Behind the buzzword: What is AGI?
Want to know what AGI is and how close we are to achieving it, if at all? Here's what you need to know about it with no fluff, only facts.
Jan 23, 2026 • 5 Minute Read
In this edition of Behind the Buzzword, we cover one of the biggest buzzwords of the last three years, and one that AI companies and the media love to throw around: AGI. Here's what you need to know about it, explained in a five-minute read.
Foundational concepts: What AI actually is
Before we launch into what AGI is, it’s important to understand what AI actually is. You might think this is obvious to 99% of the population, but surprisingly few people actually grasp this concept, even though it has a massive impact on their lives.
AI, or artificial intelligence, means something that can perform a task typically associated with human intelligence, even small tasks. Because of this, a lot of things qualify as AI, like:
Autocorrect in Microsoft Word
Your email spam filter
A fraud detection system in your banking app
The scripts that make enemies attack you in a game like Minecraft, or your neighbours visit you in The Sims
Your ebook reader that reads out text to speech
Netflix recommending your next show
All of these are AI!
These days, it’s not uncommon for people to say “I hate AI, and I’d never use it!” But chances are, they use AI every day—even multiple times a day—without even thinking about it. Usually they’re talking about Generative AI, which is only one kind of AI.
To really not use AI at all, you'd have to go off the digital grid: start using cash money, use snail mail, write and read everything on paper, and only watch free-to-air TV (if that). That would be the extent you’d need to go to in order to “never use AI.”
Now that we’ve talked about how broad AI is, let’s talk about how AGI fits into all of this.
The three different categories of AI: AGI vs ANI vs ASI
AGI stands for Artificial General Intelligence, and it’s a hypothetical type of AI that would match or surpass humans in practically all cognitive tasks. It can generalize its knowledge and solve things without having to be reprogrammed. It also goes by the name Strong AI.
In comparison, the AI we have today is called ANI, or Artificial Narrow Intelligence. It’s good at mimicking what humans do at one very specific task. For example, a Waymo car can drive like a human, but it couldn’t then go on and do your taxes. This is also known as Weak AI.
In the movies, you might have seen AI that can outperform humans at practically everything. This is called ASI, or an artificial superintelligence. You could call this Super AI.
Do we have AGI yet?
Many companies like to claim they’ve invented AGI, or redefine what AGI is, particularly makers of popular chatbots and agentic AI solutions. That’s why it’s a bit of a buzzword, even though the underlying concept is real. But despite us using ANI everywhere—and some of these being quite sophisticated—none of what is currently on the market qualifies as AGI.
The AI solutions we have now lack that combination of memory, problem solving, critical thinking, and learning that is distinctly human. ChatGPT, for instance, doesn’t really have a “mind” or “thoughts,” so to speak, but rather guesses the next word in a sentence from its training data.
Without the ability to reason, truly learn, or a working memory, these solutions fall short of true AGI. Our best general-purpose solution is what you’d call “Vox sine persona,” a voice without a person—there is no actual thinking going on beneath, just a very sophisticated word-guessing algorithm.
And while people overwhelmingly treat AI solutions like they are people, and even wanting to marry your chatbot is a thing, the underlying reality is that these aren't AGI.
How close are we to achieving AGI?
Despite the noise, we’re not even close. The domain that’s getting the most hype—Large Language Models—is limited by the architecture of the underlying solution, which isn’t really designed for AGI. What it is designed for is convincing humans that it is a person and faking having a personality, and this unfortunate misbelief can lead to unfortunate consequences like AI psychosis.
While AI companies are currently trying to “bolt on” features onto their models like OpenAI’s GPT or Anthropic’s Claude, this doesn’t change the limits of the underlying technology.
Put simply, imagine your current AI runs on the equivalent of a car engine, and AGI would run on the kind you’d use for a spaceship. Even if you were to put a car engine in a spaceship and claim you were going to space, it simply wouldn’t be up to the task. At best, you’d be able to turn some lights on inside and try to make people think that it’s working, but when you actually go to fly it, the gap would become obvious pretty quickly.
What are the pros and cons of making AGI?
There’s no shortage of projects and investments being pushed into AGI research, mainly because there’s a lot of perceived benefits in creating one. An AGI may be better able to advance science, technology, medicine, and healthcare, and solve global problems.
There’s also a lot of risks and concerns, up to and including the possible extinction of humanity. There are a lot of questions such as:
If an AI can do everything we can do, what is the point of humans?
Does it lead to mass unemployment, starting with white collar and then blue collar jobs?
Do the AGI spread and preserve the values of those who created it?
Do they have their own consciousness, and is it ethical to get them to do tasks?
With all these AI, do we then need to adopt a universal basic income?
And there are a lot of people asking these questions even with the ANI that we have now, even though it’s yet to reach the point of AGI.
Conclusion
And that’s AGI in a nutshell! Now when someone says that they’ve invented AGI or are getting close to doing it, you’ll have a firm understanding on what that actually means, and if they’re just throwing around a buzzword or not. On the flip side, we’ve also demystified what AI actually means, so you can be that annoying-yet-correct person that tells someone who claims they don’t use AI that they probably do (and you know, win all the friends.)
Further learning about AI
Want to learn even more about AI, including all the different kinds there are? Pluralsight offers a range of beginner, intermediate, and expert AI and ML courses, including dedicated courses on generative AI. Since you can sign up for a 10-day free trial with no commitments, it’s a great way to take some professionally authored courses with a set course structure. Why not check them out?
More Behind the Buzzword articles
Advance your tech skills today
Access courses on AI, cloud, data, security, and more—all led by industry experts.
