I had some thoughts on GenAI:
We have Generative AI but we don’t have Cognitive AI. Generative AI creates new content using weights, probabilities and basic context (which is based on probability from training data) but the analysis and content generated isn’t novel or based on reasoning across data – it’s not information from data. For that you need Cognitive AI.
ChatGPT, with prompts, refined it to the following:
We have Generative AI, but we don’t yet have Cognitive AI.
Generative models create impressive content — fluent, convincing, and often surprising. But the surprise lies in form, not in origin. What looks novel is usually a high-probability remix of a low-probability combination. It’s simulation, not discovery.
There’s no awareness. No purpose. No internal thread of logic connecting “why” one idea builds upon another. It’s not invention — it’s interpolation.
Cognitive AI, if and when it arrives, will reason across data. It will generate content because of insight, not merely in the shape of it. It will model the world, not just mimic our words about it.
Until then, calling what these systems do “creative” or “novel” might be more wishful than accurate. We’re not watching ideas unfold — we’re watching echoes of them recombine.
It not only corrected some false assertions, it also helped make my thought more readable and less terse.
Leave a Reply