Artificial General Intelligence (AGI) has certainly become a buzzword in AI circles in the past few years. But what does AGI really mean for the average person?

Well, imagine if you had a personal assistant who wouldn't be good at just one task, like setting reminders in your calendar, but would be skilled across a wide variety of tasks—from editing your photos to designing a marketing strategy from scratch, or even telling you what to cook tonight12. That's what AGI is about—a system that can think, learn, and adapt across all fields, just like humans3.

While today's AI tools can already do pretty impressive things (we're looking at you, Perplexity), they're still very much limited in doing what they've been trained for. They shine in specific areas—like playing chess, generating art, or browsing the web for you—but fall short of replicating true human intelligence and doing tasks outside of their reach45. Your photo editing app could be as powerful as it wants, you still wouldn't use it to help you with a cooking recipe, right? The goal of AGI is to go beyond narrow tasks and finally have an AI capable of solving the same range of complex problems as the human brain6.

In this article, we'll explore what AGI is, its potential, and the hurdles we need to overcome before reaching this new frontier. Let's get to it.

If you've been using a language model like ChatGPT, you've probably realized how powerful AI can be at answering questions or generating coherent text, to the point where you may sometimes wonder if the AI can actually think by itself. Hint: It can't1. These systems, which we usually call "narrow AI systems", have simply been trained on millions of data to do perfectly one very specific task, but don't genuinely understand or learn the way humans do2.

AGI, on the other hand, is trained to perform any task a human can handle, whether it's mathematical reasoning, creative writing, or even emotional interaction, all without having to be entirely reprogrammed each time3. Said differently, AGI learns by itself and performs a wide range of tasks that are unrelated to each other4. As you can see, the real differentiator between AI and AGI is what we call "unsupervised learning". Unlike narrow AI which needs structured data, AGI can learn from raw experiences and adapt to new situations without being supervised or needing reprogramming5.

The main reason why so many people are excited about AGI is its potential to revolutionize almost any industry. Today, AI tools are used in healthcare to assist doctors in diagnosing patients1. But tomorrow, with AGI, those tools could not only assist doctors, but autonomously review medical history, analyze symptoms, and even create treatment plans tailored to each individual, saving time and lives along the way2. Similarly, AGI could also offer solutions to complex tasks that humans can't solve alone, and help us find how to combat global warming more effectively by analyzing data and making informed decisions for us3.

However, and despite all their potential, machine learning models are still in their infancy when it comes to achieving true AGI. Right now, they're still limited to specific domains, requiring massive amounts of data and pre-defined rules.12 Building a machine with the ability to think and adapt like the human brain is no easy task, far from it. We've already made good progress with current Generative AI systems, but AGI still requires far more than what we can do today —like replicating the human capacity for creativity, ethical decision-making, and emotional intelligence as fluidly as humans do.23 Think about it: when we read an ambiguous sentence for example, our brains can rely on past experiences and reasoning to understand its true meaning. Teaching a machine to do that? Not so simple.4 Another roadblock with AGI is making sure that these models truly align with human values. Indeed, there's a growing concern that, if left unchecked, AGI could prioritize tasks in ways that aren't ethical, with systems optimizing healthcare solely for efficiency— making decisions to reduce costs but compromise patient care.56

As seen above, there's no doubt that Artificial General Intelligence could bring unimaginable advancements in industries like healthcare, education, and engineering. But as for every new technology, it's also essential to consider the risks and potential ethical issues it could raise.

What if AGI falls into the wrong hands, or worse, develops its own goals that conflict with ours? Creating AGI systems with human-like capabilities is not just about technological achievement; it's about making sure they reflect our values and work with us to improve our society. Making sure we build "fail-safes" into AGI and that we remain in control even as machines evolve.

AGI isn't just a technological leap but might be closer to a crossroads for humanity. On one hand, it offers the possibility to reshape most industries, reinvent how we live as a society, and even redefine what it means to work. On the other hand, it also brings a new set of challenges that we've never faced before. And as we're getting closer and closer to machines that can think and adapt like humans, it's not just about the "how" anymore, but also the "why"- Why are we building this? Who stands to benefit, and who might be left behind? As we enter a new era of technology, we have yet to find out if AGI could be a powerful tool for progress or a Pandora's box we're not ready to open.123