[infurl]

https://bdtechtalks.com/2019/06/05/mit-ibm-hybrid-ai/

Hybrid AI combines both statistical and symbolic artificial intelligence algorithms into a unified solution. Such a program can handle messy real world data like a neural net can, but requiring orders of magnitude less training and data than the current crop of behemoths. Likewise it can handle complex reasoning problems without requiring a massive amount of human intelligence to assemble a plethora of rules.

https://openreview.net/forum?id=rJgMlhRctm

Researchers from IBM and MIT proved that their Neuro-Symbolic Concept Learner (NSCL) could achieve a score of 99.8 on the CLEVR battery of standard tests for visual problem solving. Put simply, it uses several different neural networks to encode the data symbolically and generate a classic logical program that does the actual problem solving.

Abstract: We propose the Neuro-Symbolic Concept Learner (NS-CL), a model that learns visual concepts, words, and semantic parsing of sentences without explicit supervision on any of them; instead, our model learns by simply looking at images and reading paired questions and answers. Our model builds an object-based scene representation and translates sentences into executable, symbolic programs. To bridge the learning of two modules, we use a neuro-symbolic reasoning module that executes these programs on the latent scene representation. Analogical to human concept learning, the perception module learns visual concepts based on the language description of the object being referred to. Meanwhile, the learned visual concepts facilitate learning new words and parsing new sentences. We use curriculum learning to guide the searching over the large compositional space of images and language. Extensive experiments demonstrate the accuracy and efficiency of our model on learning visual concepts, word representations, and semantic parsing of sentences. Further, our method allows easy generalization to new object attributes, compositions, language concepts, scenes and questions, and even new program domains. It also empowers applications including visual question answering and bidirectional image-text retrieval.

[Dee]

https://bdtechtalks.com/2019/06/05/mit-ibm-hybrid-ai/

Hybrid AI combines both statistical and symbolic artificial intelligence algorithms into a unified solution.

How about adding classic algorithms also? for the instincts of AI 

[infurl]

How about adding classic algorithms also? for the instincts of AI 

I think it's a given that all artificial intelligence projects rely very heavily on classic algorithms, not just numerical and combinatorial algorithms but symbol manipulation as well. I'm sure that if it was called for, NSCL could be equipped with the ability to implement any kind of algorithm.

https://www.wolframalpha.com/

If you have ever used Wolfram Alpha then you will have a sense of the capabilities of this sort of software. This particular "software as a service" has implemented tens of thousands of classic algorithms and connected them all together with a natural language front end which has the smarts to interpret all manner of utterances and match them with appropriate calculations.

On a sidenote, it's amusing to read comments by machine learning fanboys dismissing classic algorithms and not realizing that even the most pure of machine learning projects are still utterly dependent on deterministic software.