There are three fundamentally different forms of AI:
- Symbolic,
- Subsymbolic, often called “neural networks,” or “connectionist,” and
- Model-based methods. (These are neither symbolic nor subsymbolic; and include reinforcement learning, variational inference, and active inference.)
Most of the attention now is on transformer-based methods, which are a particular form of generative AI.
Generative AI comes in several forms, of which transformers are the most recent.
Now this is an interesting article.
Thank you, Kunz – and look for more!
You were an early responder, and the blogpost wasn’t even finished – although the intended content may have to go into the next blogpost and the one after … as AI/AGI keeps rolling forward!
Appreciated! – AJM
Looking forward to your next blog post. I think we are really close to combining the two (Symbolic, Subsymbolic) and your architecture also looks interesting.