Each week we find a new topic for our readers to learn about in our AI Education column.
After spending some time diving into some of the specifics around artificial intelligence and AI applications, we’re continuously going to zoom back out—rewind the tape, so to speak—and tackle some fundamental artificial intelligence topics. This week, we’re going to discuss a non-technical manner in which artificial intelligence technology has been classified: Weak artificial intelligence, strong artificial intelligence, and, just to take a good look into a potential future for humanity and technology, what has been deemed as “super AI.”
We should note that last year, not long after taking over this column, we touched on this method of AI classification in our discussion of artificial general intelligence, and thus it makes sense for us to start there, and, with a discussion of strong AI, because they are usually one and the same: strong AI is artificial general intelligence in most discussions, and weak AI, also called narrow AI, encompasses all of the technology and applications of artificial intelligence up to artificial general intelligence and strong AI. Got that?
To be perfectly honest, until a few months ago the abbreviation “AGI” was for me a tax term, but now I use it more to describe artificial general intelligence. Hopefully, we’re all becoming more literate in artificial intelligence, because the genie isn’t going back in the bottle. Anyway, here’s what I wrote way back then bout AGI:
“Simply put, it is artificial intelligence with the ability to perform tasks within the limits of human capabilities… Artificial general intelligence would be able to not only think and perform with the cognitive abilities of a person, it would be able to multitask like a person, like a gigantic Swiss Army knife where every tool was a high-quality tool.”
That’s not a bad description.
How We Got Here
Keeping this part short and sweet, weak AI and strong AI are terms that tumble out of people’s mouths. Without definition. All. The. Time. They’re insider jargon words that serve to confuse everyone else because seemingly no one takes the time to go back to tell their readers or listeners or viewers what they actually mean, and sometimes they’re tossed around by people who don’t really know what they mean. Making matters worse, they’re used interchangeably with the terms “narrow AI” and “AGI,” and maybe they shouldn’t be. Adding even more to the confusion, a blurring of the lines between weak AI and strong AI may be occurring (we’ll get to that later).
What Is Weak AI?
If you’re using AI in any capacity, and you’re not privy to technology and information inaccessible to the rest of us, you’re using weak artificial intelligence. This is not a pejorative term. While strong AI would have a human capacity to think, weak AI is far from thoughtless, it just uses a particular slice of cognitive ability, or it’s just narrowly applied—in other words, it’s built and trained with specific cognitive tasks or use cases in mind.
So weak AI encompasses speech recognition software in voice assistants like Amazon’s Alexa and Apple’s Siri, but it also includes the self-driving technologies now being implemented in vehicles. It powers the image recognition and text completion tools available in today’s multimedia suites, but also tools that can translate speech to text or between two or more languages instantly. Weak AI is also providing recommendations and next-best actions to everyone, from financial advisors looking for conversations to have with their clients, to Amazon and Netflix users considering their next purchase or binge-watch.
Back to strong AI. Remember our description of AGI, or artificial general intelligence? As it turns out, that’s just one component of what could be considered strong AI. In other words, most everyone does use strong AI and AGI interchangeably in technology journalism, but perhaps we shouldn’t. Strong AI, as a term, should also include the conceptual AI that goes beyond AGI, both super AI and conscious, sentient machines.
Does Strong AI Exist?
As of right now, strong AI does not yet exist. AGI, for one thing, is software that has human cognitive capabilities, and that hasn’t been achieved. There is some debate, however, as to whether the learning capabilities of modern generative AI, particularly the transformer technology underpinning large language models, could become, in reasoning and language skill, akin to an artificial general intelligence over time. It’s possible that today’s LLMs are a necessary precursor to AGI—and a weak AI will give birth to a strong AI.
Until recently, our ability to build infrastructure—to stand up modern data centers and lay down the fiber connecting them—loomed as a potential throttle to the evolution of strong AI from today’s narrow applications. But new developments both in infrastructure and in AI models hold out the promise that the explosion in AI development can continue for some time, and that AGI is within our reach.
What’s Beyond AGI?
AGI is just coming out of the realm of science fiction and becoming more than a theoretical possibility—but nothing (yet) runs at the speed of our imagination. Our ability to conceive of technology still far outpaces our ability to build and implement it.
There’s more to being human than having human cognitive abilities, and thus, there’s more to strong AI than artificial general intelligence. If people manage to create AGI, then they’re going to start finding ways to introduce other elements of humanity to software and machines—the genuine ability to feel and emote, for example, or genuine self-awareness and consciousness. Such artificial humanity and genuine humanity could eventually converge.
There’s also super AI, or artificial super intelligence, which describes software that has superior intelligence to human beings. If we can create software that learns, that trains and teaches itself, it becomes possible that software exceeds us. We’re probably a long way from artificial super intelligence becoming a reality—or maybe not. That’s the great thing about technology: it advances in leaps and bounds, and the next big thing is always right around a corner, but it’s always a corner that few people see.
