How does abductive reasoning relate to the current developments in artificial intelligence? It is possible to see traces of deductive and inductive reasoning in most recent artificial intelligence studies. On the other hand, some people think that developments are still far behind human-type intelligence. There are various reasons, such as the consciousness argument. Some people, on the other hand, think that consciousness is not part of intelligence. If it is, then only slightly related… This is a topic I enjoy talking about, but I will proceed to the main subject.
BTW, I am on the side of people who thinks artificial intelligence studies are merely advanced statistics.
People, like me, think in this way because the human mind has an amazing and bizarre feature: we can make inference using seemingly unrelated premises. Logical reasonings such as analogy (a type of inductive reasoning) and abduction. For example, if your friend feels bad after having dinner with you, you try to find (abduct) the cause:
- Sensitive stomach.
- It might be because your friend is currently sick.
- Bad food.
For instance, you can eliminate the third possibility by knowing that you do not feel bad. The second would be obvious to you if you know that your friend is sick already. The first one would be obvious depending on your background knowledge about your friend. We can abduct such cause-effect relations because we have previously experienced them (or heard of them). Hence, artificial intelligence can only make such reasoning if they have a human-like life experience.
There could be possible ways to substitute such experience at a certain level. For example, in the Ex Machina movie, such experience is substituted by the search engine data. Still, we need a huge leap in our statistical models to make abductive reasoning.
A Computer Analogy
Analogy is from the following article: What is Abductive Reasoning? Examples of Abduction and Induction.
Difference between deductive, inductive, and abductive reasoning
The computer processors work with logic gates. You can see that how this relates to deductive reasoning. Deductive reasoning produces certain outcomes – definitive conclusions (Despite humans may fail at deductive reasoning). The processor is our computer working with the logic gates. By this means, processors are substitutes for human intelligence in certain ways. In fact, they are smarter in certain aspects. The clearest advantage is their computation capacity. However, it is not completely a substitute in this way.
With the statistics models, however, we make computers smarter. Especially in recent years, with the increased computation capacity, we can train complex statistical models such as artificial neural networks (ANN). ANNs have interesting uses, making the “artificial intelligence” studies popular in the last decade. Image recognition, disease diagnostics, self-driving cars, and so on. These models are beyond what a human mind is capable of.
Abductive reasoning is one of the things very difficult for computers to do. A large computer with complex statistical models cannot achieve something we can. We do not know how to model computers so that they can infer causality and associate seemingly unrelated things. However, still, we all appreciate the sophistication of the human mind.