The most popular and comprehensive Open Source ECM platform
Deep Reasoning: Hybrid Combinations with Neural Networks
Over the last few years AI has made stunning progress in visual object recognition and interpretation. Given typicaly hundreds to thousands of examples, AI algorithms are able to sort out and to recognize hidden patterns and relationships and often outperform tasks like recognition and diagnosis than humans trained to perform the same task.
But recognition isn’t the same as reasoning, it’s more like being able to deduce or solve a problem only because of close matches with patterns and similarities that were extracted from large amounts of historical data.
Carla Gomes, professor at Cornell, said that “you can teach a machine to recognize a dog by showing it 1,000 pictures of dogs, but scientific discovery is not like that. You are not going to have lots and lots of labeled data. And in general, the examples you have are not exactly what you are looking for, but then you reason about what you know scientifically about the domain, and you can infer new knowledge.”
Even solving simple questions of reasoning is very challenging for current AI. One approach to improve this is to combine AI techniques with more traditional computer algorithms and with domain knowledge. For example, game programming frequently uses physics engines that can assist in the graphics simulation, making the rendering more lifelike by adding gravity and kinetics.
Another idea is to combine different types of AI algorithms. Ben Dickson, software engineer and tech analyst, wrote for TechTalk that “the benefit of hybrid AI systems is that they can combine the strengths of neural networks and symbolic AI. Neural nets can find patterns in the messy information we collect from the real world, such as visual and audio data, large corpora of unstructured text, emails, chat logs, etc. And on their part, rule-based AI systems can perform symbol-manipulation operations on the extracted information.”
Julian Harris, technologist and product specialist, wrote that “many argue that while it may be conceivably possible to build practical reasoning systems by example (deep learning), sometimes itâ€™s vastly more effective to have some rules in there too. I think the real answer will be a mix, so expect to see a growing number of hybrid architectures.”
Leave a Reply