Every Internet company has opened artificial intelligence labs, acquired start-ups in the field and hired the best scientists in the world. But what will come of it all, no one knows.
What lay people still think of as science fiction is already reality for the experts. While we might not yet have a personal robot that can think, artificial intelligence (AI) has already arrived in one form or another in our smartphones, in image processing or in driver assistance systems.
It can recognize contexts, evaluate situations independently and, just like people do, it constantly learns on its own from through experience. The latter in particular has given fresh impetus to artificial intelligence in the form of deep learning. The time has only now arrived for what is actually an old method. Thanks to huge volumes of digital data and enormous computing power, AI recognizes images, understands speech and solves problems that are not prestructured in a specific form.
To do so, the accompanying mathematical model of artificial neural networks follows the pattern of the nerve cells in our brain. When learning, new connections are often formed between the individual neurons while old ones are strengthened, weakened or even completely deleted.
IBM Research scientists Murray Campbell, Gerry Tesauro, and Eric Brown discuss “Extending Game-Based AI Research into the Wild.”
“Augmenting Human Intelligence AI and the Future of Cognitive Computing”
Yoshua Bengio, Professor, Department of Computer Science and Operations Research, Universite de Montreal,
John E. Kelly, Senior Vice President, Solutions Portfolio and Research, IBM.
Clever even without reason
Deep learning is sometimes compared to machine learning or machine learning with neural networks. Machine learning, however, is the generic term, and neural networks or, more recently, so-called support vector machines, are mathematical models. Deep learning is the term used when machines continue to learn without end. Like Google’s speech recognition.
That doesn’t actually happen on our smartphones, of course, but at a U.S. data center. The computing power of the smartphone would be way too little. The computer program has been “listening” to spoken questions from the entire world for years and attempts to draw conclusions from the vast amount of data. So it gets better with every call. No one knows exactly how the program works.
And that’s exactly where the problem lies. It can be tolerated for speech recognition and many other applications. But for safety-related uses, it’s too dangerous to have a black box, even if it works beautifully. For instance, there is no line of code that would be responsible for approaching a passerby.
Even Google is concerned about what it would do in the event of an out-of-control AI. It would simply be necessary for someone to activate the “big red button” from time to time.
5x artificial intelligence in numbers
From investor to technology decision-maker, everyone is interested in the growth prospects of artificial intelligence. But when looking at the figures, it’s important to not forget that the term itself is “blurry.” And not everybody means exactly the same thing.
- The research firm Markets and Markets estimates that the AI market will grow from $420 million in 2014 to $5.05 billion in 2020.
- Tractica expects around $40 billion to be spent on deep learning software between 2015 and 2024. The market will grow from $109 million in 2015 to $10.4 billion in 2024.
- Gartner foresees 6 billion devices such as cars or wearables that will be supported by artificial intelligence by 2018. Gartner also expects 85% of all customer interaction to be conducted without human contact by the end of 2020. In addition, almost 50% of the fastest growing companies in the world will “employ” more intelligent machines and virtual assistants than humans by 2018.
- Entering “artificial intelligence” in AngelList, returns more than 1,000 start-ups with an average valuation of $5 million, which corresponds to a total investment volume of some $5 billion.
- Forrester Research expects artificial intelligence to replace up to 16% of all jobs in the U.S. in the next ten years. New jobs will be created to operate AI platforms and manufacture corresponding devices. The research company thus sees an ultimate job loss of just 7% due to AI.