I heard about Neuro-Symbolic AI as a good enhancement to neural networks, because they consume less power and more transparent (for example - quanta magazine - Vector-Driven AI). Kind of "best of both worlds" solution.
My question is - can symbolic models stand by themselves? Can they reach the same capabilities as neural networks or Neuro-Symbolic AI? I haven't found any evidence for that.
Asked
Active
Viewed 237 times
0
CuriosityChaser
- 1
- 1
1 Answers
1
Symbolic AI still has a future, as there are many real-world applications where having a 'black box' driving decision making is not acceptable/compliant. In many fields you need to be able to explain how and why a decision was reached, and in symbolic AI you can do that.
Maybe at some point in the future we will know more about explainability of neural models, but it would also require being able to fine-tune them in a reliable way if you found an error you wanted to correct.
There could be scope for hybrid systems where a neural model 'generates ideas', which a symbolic model then works with in an explainable and transparent way.
Oliver Mason
- 5,477
- 14
- 32