Neurosymbolic AI combines neural networks (the pattern-matching brain) with symbolic logic (the rule-following brain).
Deep learning is great at recognizing faces or translating text, but terrible at reasoning step-by-step. Symbolic AI is the opposite: rigid, logical, but inflexible. Neurosymbolic tries to get the best of both.
Some researchers think this is how we build AI that can actually think, not just guess. It might be the end of the ‘bigger is better’ era of just throwing more data at a model.