Are AI chatbots examples of general intelligence or cognitive automation?
As artificial intelligence (AI) chatbots like ChatGPT become famous, I can’t help but notice how a perpetual topic of debate about AI continues to reemerge. Are these chatbots examples of general intelligence or cognitive automation?
Defining artificial general intelligence vs. cognitive automation
Artificial general intelligence (AGI) refers to a hypothetical idea, which goes something like this. Someday, we’ll be able to build machines that can perform (if not outperform) anything and everything that people do. In other words, robots will be as smart as humans (if not smarter).
AGI is related to the dream (or, if you read science fiction, the nightmare) of the “technological singularity.” It’s a future in which AGI technologies possess a general intelligence that will surpass human intelligence. The implication of this future is that AGI will become a runaway technology that we won’t be able to control.
Losing control of AGI in this futuristic scenario could result in a terrifying vision popularized in science fiction, not unlike “Judgment Day” in The Terminator. After all, if machines become more intelligent that people, what reason would they have to keep us around? Wouldn’t they just terminate humankind?
In contrast, cognitive automation is a more practical idea. Cognitive automation refers to AI programs—for instance, machine-learning algorithms—that perform specific tasks. The purpose of this technology is simply to automate activities, including cognitive tasks, previously done by people. In other words, machines are designed to replace humans, especially at work.
Common examples of cognitive automation may include:
- AI-powered data analytics, or software that automatically collects, organizes, and interprets large sets of data, such as using customer information to model and predict patterns of consumer spending.
- Image processing (a.k.a. computer vision), or software that sees and recognizes objects, such as facial recognition technology for security and surveillance.
- Natural language processing, or software that reads and writes in plain language, such as chatbots that answer questions on websites and apps.
Let’s take that last example, which brings us back to our topic of debate. Are chatbots just another example of cognitive automation? Or are they more than that—are they novel forms of artificial general intelligence?
The debate on whether AI chatbots are general intelligence or cognitive automation
For instance, isn’t it true that AI chatbots like ChatGPT are incredibly flexible in terms of how much they can talk about? This technology seems to be able to do more than respond to task-specific inquiries. It almost appears to display a sort of general intelligence on multiple subjects.
So, aren’t AI chatbots examples of artificial general intelligence, as opposed to mere cognitive automation? In truth, that’s highly unlikely. Here’s why …
The GIGO problem
One problem has to do with the input and output of AI chatbots. The output of these technologies will only be only as intelligent as the information they receive for input. And unfortunately, sometimes the input is terrible, which means the output will be terrible too. I’m reminded of what computer scientists often say: garbage in, garbage out (GIGO).
The GIGO problem is why a technology like ChatGPT is so prone to spreading misinformation and disinformation, which are rampant all over the Internet, its main source of input.
The creativity problem
Even when the input is of good quality, it’s important to keep in mind that AI chatbots don’t really create original content from complete scratch. Rather, they recombine existing content created by people.
To put it slightly tongue in cheek, this AI-generated content is kind of like a remix of a reboot of a rerun of something that content creators made years ago.
AI’s future: small-scale cognitive automation
For these reasons, the future of chatbots and other forms of AI will most likely be about small-scale cognitive automation that can perform specialized work tasks, similar to what Microsoft Copilot can do. The future of AI probably won’t be about large-scale displays of AGI that can ostensibly do anything and everything.
Plus, if you’re a tech company, building small-scale automation makes a lot more economic sense. As the Financial Times reports:
The future will belong to smaller, specialist generative AI models that are cheaper to train, faster to run and serve a specific use case, says Yoav Shoham, co-founder of the Israeli start-up AI21 Labs. “The moat is not technology. It is the relationship with the consumer … I think it will be a ‘few-takes-the-most’ market, not a ‘winner-takes-all’ market.” Scores of other start-ups, which only provide a generic service and have no traction with customers, will fail.
Related posts
Technological innovation and the future of work: synopsis of ‘The Second Machine Age’
Are relationships with robots possible, or will we all be ‘Alone Together’?