding, and being someone others can count on.

We've Rehearsed This

We've lived through one fast, powerful rollout already. Social media brought real gifts: photos of family, old friends found again, news from people far away. It also took our attention in ways none of us chose, through feeds, alerts and autoplay designed to keep us scrolling. Tristan Harris, who co-founded the Center for Humane Technology, spent years warning about exactly that. In his 2025 TED talk, at the end of this post, he argues that those harms were preventable, and that we don't have to repeat them with AI.

People say AI is a tool that will be used one way or another, and I think that's true. It will be used, for good and for harm, whether or not any of us likes it. But how carefully it's built, and how wisely each of us uses it, are still choices worth making. Inevitable doesn't mean we have no say in how.

Those choices aren't spread evenly. Only a few companies can build the most capable AI, because it takes those enormous data centers. In January 2025, staff at the Federal Trade Commission studied the partnerships between the big cloud companies and AI developers and flagged them as an area to watch: they could affect who gets access to computing power and engineering talent, and make it harder for the AI developers to switch providers. The study didn't assess whether anyone broke the law. Harris warns against two dead ends: powerful AI in everyone's hands with no one responsible, which he calls "chaos," or all of it controlled by a few companies or governments, which he calls "dystopia." My own worry is that the first could become the excuse for the second. If anyone can fake anything, the pressure grows to verify who we are for everything we do online, and checking everyone that closely would mean more surveillance, in fewer hands. Between the two, he says, "we should seek something like a narrow path, where power is matched with responsibility at every level."

Nobody really knows how fast all this will move, and the people closest to it disagree. Some expect huge change within a few years; others think it will take decades.

Choosing, This Time

I said at the start that every tool that does some of our thinking gives us something and takes something back. I think how much AI takes depends partly on the lines we each draw.

For my part, I let AI take on work I'd never try myself, and I lean on it to explain things and to raise questions I hadn't thought of. But the final word stays mine, and when it gives me a fact, a figure or a deadline that matters, I check it against the original source before I act on it.

Drawing your own lines takes no technical skill. Keep doing the things you want to stay good at, even when AI could do them faster. An AI can sound sure and still be wrong, so double-check anything important. Don't type anything into it that you wouldn't want a stranger to read. In a small office, agree on what never goes into an AI tool, like passwords or private customer details, and on who checks AI's work before it goes out. And if your family gathers this holiday season, take a minute to agree on that secret word in person. Then ask the young people at the table how they use AI. You might learn from each other. And some days, leave it closed. If you'd like a hand with any of it, you know where to find us.

None of us can decide alone where AI goes, but each of us can decide how it fits into our own lives. That gives me hope.

Every new tool changes us. This time, we get to notice while it's happening, and choose.

Blessings,
Jack

If You'd Like to Go Further

The Center for Humane Technology, the nonprofit Harris co-founded in 2018 with Aza Raskin and Randima Fernando, applies the lessons it draws from social media to technology like AI. Tech Medic has no connection to it; I just think its work is worth your attention. Following its work is free: watch the talk below, listen to its podcast, Your Undivided Attention, or sign up for its newsletter.

If you're thinking of giving, know what a gift supports. This fall the Center narrowed its work to reaching the public through films, podcasts and talks, and briefing the people who make decisions about technology. It's closing its own programs for policy, legal cases and research. Its update has the details, and if that's work you want to support, you can give at humanetech.com/donate.

Here's Harris's 2025 TED talk. It's about 15 minutes and more sobering than this post. It includes policy ideas of his own, which you may or may not share; I'm sharing it for the big picture, and because it ends on hope.

Harris points to times the world faced what looked like an unstoppable race and chose another path: the Nuclear Test Ban Treaty, limits on editing human genes, and fixing the hole in the ozone layer. Near the end he says, "Restraint is the central feature of what it means to be wise."

Claude, an AI assistant, helped me research and check the facts for this post; the views and the final words are mine.

Sources

Questions about your tech? We're here to help.

Book an Appointment

More from the Tech Medic Blog