The intended audience of this post is not the average Lesswrong reader; it is more for our family and friends that do not grasp the transformative power of AI in its current form. I am posting it here so that I can come back in a year and see how my predictions are doing. I hope that some of you also comment your predictions.
Over the last two years, I’ve been thinking a lot about how this new dawn in artificial intelligence (AI) could potentially affect things in society that we take for granted. I’ve consumed an enormous amount of information on the subject via books, podcasts, blog posts, conversations, and other forms of media in an attempt to grasp what lies ahead for us.
Unfortunately, I haven’t been able to find a good online community with a repository of ideas about potential near-term changes to society in this era of pre-Artificial General Intelligence (AGI). Most discussions seem to eventually drift toward what happens to the human race if or when AGI arrives. My feeling is that things will get too strange to even think about when that happens, so I’ve been trying to focus on the near future and how currently available AI technology—and the next few iterations—could change society.
The following paragraphs are my thoughts on the subject. I didn’t come up with any of these ideas entirely on my own; most of these thoughts were formulated through things I’ve read, heard, or discussed with others. Some of these predictions are already on their way, while others you might consider pure fiction. I’m using this post as a starting point to examine what we can expect over the next 2–10 years.
I’m going to start with the low-hanging fruit, as this is something most people reading this post have likely seen in their work. AI, in its current form, can significantly increase your productivity. If you’re not using a frontier model to help with your work, start using it now. If your work limits the use of AI, I suggest you experiment with it in your free time. Think of learning how to use AI as a self-directed educational course—just without the certification at the end. If you need a place to begin, start by reading Ethan Mollick’s Substack.
You might not realize it, but hundreds of companies are spending billions of dollars creating tailor-made versions of AI for your type of work. Are you in HR? There are companies developing AI assistants specifically for HR professionals. Sales, marketing, trading, accounting, management, consulting—you name it, and there’s someone racing to create the perfect AI assistant to boost productivity and streamline workflows.
Many people see this as a sign that white-collar work is on its way out. I disagree. In the pre-AGI era we find ourselves in, humans will still be crucial in white-collar professions, especially in roles that involve face-to-face interaction with clients. That said, we’re starting to see an explosion in productivity per worker, which leads to my prediction: In about two years, we’ll see the headcount of mature corporations stall and even begin to fall. Firms worldwide will feel the pressure to reduce staff to maintain competitiveness.
The good news is that this shift is happening at the same time a large cohort of the workforce is retiring, and there aren’t enough younger people to replace them. The bad news? These tools might become so effective that productivity per employee explodes—meaning one worker could be doing the work of 20. If that happens, we’ll see a significant shift in the type of work available, bringing its own set of challenges.
If you’re in university or have children in primary or secondary school, you might already be noticing the impact of AI on formal education. The change is fundamental, and educators are struggling to keep up.
In 2023, my wife and I lived in Tanzania for a month, where we spent many afternoons tutoring teenagers with their homework. In many cases, the language barrier made things challenging. By using ChatGPT 3.5 as an assistant, we were able to help the kids understand concepts we hadn’t revisited since high school—and in their mother tongue. That’s when I realized how incredible these tools are for learning.
Today, frontier AI models are far more advanced than the one I used in Tanzania. Parents are doing themselves a disservice by not leveraging these tools to advance their children’s education. At the same time, many companies are working on refining frontier models to make them safer and more effective for teaching children.
Here’s my prediction: within the next five years, families in large cities in developed countries will have the option to send their children to experimental schools where most instruction will be led by personalized AIs. Imagine this: children sit at desks, each with an AI-powered electronic device that provides tailored instruction. The AI would identify each child’s strengths and weaknesses in real time, allowing them to progress at their own pace. Parents and counselors would receive continuous feedback, while human monitors in the classroom ensure order and troubleshoot technical issues.
This kind of education wouldn’t replace all human interaction. Group projects, physical education, and arts would still involve socialization. However, traditional teaching methods will likely feel increasingly outdated as society recognizes the benefits of personalized, AI-driven learning.
By now, you’ve probably heard that cities like Austin, San Francisco, and Shanghai have started experimenting with self-driving cars that taxi people without a human driver. Although I haven’t experienced a Waymo ride yet, I plan to visit these cities this year and see what it’s like to be driven by a fully autonomous car.
The data coming out of these experiments is groundbreaking. Simply put, these cars are already driving better than the average human. This development has major implications for industries like taxis, deliveries, and even car insurance.
Once this technology scales, there will undoubtedly be pushback from society. However, market forces are strong, and the economic benefits of autonomous vehicles will be hard to ignore. Here’s my prediction: within the next 5–10 years, some regions in the U.S., China, and Europe will allow private ownership of fully autonomous cars that can also function as income-generating assets.
Imagine owning a Tesla with an Uber account. Your car drives you to work in the morning, then spends the day autonomously giving rides to other people. After picking you up in the evening, it continues working while you’re home. Over time, cities might even introduce fees for human drivers, arguing that self-driving cars reduce traffic and accidents, making urban mobility safer and more efficient.
Scaling this technology globally will take time, given the high costs and societal resistance, but the wheels are already in motion.
With the advent of AI models like openAI o3 and others that perform at near-PhD levels in certain fields, we’re on the brink of a new era in scientific discovery. Long-standing challenges in science and engineering are likely to be tackled more effectively than ever before.
Starting in the next six months, I predict we’ll see a noticeable uptick in breakthroughs. These might initially emerge in industries with lower regulatory barriers, such as telecommunications, energy production, aerospace, and large-scale infrastructure. The application of these discoveries will likely follow more slowly in heavily regulated industries like healthcare.
For many, this will be a thrilling time. However, the lag between scientific discovery and practical application means we’ll need patience to see these breakthroughs reshape our everyday lives.
This is the most unsettling section but also one of the most pressing.
AI is already transforming warfare. For example, drones are changing the nature of conflict in Ukraine, where kamikaze drones are being deployed en masse. These drones, piloted remotely, are cheap, effective, and deadly. Now imagine adding autonomous driving technology to drone warfare. Suddenly, you have an AI system capable of commanding swarms of thousands of drones with minimal human input.
Policymakers are slowly beginning to grasp the importance of winning the AI military race. My prediction? In the second half of Trump's administration, it’s likely we’ll see the U.S. shift to a “wartime economy,” reallocating significant resources to develop and produce AI-powered weapons domestically or near-shore.
This isn’t a prediction I make lightly—I hope I’m wrong. But the race to develop AI-driven military technology feels inevitable. If AGI is truly on the horizon, we must ask ourselves: where should it be developed, and by whom? For me, the answer is clear—it should be aligned and developed in societies that value ethical oversight, such as those in the West.
This final section might feel a bit out there, but it’s worth considering.
The rules and systems we rely on to govern society—laws, political structures, economic models—are all human creations. They’ve evolved over centuries, tested and refined through experience. But they’re not immutable laws of nature; they work because we collectively agree to follow them. So, what happens when AI becomes smarter than any human who has ever lived?
I believe we’ll start seeing small-scale experiments in governance within the next 5–10 years. Perhaps it will begin with towns or autocratic states testing AI-driven decision-making systems. These could range from new economic models to AI-assisted policymaking. Imagine a local council where a human leader collaborates with an AI advisor that analyzes data, suggests policies, and even predicts societal outcomes with uncanny accuracy.
There’s also the possibility of communities voluntarily choosing governance models that integrate AI more deeply. For instance, an AI system could handle administrative tasks or mediate disputes based on fair and impartial algorithms.
Will democracies embrace this shift? Probably not right away. Democracies are built on human input, debate, and decision-making, and introducing AI into that mix could feel like a threat. But as global challenges grow more complex, the pressure to experiment with alternative governance systems might increase.
And here’s a bigger, more abstract question: could AI itself become an entity with specific rights or legal recognition? As AI models grow more advanced, this might not be as far-fetched as it seems. Whether we’re ready for it or not, governance as we know it could look very different in a decade.
AI is changing society faster than most of us realize, and we’re only just beginning to understand its potential. While AGI might still be years away, the pre-AGI era is already reshaping how we work, learn, and live.
Over the next few years, we’ll see significant shifts—some exciting, others unsettling. My hope is that by focusing on the near term, we can better prepare ourselves for these changes and navigate them thoughtfully.
The journey ahead will be anything but predictable, but one thing is certain: it’s going to be a wild ride.