What is AI doing to our memory?

Looking Outside is a newsletter offering a fresh perspective on familiar topics. Written by Joanna Lepore, without AI aid. For more stuff like this, check out the podcast.

We're talking a lot about the upsides (like productivity) and downsides (like bias) of AI. There’s one thing missing from the conversation: its impact on our human memory.

We rely on our memory to learn, but we also strengthen our memory by learning.

Whether in scientific experiments or artistic creations, the process of doing is how we push ourselves to create those things that make us uniquely human; things seemingly at the limits of our capabilities. Sometimes beyond our rational and moral limits. When we create, design, draw, write, snap, paint, we curate. That process matters, perhaps more than the outcomes. That's why we often quote Thomas Edison -

“I have not failed 10,000 times. I've successfully found 10,000 ways that will not work”.

What happens when we outsource the learning process?

The process is a critical part of invention. That’s how novel ideas come to us. Adam Grant writes about this; how some of the best inventors throughout history didn’t have a lightbulb moment, that they kept writing, refining, spitting out and playing with hundreds of ideas.

Sharing memories is also a way we learn. And how we attach meaning to moments. You want to take a long drive through snowy mountains and stay in a secluded cabin with a hot chocolate that burns your lips and a heavy book in your lap because you’ve likely already done this before, or that feeling has been painted so viscerally for you by another storyteller.

The oldest and one of the most successful ways for us to learn is through storytelling; passing down of our experiences. We know that when one person tells a story the listener’s brain mimics it. The brain lights up. It feels real, even if you didn’t personally experience it.

Do we have that same effect when we read a story from AI? Can you even remember the last question you asked your favorite AI copilot?

The reason may be simple. Following five decades of research on memory, neuroscientist Eric Kandel, describes this beautifully:

The brain’s memory system works something like a pen and notebook. For a brief time before the ink dries, it’s possible to smudge what’s written. But after the memory is consolidated, it changes very little.

Skipping to the answer is surface level learning.

Memory is also a feeling; it is tactile.

Like many others, I’ve played around with AI generated headshots, out of curiosity more than anything else. How much further from the truth is it, I wonder, than a photo I take and photoshop, or put a filter over? They’ve come a long way and they’ll go much further. Recently, I was having fun playing around with the Apple Intelligence AI feature, turning myself into a painter in space. Cute, but likely a memory I’ll quickly forget.

This photo, in contrast, was taken by photographer Mindy Briar in New York. It was part of a photoshoot to get some professional looking photos for more public speaking I was starting to do. I got maybe 5 photos I really like out of 200. But more importantly I got a great memory.

I remember distinctly the moment I stepped out of my Uber in front of the steps at The Met and looked around for my photographer. We started taking photos outside the museum and, man, was I nervous. There were so many people out the front on a very warm NYC day, enjoying their lunch on the steps of The Met. We then took the photoshoot into Central Park and I thought for sure it would be more quiet and I could relax but no luck. Particularly around the famous Bethesda Fountain, there were crowds of people looking at us, wondering perhaps if I was someone important enough to have a photoshoot in Central Park and in truth that’s what I was second guessing at that moment. After about an hour, the sun rose higher in the sky, and I started to really feel the heat, not just metaphorically. I was squinting at the camera and sweating in my dress, my hair started to stick on my forehead and I was wondering why I had chosen an outdoor setting in summer in the first place. It was only at the very end when we called it a day, I put my flats on and we started walking out that I felt like myself. I stopped near the entrance where a man was selling his paintings of scenes of Central Park, and that print now sits opposite my desk at home. I see it each time I look up and away from my work, and think of how incredible it is that I got to live so close to this beautiful part of the world for a little while.

So many feelings are tied into my LinkedIn profile photo. I don’t care that it’s not perfect, or perfectly professional, each time I see it makes me think of that day in Central Park, of that moment in my career when I warranted a photoshoot, of how much has happened since that moment.

Introducing ... AI induced cognitive atrophy

It didn’t take much effort to recall those memories, and all the feelings that came with them. We know from studies that the more we rely on AI to go and find us that thing we want, or prompt something we’re thinking about but quite can’t place, the more it deteriorates our own memory.

A 2024 Frontiers white paper, published on NIH, flags the unknown ways this could impact our cognitive abilities and our memory. They call this potential decline of our abilities?AI induced "cognitive atrophy”. Specifically, the over-reliance on AI and chatbots to retrieve, problem solve, and go-do may weaken our brain’s ability to develop the skills we need for pretty important stuff, like critical thinking, analytical acumen and creativity. ?

“[It] draws parallels with the 'use it or lose it' brain development principle, positing that excessive dependence ... without concurrent cultivation of fundamental cognitive skills, may lead to underutilization and subsequent loss of cognitive abilities.”

The paper rightly flags what a concern that may be for young developing brains, which we know are already outsourcing the process of learning. But I also worry what this means more broadly for our future workforce, who the World Economic Forum has already highlighted will need more critical thinking and advanced analytical skills in a technological future.

Use or it or lose it also applies to employers, who wear this risk when prioritizing productivity gains from AI over upskilling and reskilling the human workforce.

In this year's Future of Jobs report, the WEF flags global macro trends driven by technology that are creating a massive shift in the labor force. “Employers expect 39% of key skills required in the job market will change by 2030,” the report says. When you see the list of required skills in the future, aside from the obvious one, you notice they all require that process of learning; of doing and experiencing things to figure out what works and what can vs should be done.

Leadership skills on the list, including social influence and talent management, can really only be learnt on the job, applied by dealing with different personalities and team dynamics.

Then there is a very interesting bucket amongst this that I see often discounted by senior leaders in business:

“Creative thinking and resilience, flexibility and agility are also rising in importance, along with curiosity and lifelong learning.”

While organizations talk about programs that strengthen employee development and learning, that build resilience, adaptability and flexibility, and encourage creativity and curiosity, these programs are often under-resourced and de-prioritized in times of short term fire fighting. And yet, in a bitter irony, these are the skills that make us irreplaceable by AI, and uniquely able to generate truly innovative ideas for business and for ourselves. Forbes wrote about the various benefits of developing these soft skills a few months ago, citing recent studies by Gallup and others.

So it’s important that even if our organization doesn’t support or reward the building of these skills, that we do this ourselves, for ourselves.

Back to memory, unfortunately a decline is inevitable.

“For anyone over the age of 30 reading this article, here's some bad news for you: Your brain is already on the decline.”

Stanford released this upbeat statement recently. The average age of an American worker is 42.3. The workforce is getting older, according to the Bureau of Labor Statistics, and our younger generation is learning with AI.

According to research done by Harvard and the Center for Digital Thriving, half of American students aged 14-22 are using AI to study. Young kids see it not as a way to cheat, but as the modern way of learning. I wonder if it would be better the other way around.

The Center for Digital Thriving has a worthy mission of making sure children advance in a technological future but adopt habits fully conscious of their consequences. They state, “We can have more meaningful choice, intentionality, and control over how technology fits into our lives.”

The solution isn't to completely bury our heads in books. According to Dr Sharon Sha, Stanford Neurologist, an important way to build concentration and strengthen memory is to take mental breaks, particularly by going outside into nature. She says, “The thought behind this is that the constant stimulation of technology places?a?high demand on our attention (emails, text messages, etc.), but disconnecting and focusing on the sights and sounds of nature improves our attention and working memory.”

As it turns out, this is where we see a similarity with AI and the human brain.

Working memory, as it turns out, is similar to a ‘working set’ of data in computer science. In both cases we’re looking at how much ‘data’ a single processor can handle at any one time to solve problems.

Working memory or a 'working set' is essential for critical and complex decision making. Evolutionarily we have a limit. Even the smartest of us who are mathematicians and chess players have a ceiling for how much we can process. Theoretically, AI does not.

At best, AI problem solving can compliment human problem solving - an expansive AI model that can tackle larger data sets (and one day spit out fewer flaws), while the human brain approaches deeper problems that require critical thinking. That isn't just our advantage, it's our human agency.

Experts surveyed by Pew are split about the impact this may have on human agency; our ability to make decisions. They can't quite seem to agree on how the machines will be designed for human intervention. Even though we're the ones designing them.

Futurist Paul Saffo in the survey warned of an unnoticeable captivity AI grants us under the guide of agency.

“We have already turned the keys to nearly everything over to technology," says Saffo.

Memory isn't something we can extend to AI though, at least not yet, and not if we don't choose to.

Parking existential fears for a moment, we simply don't know what effect AI will have on our memory. For now, memory is a biological process. Even if our cognitive systems will adapt, we don't know in what way.

It's too early to tell.

The question isn't if it will happen but how. And aside from our memory, what are we at risk of losing?