AI Whiplash and the Future of Humanity

These days, everything I read about AI gives me a bad case of whiplash.

Educators are fed up with Ai, so much that they are going back to pen and paper assignments in class. At the same time, the government is investing untold sums of money so that AI can take over literal functions of the government, just one of many efforts to make it clear that AI is the future of work. College graduates are entering a workforce that is completely upended by AI, signaling that the ability to use AI is paramount to participating in today's economy.

In medicine, AI is predicted now to be one of the greatest innovations, with huge potential benefits in areas like radiology and diagnostics. Yet, no surprise, AI has limitations and cannot replace your doctor.

And then there's the whole question of whether or not AI is taking our jobs. Is it? Is it not? It depends on who you ask.

The Godfather of AI kind of predicts an AI bloodbath when it comes to AI replacing employees, and Dario Amodei, the CEO of Anthropic, warns that AI is going to replace most entry-level jobs. Yet a lot of what is actually happening in the workforce has nothing to do with AI, argues economists and Ezra Klein. It turns out there's quite a bit of evidence that AI isn't replacing jobs like many people claim.

Silicon Valley is talking about and preparing for a future with a permanent underclass and Dario Amodei has not been shy to warn us about the upcoming white collar bloodbath, courtesy of the very tool he's developing.

But if these things haven't happened yet, are they going to happen? And when? How do we prepare? Should we spend time worrying about any of this? And what is the impact on us as humans, even in the present?

Obviously, no one knows.

We don't know what AI will and won't do -- to our economy, our environment, ourselves. We actually don't know how it will affect employment or the economy beyond what it is doing at this very moment. This means preparing for it is pretty much impossible. Everything is conjecture.

What I can do is observe how AI is changing the work I do today and extrapolate from that experience. Because beyond the clickbait headlines designed to glue your eyeballs to a screen, there is reality.

I am considered a "knowledge worker." My job involves research, generating written products, gathering evidence, and drawing conclusions. It involves working with data, both data that already exists and data freshly gathered.

I interact with a computer for about 50% of my day and with humans for the other 50%. Human collaboration is used to make decisions, generate ideas, and think about how to approach our research. We meet with experts and discuss topics, learn about how a program is being implemented on the ground, hear about the unforeseen challenges generated by a particular policy or legislation. We gather new information and data that is not available anywhere else, and synthesize that information and data into reports and briefs and oral presentations that (hopefully) will be used to make important decisions.

We answer questions based on what we observe and learn, draw new conclusions, and challenge old assumptions. Conclusions are drawn not just based on existing ideas and frameworks, but also new ones that we sometimes create. 2

About a year or so ago, I became moderately concerned that my job could be threatened by AI. Couldn't a robust AI tool do most of my work? Even if the tools available today couldn't, then what about the tools of the future? In an attempt to get out ahead of this concern and maybe brace myself for what's to come, I decided to aggressively start using AI tools (primarily Claude, Perplexity, and Gemini) to really see what they could do.

The biggest limitation I've encountered is that AI cannot generate new data or information. It cannot go out and talk to someone working in a field and elicit their thoughts about something happening right now. AI cannot field a survey to generate a whole new data set. It can't take in context about things currently happening and combine that with other contexts. Even the best models have pretty significant limitations when it comes to knowledge work. Turns out, replacing a human brain with all its connections and a live human with all of its experiences isn't as feasible as the AI tech overlords want you to believe.

AI tools can be beneficial

That said, there are clearly efficiencies that come with using AI. I have found it immensely helpful for conducting background research, locating information, drafting emails, helping me get started with writing, or brainstorming questions that I should be asking someone. AI tools can't do these tasks for me with absolute perfection, but they can make these tasks easier and more comprehensive. AI can shave off quite a few hours in my work week that I'd otherwise spend trying to solve a problem or find information.

AI is a tool, and a good one. But it is still just a tool. It is a faster tool than previous tools. An improved tool, much like all the tools we've improved throughout human history. To me, it is similar to when Google and powerful internet search engines were released. Those too brought about concerns that certain jobs (research assistants, librarians) would disappear. And what we saw was that those jobs didn't disappear entirely but rather changed. Google's search engine didn't shrink the workforce, but it did change how people worked. People who didn't adapt to the new technology were left behind. Those who adapted were more efficient, more marketable.

I see the gain from AI as being much of the same. In using it, I've found ways it can make my job faster. It can do some low-level tasks that don't require knowledge building or a foundational understanding of what came before (drafting an email, finding legal citations, formatting text, helping me code software, proofing or translating text). It can, however, speed up my day, leaving room for more work, more thinking, more complex tasks that aren't suited well for AI. And the more I used AI, the more I found tasks that it simply could not do as well as I could do them, because while my brain and the brains of people I interact with can generate new information and new ideas, AI is stuck working with the ideas that humans have already put out there.

But what do we lose by using AI?

Because we absolutely do lose something by using AI, and I think it's more fundamental than the loss of the foundational knowledge of the Dewey Decimal System or ability to use a microfiche machine. It is a loss that, as a society, we need to talk about because it could be devastating for humanity if we don't address it.

When you use AI to do the heavy thinking and don't actively engage in it, I do think there's an undeniable loss in our ability to think critically, to do the very kind of thinking that (ironically) led to the development of these AI tools.

In a recent interview, Steph Ango (CEO of Obsidian) points out what is lost by relying on AI, how it erodes our ability to understand how something is built, which then damages our ability to make higher-level decisions, because those decisions are no longer backed by an understanding of the very foundation on which you're building. It is, he says, why at Obsidian they don't rely on AI for coding tasks. Because when you lose an understanding for something, everything built on that something going forward is shaky.

In my job, to make higher-level decisions, I have to understand the decisions and ideas that come up from the bottom. If I turn all those bottom tasks solely over to AI, I lose that foundation, and the longer I use AI, the more of that foundation I lose.

This struggle is taking place right now on college campuses, the very place where young adults go to learn how to think. I know personally that my undergraduate experience was transformative in this way; it was the first place I was truly forced to think differently, to go beyond a text book or conventional wisdom, to push back on my own perception of the world. A recent New Yorker article captures the wide-ranging perspective of college professors trying to teach the next generation in this moment. The anecdotes from professors in this article represent a huge range of this struggle - from the dire pessimism that AI is destroying critical thinking to the optimistic professors that are incorporating it into their lessons in the hopes that they can teach students to think critically with AI.

The Future

I often wonder what things will look like in 10 years. AI may continue to evolve exponentially or it may stall out and be nothing more than a much better version of Google. Will future me manage a team of AI bots instead of a team of humans? Will my child be unemployed because there won't be entry-level jobs in which she can learn and grow her own knowledge base, or will the workforce remain largely the same and simply adapt and incorporate this new tool into its existing structure?

To some extent, the future is happening now. Entry-level jobs are disappearing in some sectors, but the negative effects of that disappearance are already being felt by the companies making these decisions. In a world where entry-level jobs cease to exist, we don't pass on knowledge to the next generation. We don't teach the next generation how to do something. The implications of this are frightening, because then there will be no one to manage the AI bots when I'm gone.

As a parent, it is terrifying to try and prepare a child for this future. While on one hand my experience today tells me that AI can't do everything, the reality is that it can and will do some things, and some fields will be drastically affected. But which fields? I used to think all areas of medicine were safe from AI, and then someone told me that orthodontics is being fundamentally altered by AI tools (if your kid has been treated with Invisalign you know what I'm talking about). I used to think doctors were irreplaceable, but a lot of medicine is about following existing guidelines and steps, something AI is actually pretty good at. I thought that there was no way AI could replace educators, but then I read about a school that uses AI to drive all of its education. YIKES.

As a human, I worry about what we’re losing by becoming more efficient. Undeniably, we’re losing our ability to think critically, to come up with new ideas, to do the things that make us human.

I think it comes down to what we, as a society, want. Not the tech CEOs who are making millions by the day with their constant release of frontier models, but us as a society. Do we want our children taught by AI or humans? Do we want our most intimate medical decisions made by AI or humans? Do we want to maintain our knowledge base or turn it over entirely to machines?

These are the questions we need to be asking now. Because I think the vast majority of people don't want their kids taught by AI, their medical care delivered by AI, their knowledge foundation eroded by AI. This will require fighting back against immense forces - AI tech billionaires who are not exactly altruistic, politicians who are receiving huge sums of money from those tech billionaires. I don't know if we can win that fight, but I think it's certainly worth realizing that it's the fight we're in and engaging.