It’s 1764, northwest England, 8 o’clock in the morning.

Yes, this is still Crash

Course: Futures of AI – just stay with me. You put on your linen shift and petticoat and start your day at the neighborhood cottagecore textile industry, using your trusty spinning wheel to turn wool into thread. But today, your spinning wheel has been replaced by a new, giant machine that can spin eight threads in the time it takes you to make one. It’s called the Spinning Jenny, and in one fell swoop, it’s revolutionized the textile industry…and stolen your job. –wait, rapid technological advances replacing jobs?

Where have I heard that before? Hi, I’m Kousha Navidar, and this is Crash

Course: Futures of AI. [THEME MUSIC] To help us understand exactly how it could transform our world, some researchers like to break down AI into levels. The first level is “narrowly transformative AI.” This is the kind of technology that could lead to “practically irreversible change,” but only in a specific area– like the Spinning Jenny did for textiles. But some AI models have the potential to change all the most important aspects of our society. We call that second level “transformative AI,” and it’s on par with inventions like electricity and the internal combustion engine (which, you know, transformed our entire world).

And then there’s another third level, a theoretical level that we’re building up to that could make even bigger changes – ones we can’t exactly imagine yet. “Radically transformative AI” could impact not just our day-to-day lives, but how we measure progress, find worth, and think about our place in the world. Most people think AI with human-level intelligence, known as Artificial General Intelligence or AGI, would be radically transformative. It’s hard to predict when AGI will hit the streets – or the cloud.

Some people think we’re really close, like a year or two away. But some others put that number closer to the year 2030, or even 2050. A small subset of people even think we’re centuries away from AGI– if we achieve it ever.

But even without AGI, we’re already in for some pretty big world-restructuring. One really clear way we’re seeing AI restructure our world already is in the economy. And to help us understand exactly how, the Industrial Revolution is probably the most helpful example.

See, the Industrial Revolution saw a lot of manual labor become automated, done by fast, accurate machines, like the Spinning Jenny, instead of slow, fallible humans (like me!). And automation turned out to be really good for the economy, mainly because it increased production, or the amount of stuff that could be made. That’s because normally, economies are limited by the amount of labor – or workers – they have available.

But the Spinning Jenny and other Industrial Revolution era technologies helped industries scoot around that bottleneck– and the era’s economy absolutely exploded. These days, thanks to AI, machines don’t just spin thread – they can design the clothes it gets made into, manage inventory at the store that sells them, run the algorithm that recommends them to you, and handle returns and exchanges when you call customer service. And it doesn’t end with the textile industry– this is happening across tons of different markets.

And, just like all those centuries ago, that automation could be really good for our economy. Consider this: our Gross World Product– the final market value of all the stuff created in the world’s economy in a given year – usually grows around 2-5% a year. But if AI keeps advancing the way it has been, some people predict that by 2100, our Gross World Product could be growing not by 5% but by as much as 30% every year.

It’s…kind of unimaginable. And even though many economists are skeptical of this number, it’s safe to say that this could potentially mean really big things for the economy. But just like back then, there’s a dirty, soot-blackened flip side to the automation coin.

What’s good for the economy overall isn’t always good for individual workers. The more jobs that get taken over by automation, the more human employees are pushed out of work. And that could change a lot more than how long it takes us to spin eight threads.

Narrowly transformative AI is already changing the workforce. Since the 2000s, automation has taken over something like 1.7 million jobs, from customer service to data entry to manufacturing to writing and editing. But not here on Crash Course, don’t worry.

Humans! And if AGI were to enter the chat, even more jobs, or even entire professions, would be cooked. AGI isn’t here yet, but theoretically, it would be able to do anything human cognition can.

And it might even be able to do that work better than we can, transforming not only the economy, but all of human society in a way we’re not ready for yet. But let’s see it wear a top hat as well as I can. Like, AGI could take over the entire workforce, restricting humans to narrow positions like AI oversight, which could be few and far between.

And that’s not even mentioning the environmental impacts of AGI. Even today’s AI uses huge quantities of energy and water, and pumps out huge quantities of greenhouse gases. Recent data suggests that the average AI data center consumes anywhere from 500,000 to several million gallons of water every day.

And if AGI becomes ubiquitous, that number will probably skyrocket. All of which is gonna exacerbate the pretty gnarly climate change we’re already seeing, impacting not just the natural world, but our human communities, too. If the radical transformation happens slowly enough, and governments start planning for it, like, now, we might have time to make it work.

Governments could introduce environmental standards that require energy usage to be transparent and push companies towards greener data centers. They’re not currently where  I sit in the U. S. in 2025,  but theoretically they could be.

Governments could adopt structures like job guarantees that would keep us from being swapped out by AI. Or they could take things even further and adopt Universal Basic Income to make sure people still have money even if our labor is replaced by machines, or implement dividend schemes, where humans would get small payouts from the profits AI makes. And that could be really transformative.

If AI takes humanity’s place as the beast of burden for our economy, there’s kind of no telling what we’d get up to. It could totally realign our values, our social structures, and the ways we spend our time. It could change us from cogs in the capitalist machine into …maybe something else.

I might finally have time to catch ‘em all! I’m at…one. But if the transition happens too fast, without the proper safety nets, AGI could leave the majority of humanity unemployed, broke, and, well, chat, cooked.

Even if robots don’t take over the whole workforce, these advances in AI could spell disaster for lots of regular people. One way this could happen is that the first people to develop AGI could get really powerful, really fast. That could mean a single company starts to amass tons and tons of wealth, making the wealth inequality in 2025 look like chump change.

And think about it on a governmental level: what if a single government could gain huge political, economic, and military advantages over all the other ones? The prospect of that huge advantage could lead to an AI arms race, where countries, corporations, and maybe even extremist organizations hustle to build that  super-powerful AGI model first. All that stuff could be pretty bad news for most of humanity.

It could concentrate wealth and influence in the hands of big companies like Microsoft or Google, billionaires like Elon Musk or Mark Zuckerburg, or greedy governments, who don’t always have the people’s best interest in mind. It could be used to support authoritarianism around the globe. And I’m going to level with you: that process is already starting.

Deepfakes are images, videos, or audio recordings that can make it look like people are saying or doing things they didn’t really say or do. They work like lots of other AI does, where neural networks train on tons and tons of data, until they can start to generate new data that looks a lot like it. Only, instead of chess games or novels, these neural networks train on faces, voices, and body language to learn how people look, sound, and move.

And they can get pretty good at imitating people. Like, say, US presidential candidates. In July of 2024, Musk himself shared a campaign video from democratic candidate Kamala Harris where she called herself a “deep state puppet” who’s hiding her “total incompetence.” Only, Harris never said that.

The video was a deepfake, made by AI models trained on Harris’ face and voice– and it was viewed over one hundred million times. Deepfakes are an example of how AI can be used to influence voters, spread disinformation, and throw elections to those willing to play dirty. And it’s not just political parties who are using AI to influence politics, but foreign nations, too.

Lots of the deepfake content from the 2024 US election seems to have come from Russia, a country that…might have some reasons to mess with the US’ democracy. Deepfakes’ actual impacts on voters aren’t totally clear yet, but there is evidence that they’ve already influenced voter behavior not only in the US but also in countries like Slovakia, Indonesia, and India. And things could only get worse as Deepfakes get more advanced, manipulating voters with disinformation, generating doubts around legitimate election results, or convincing people not to vote at all.

Messing with elections like that is a direct threat to democracy, and if AI can help people do it, then it could easily become a tool of authoritarianism – a pretty powerful one. And the more AI gets implemented in things like politics and economics, the more power it’ll have in those systems– and the more it will be able to enforce the values of those (potentially pretty evil) groups who leverage it. This hypothetical situation is known as value lock-in, and it could leave us stuck with systems that are inequitable, anti-democratic, and a threat to human rights– all supported by AI.

So that all sounds…scary. But it also kind of sounds like science fiction. While things like deepfakes are real and dangerous, radically transformative AI isn’t actually here yet, and we can’t be sure exactly how things will play out as we move into our AI future.

Things could go down the radical road, where AGI takes over our jobs and political systems, robbing us of what it means to be human– or even going rogue and throwing off its human overlords entirely. And, true, AI development has been increasing exponentially, blowing benchmarks and expert predictions out of the water faster than you can say “Alexa, how do you spell antidisestablishmentarianism?” All of which makes that robot takeover kind of scenario seem more and more likely by the day. But there are any number of other things that could happen, too, making our AI future nothing more than a twinkle in Wall-E’s melancholic robot eye.

For instance, governments could intervene, taking action now to put a leash on things before they get too out of hand. Or we could start to run out of the new, relevant, high-quality data that AI models need for  training, stalling progress. Or maybe AI development itself could end up slowing way down.

And, of course, there’s always the possibility  that AI just can’t do as good  a job as we think it can. So, the AI takeover could be fast and shocking– or maybe never happen at all. Or it could turn out the dolphins were behind everything all along.

Anything’s possible. But that doesn’t mean we should get complacent. A long time ago, the Spinning Jenny probably sounded like science fiction, too.

And if we look the other way for too long, we could end up trudging sadly home in our petticoats, choking on factory fumes, wondering if this is the end of life as we know it. But to really prepare for the possibilities of our AI future, we have to answer two really important questions – how powerful can AI become? And how quickly could it get there?

That’s next episode, here on Crash

Course: Futures of AI. Crash

Course: Futures of AI was produced in partnership with the Future of Life Institute. This episode was filmed at our studio in Indianapolis, Indiana, and was made with the help of all these nice people. If you want to help keep Crash Course free for everyone, forever, you can join our community on Patreon.