Machine Learning and Artificial Intelligence: everyday life goes 4.0?

Machine Learning, Deep Learning, Internet of Things, Industry 4.0: these are only some of the most popular terms that are entered into the everyday language during this year, and these words are used also by those who, apparently, are not real AI-insiders.

Since these terms belong to the more specialistic topic of the Artificial-Intelligence, their use is not always done in a proper manner; however, if words like "Machine Learning" or "Internet of Things" are used with greater frequency even in "non-scientifical conversations", the main cause is that the AI itself is now applied in a bunch of brand new fields.

This is the reason why these terms, once used only by engineers and researchers, are now a mainstream trend that is penetrating the most different business environments and medias like radios, TV shows or generalistic web magazines. It is no coincidence, for example, that about a month ago in Italy was presented by Carlo Calenda, Minister for Economic Development, a plan for Industry 4.0, the so-called Fourth Industrial Revolution, that involves the State intervention to prevent a possible digital divide between enterprises on the Italian territory and foreign markets.

But these new Artificial Intelligence technologies are not limited to the economic and industrial field: the AI is also implemented in applications and tools that we could use every day and that already are part of our lives.

Google RankBrain

Some principles of Machine Learning seem already implemented in the most popular and used search engine. The Google RankBrain is developed through the use of an Artificial Intelligence system; this allows at the same time to identify and understand the most complex queries and keywords connecting new meanings and providing more valuable results, and to evaluate and sort with more refined criterias the content that will be part of the SERP.

Obviously, it's hard to understand what there is "behind Google", but, reading the news leaked from Mountain View and the interviews and studies made by the search engine experts, it seems that, thanks to the Machine Learning technologies, Google is able to pick up a large amount of selected signals and use them as a source for learning new informations.

Sony Flow Machines

Since 1997 the Music Team of the Sony Computer Science Laboratory in Paris is focusing on the study of the informations derived from the many facets of interaction and on the malleable "descriptions" in music. Over the years the team studies have been awarded with prizes and accolades, not least Flow Machines, an Artificial Intelligence software whose purpose is to catch a style from a song (from the text to the melodies) and then use this informations as the "matter" to compose new tracks.

Thanks to this sofware, a couple of weeks ago, are been released "Daddy's Car" and "Mr. Shadow", two songs composed with the informations taken directly from a database called LSDB, which counts 13,000 tracks. The two songs, "Daddy's Car" and "Mr. Shadow", were produced by Benoît Carré: he has selected a style of music and has generated melodies and harmonies with the FlowComposer system. And, although this may seem "just the beginning", the first album of AI-pop is expected to be released during the 2017 .

The Italian Deep Learning in 2016

Meanwhile in Italy, something is happening. As we already seen, the Government is planning to launch a plan to increase the investments in Industry 4.0; however there are several companies in the Country that are making pretty interesting things. In September Add-for, a Turin-based company and one of the most interesting Italian reality in the fields related to Machine Learning and Artificial Intelligence Applied to Engineering, has participated as a speaker at "AI with the Best", one of the most important forum on the AI topic. And just few days ago Add-for has released a benchmark test between GPUs that has aroused the interest of many experts.

Add-for has choosen to this test two GPU that are famous also in the gamers world such as the GTX 1080 and the Titan X. The Add-for team has tested the performances of the GPUs with different Deep Learning algorithms and libraries such as Neon, TensorFlow and Café. These Deep Learning Benchmarks require specific conditions and need appropriate structures, tools and knowledges. But such a benchmark test requires also a large amount of money that not everyone can afford: for these reasons the consultation of this data is very interesting for professionals who want to get an idea of the performances of these products in order to evaluate a purchase.

The Add-for team is already planning to publish new Deep Learning benchmark tests; we suggest to check the updates on their blog and Benchmark pages periodically, in a way to be up to date with the most interesting technical studies on the Artificial Intelligence.