4 ms·
There's a whole research area that tries to combine some ideas from symbolic AI with neural architectures. IMHO, this might eventually become mainstream. For e
by nextos 2y ago
There's a whole research area that tries to combine some ideas from symbolic AI with neural architectures.
IMHO, this might eventually become mainstream. For example, see all the work that merges theorem provers with RL and NN.
- CyberDildonics 2y agoWhat are examples of "symbolic AI", "neural architecture" and a system that is "neurosymbolic" ?
- optimalsolver 2y agoAlmost everything that comes out of Deep Mind.
- CyberDildonics 2y agoWhat makes it symbolic?
- optimalsolver 2y agoThere's usually a DL part (the neural network), welded together with a classical AI algorithm, e.g. Monte Carlo Tree Search, or some kind of theorem prover.
- CyberDildonics 2y agoWhat's the symbolic part? What does 'DL' mean here?
- optimalsolver 2y agoThe symbolic part is the MCTS and/or theorem prover. DL is deep learning, i.e. the multilayered neural network. Neurosymbolic approaches form a significant branch of current machine learning. You can read up on all of it if you're actually curious and asking all this in good faith, but I suspect you're not.