4 Predictions for AI in 2017

2016 saw great advancements in the artificial industry and machine learning industry, but 2017 is looking to bring even more about. We take a look at the top four predictions for AI that everyone should keep an eye out for:

Language

With the immense progress seen in terms of voice and image recognition by AI, one of the next big focuses is language. Researchers are honing on their vision on helping computers generate language with more ease and more efficiency. Full-blown conversations with machines has long been a fascination of humanity, but with all the complexities and subtleties of language, it’s going to be a herculean effort. While 2017 will surely see great advancements in this area, for the time being our Siri’s and Alexa’s won’t make for great debate club partners. However, it’s probably only a matter of time!

Neural Networks

The new buzz around machine-learning is centered around a technique called generative adversarial networks. The brainchild of Ian Goodfellow (currently a research scientist at OpenAI), GANs are systems made up of one network that generates new data after learning from a training set. It’s completed by another one that tries to tell the difference between factual and fake data. These dual networks, by working in tandem can produce realistic synthetic data. So, what can this be used for you ask? Well it could be used to de-blur pixelated footage or generate video-game scenery among other things. According to Yoshua Bengio, who was Goodfellow’s PhD advisor, said this new development is highly exciting due to the possibilities that it brings. This system that allows machines to learn and make sense from unlabelled data appears to be intrinsic in making machines far more intelligent in the upcoming years.

China and the AI focus

2017 may be the year when China becomes a major player in AI. It’s leading AI focused company, Baidu, has been focusing on artificial intelligence for quite some time now and is making advances in terms of natural language processing and voice recognition. China’s tech industry is no longer solely taking from or copying Western companies, having identified AI as the next big thing in innovation.

Baidu however, isn’t the only company with its eyes set on AI. Other companies causing a stir in the field are Tencent (mobile-first messaging and WeChat app) and Didi (ride-sharing) and the Chinese government has pledged to invest around £12 billion by the end of next year.

Reinforcement learning

Last year saw AlphaGo’s historic victory against one of the best Go players of all time (Lee Sedol) and this year will see AlphaGo take on 5 different players at the same time. The victory was a landmark particularly in the advancement of deep reinforcement learning. This technique is inspired from how animals learn that certain behaviours have a positive or a negative outcome.

By using this approach, machines can for example learn how to navigate a maze and through trial and error discern the positive outcome. Researchers and academics have been toying with this notion for decades now, but now thanks to the deep neural networks, the power to make it actually happen is there. In the case of AlphaGo, it figured out how to win the game at an expert level through relentless experimentation and analysis.

It’s hoped now that with the evolution of reinforcement learning, it’ll prove to be useful in many instances. 2017 will most likely see attempts to use reinforcement learning when dealing with issues in industrial robotics or automated driving.

Do you have any other predictions for AI? Let us know in the comments!

Are you looking for a new position in the Salesforce or Dynamics 365 markets? Click here to see our new roles.