Industrial automation’s deep learning curve

11th November 2014

Jacqueline Regnier

0 0

Artificial intelligence (AI) has been a bottomless source of inspiration since the early days of science fiction. The critical acclaim of the most recent film on the topic - entitled Her - is proof that the subject is still particularly fascinating. Perhaps the strangest thing about this, supposedly futuristic, film is just how familiar it seems. We live among machines that are becoming increasingly intelligent – often able to see, speak or even imitate patterns of human thinking. This is called deep learning.

In a slightly more practical deep learning effort, Microsoft created a piece of software that instantly translates spoken English into spoken Mandarin Chinese, while also preserving the unique speech rhythm and voice. The system is still far from perfect, but while in older applications approximately one word in four was incorrect, the error rate is in the most recent generation is only one in seven or eight words. The software is learning. Deep learning.

Tech giants have been jumping on the deep learning band wagon like there’s no tomorrow (which there may not be if robots rule the world). In September 2013, Facebook launched an advanced AI effort to find meaning in its user’s posts. Google has now worked out how to transcribe house numbers captured in Street View photographs by recognising arbitrary multi-character text in photographs. Even the infamous Netflix has been experimenting with methods of improving recommendations by mimicking the way the human brain works.

In industry, the applications of deep learning are just as varied as in the consumer sector. For us at European Automation, deep learning has the potential become a particularly useful tool. If our system was able to use extensive customer data to predict what type of equipment our customers need to purchase and when, it would most definitely help us anticipate our client’s needs.

While it’s no Samantha – the intelligent OS from the movie Her – our new generation CRM system has resulted in obvious customer service improvements. It has helped us track purchasing history in order to partially automate our sales efforts and marketing strategy for instance. And this is only the beginning – we have a lot more pleasant surprises in store for our customers this year.

Deep learning refers to a set of algorithms that machines use to model high level abstractions in data. This allows computers to see objects and understand what they are, by using very similar processes to those used by the human brain. In a previous blog post, entitled the “big data Lebowski”, we touched on this. Deep learning refers to computers that are able to independently process the huge amounts of data they come into contact with and to extract patterns from this data, whether it’s images, videos, human speech, different languages or codes.

These abilities could be used for a variety of applications, from personalising social media services to identifying house numbers on Google Maps or – less important, but more entertaining - recognising cats in YouTube videos. By the way, this last application was actually perfected by Google back in 2012. Initially, it took 16,000 interconnected computer processors, which were then left to roam the internet freely in search of feline footage.