Many persons believe of artificial intelligence along logic line will take the road to ruin, they adhere to neuro-network is main way to artificial intelligence.
Artificial intelligence includes three functions: sensation, recognition, thinking and reasoning. As one of recognition methods, the neuro-network has ablity to solve the problem about thinking and reasoning in artificial intelligence?
Let us see it positively as that which allows people to improve their performance over time. Artificial Intelligence (AI) is the property of computerized machines and systems to perform the intellectual functions independently to solve problems, and draw conclusions for making decisions.
https://arxiv.org/pdf/1804.01396.pdf
We should never forget that AI is a set of tools to solve decision making problems. Although neural networks have the capability of processing high dimensional data to perceive, recognize patterns, computing policies and so on, they have shown many limitations like bias, low experimental reproducibility (which means a lot) and fully depends in the specification of the function to optimize. Also, NNs have poor perfomance on combinatorial problems that require to plan ahead, and good luck if something fails testing the black box. On the other hand, people use logic in their daily lives and is a field that have been studied since the ancient greeks which is a good indicator to build an AI on top of it. By the way, there is no AI if it cannot plan for the future, and logic have been the best option for planning so far with computers that are also logical architectures :).
This is the typical discussion between planning and learning communities, and the near future looks that both should be considered together to build a stronger AI.
Thank you to Krishnan for the previous paper reference.
Hi Yaozhi Jiang ,
Do consider limitations of digital computer based systems:
See:
https://researchcommons.waikato.ac.nz/handle/10289/1143
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking