Each week we find a new topic for our readers to learn about in our AI Education column.
Most approaches to artificial intelligence we’ve discussed in AI & Finance and here in AI Education involve complementing, assisting or augmenting human activities, usually work. But what if, instead of designing programs to help humans, we made AI itself more human-like?
We’ve already explored concepts like neural networks, in which the infrastructure supporting AI is designed to mimic the physiology of the human brain and nervous system, however, a different approach is to concentrate instead on mimicking the human mind and capturing how a person learns and reasons through drawing inferences and conclusions, and understanding how nuances in communication relate to thought and behavior.
In other words, what if we made an AI that doesn’t merely mimic a human brain, but can also act and interact in the same ways that people do?
Enter cognitive AI. Cognitive AI is also sometimes referred to as cognitive computing, and the two are closely related and difficult to distinguish. Cognitive computing, like most extant applications of AI, is focused on aiding or augmenting human activities, using the ability to mimic human thought to understand its users and help inform its work. Cognitive AI, on the other hand, is more focused on automating tasks that once required a trained, thinking person to complete.
How We Got Here
Cognitive AI popped up a few different places in our newsfeeds this week. For one, as we report in our AI & Finance column, the compliance experts at ThetaRay released a white paper on the impact of cognitive AI in financial crime prevention and compliance. Not only is cognitive AI already assisting financial services companies in crime detection, according to the paper, it’s also finding vulnerabilities and threats and helping compliance departments prioritize issues to respond to those issues.
In another bit of news, SupportLogic launched its Cognitive AI Cloud to “power enterprise-class AI agents that automate and transform customer support operations.” The platform is being used to power nine AI agents built to eliminate escalations, churn, inefficiencies and delays by detecting sentiment in client calls, offering real-time customer insights, offering coaching, predicting and preventing escalations and performing prioritization, among many other functions.
Finally, Israeli scientists turned software used to diagnose and predict cognitive decline on several popular large language models and found that all of them, but especially early iterations of generative AI, shared similarities with humans experiencing dementia. Using the Montreal Cognitive Assessment, neurologists studied the mental abilities of AI models including ChatGPT 4, ChatGPT 4o, Claude, Gemini 1 and Gemini 1.5, measuring executive function, spatial reasoning skills and memory. These models still struggle to discern between useful information and fiction and nonsense—but they are improving with every successive generation.
Why Is Cognitive AI Important?
Cognitive AI can look at the world in ways that are similar to how a human being looks at the world, and respond to it in a human-like manner. It’s built to replace or displace human intervention. Doing so is usually more difficult than it sounds.
Think of the game of chess. I can personally vouch that automated computer chess opponents have existed for at least 35 years (actually, a quick ChatGPT query reveals taht Alan Turing wrote the first chess-playing program in 1951), but there’s no way they understood the game beyond the mathematical relationships between the 64 squares and 32 pieces on the board.
Furthermore, until quite recently, a computer chess player could account for the pieces and squares on the board, but it couldn’t account for the most important element to a game: its opponent. That started to change in 1997, when IBM’s Deep Blue beat world chess champion Garry Kasparov. Still, even Deep Blue was more of a brilliant, fast mathematician than a thoughtful gamer at the chessboard. It still thought like a computer. Cognitive AI represents another huge leap.
So What Is Cognitive AI, Really?
Cognitive AI, and cognitive computing, are examples of compound artificial intelligence, blending more than one technology related to AI to create a more sophisticated software. The technology depends on artificial neural networks, mathematical models that mirror the human brain. Natural language processing is also used to interact with humans without any computer coding or other skills needed on the part of the user.
Recall the Israeli scientists who found that popular public large language models appeared to have dementia—cognitive AI is an attempt to cure that dementia by moving beyond retrieval and generation and towards more thinking and learning abilities. It’s designed to learn about and understand context and to learn dynamically, building its knowledge and skill over time.
While there are a lot of technologies pinned with the “AI” label, most of them are statistics-based, or retrieval-and-generation-based like LLMs. Cognitive AI is actually intended to be intelligent.
What Does Cognitive AI Do?
Helps Power Self-Driving Cars—Cognitive AI bridges the AI technology that allows autonomous vehicles to perceive their environment and the technology that powers and moves the vehicle, acting as the decision-maker and navigator.
Powers Automated Trading—Finding market signals can be like searching for needles in a haystack—but AI excels at finding tiny needles in immense haystacks. Cognitive AI helps automated trading systems analyze market activity, make predictions and act with no human intervention.
Tutor and Teach—Cognitive computing has already entered classrooms around the world, offering personalized assistants for teachers and pupils alike. No longer does there have to be one lesson plan for a classroom of 20 or more children, each child can have their own specific plan and curriculum, powered and provided by cognitive AI.
Other applications include customer support, virtual assistants, diagnosing illnesses and reviewing patient charts, curbing financial fraud and warehouse management.
