Having humans scout for defects during steel production has proved unworkable but ArcelorMittal, the world’s largest steelmaker, says it has proof that deep-learning neural networking technology — artificial intelligence (AI) software that learns on the fly — can do it with 93% accuracy.
It is a sign of the times.
Thanks to advances in AI and computational power, technological singularity — the moment predicted in 1993 by mathematics professor and sci-fi author Vernor Vinge when machines surpass human intelligence — may be near.
German computer scientist Jürgen Schmidhuber, co-author of the ArcelorMittal report and Scientific Director of the IDSIA Lab in Switzerland, says he believes we are “not so many years” away from being able to produce a $1,000 computer with the raw computational power of a single human and perhaps 50 years away from a single, cheap computer surpassing “all of mankind combined.”
While some fear the technological singularity will usher in an era of escalating unemployment and robots running the world, Schmidhuber and others — including big tech firms, venture capitalists and start-ups — see it as an opportunity to transform business and society for the better.
For example, the world spends over 10% of GDP on healthcare — over $6 trillion per year — much of it on medical diagnosis through expensive experts, says Schmidhuber, a scheduled speaker at DLD 2016 in Munich.
Help Cancer Detection
Partial automation of this sector could not only save billions of dollars, but also make expert diagnostics accessible to many who currently cannot afford it.
Business Sectors Being Disrupted by AI
AI-based services will play a growing role in automated diagnosis, support and advice in healthcare.
Financial And Legal Services
Advances in robotics and AI will likely disrupt 25 million workers within these sectors globally, generating up to 55% in productivity gains.
Virtually no manufacturing company will be unaffected — AI and robotics will impact corporate strategies and fundamentally change the way companies operate.
Autos And Transport
Self-driving cars will operate without human input.
Take-up is expected to encompass adaptive robots, autonomous navigation in fields, collaborative robots, computer vision, site-specific crop management, and Unmanned Aerial Vehicles.
Big Data and automation of trucks, trains and drills are set to transform the sector.
Aerospace and Defense
The U.S. military is expected to ask for up to $15 billion in its 2017 budget to experiment with AI and robotics.
The next three years will see growth in sales of robot companions/assistants/humanoids to perform typical everyday tasks in the home. The global personal robot market, including CareBots, could reach $17.4 billion by 2020.
Sources: Bank of America Merrill Lynch, WinterGreen, AUVSI, McKinsey, Strategy&
Schmidhuber’s team won international competitions by using advances in AI for cell-division, or mitosis, detection, which is important for cancer prognosis, but difficult even for trained experts. By rapidly reviewing millions of cells, the lab’s technology lets medical experts focus on the suspect ones, allowing the handling of far more cases.
AI is expected to transform not just medicine but the manufacturing of all sorts of products as well as tasks ranging from “strawberry plucking to driving cars,” says Schmidhuber.
Indeed, the rise of AI and intelligent machines is expected to usher in the next industrial revolution, according to a November 2015 Bank of America Merrill Lynch report [PDF].
Within 10 years AI and robotics are expected to create an estimated “annual creative disruption impact” of up to $33 trillion globally, including $8 trillion to $9 trillion of cost reductions across manufacturing and healthcare, $9 trillion in cuts in employment costs thanks to AI-enabled automation of knowledge work and $1.9 trillion in efficiency gains via autonomous cars and drones, according to the report.
Deep Learning Coming Of Age
The report predicts adoption of robots and AI could boost productivity by 30% in many industries, while cutting manufacturing labor costs by 18% to 33%.
These advances are possible through the work of AI researchers like Schmidhuber who has devoted decades to building a self-improving AI.
“As a teenager I said this is the last significant thing man can do,” says the German computer scientist. “I am still saying the same thing. The only difference is that more people are listening. Why? Because some of the simple methods we have developed on the way to this goal are now massively used by the world’s most valuable public companies.”
What’s changed is that Deep Learning, a type of AI which Schmidhuber helped pioneer, is coming of age.
Deep Learning involves training a computer to recognize complex and abstract patterns by feeding large amounts of data through networks of artificial neurons, and refining the way those networks respond to the input.
Some 20 years ago Schmidhuber’s team started working on an artificial neural network called “Long Short-Term Memory” that to a certain extent can “think,” “remember” and adapt, in a way that is reminiscent of the human brain. Limits to computing power held the technology back but that is no longer an issue and the approach has proven superior at recognizing spoken words or other audio, or classifying video-like information.
Available To A Billion Smartphone Users
“We had the first Very Deep Learning neural networks already in the early 1990s, much deeper than those of Ukrainian mathematician Alexey Grigoryevich Ivakhnenko, the great father of Deep Learning,” says Schmidhuber. “Later, when computers had become fast enough, my team’s Deep Learners became the first to win connected handwriting and object detection and image segmentation contests, and the first to achieve super-human visual classification results, winning nine international competitions in machine learning and pattern recognition (more than any other team).”
“Our Long Short-Term Memory recurrent neural networks have recently defined the state-of-the-art in handwriting recognition, speech recognition, natural language processing, machine translation and image-caption generation,” says Schmidhuber. “Google and other companies have made them available to over a billion smartphone users.”
Alumni from Schmidhuber’s lab strongly influenced successful start-ups, such as the UK’s DeepMind, which was bought by Google in 2014 for $600 million.
Google, which has since bought several other European AI start-ups, has also made an investment in the German Research Center for Artificial Intelligence (DFKI), which counts 16 other private companies, including Airbus, BMW, Intel and Microsoft as partners. As yet another testament to Europe’s leadership in the field, the U.S. search engine giant has also invested in Oxford University’s AI research efforts.
Moves from U.S. tech players are also advancing the commercialization of services based on deep learning AI.
Take X and add AI
Last summer IBM announced that it will upgrade its Watson supercomputer by combining different AI techniques, including deep learning.
In early November, Google announced that it is open sourcing the software engine that drives its deep learning services.
Facebook has open sourced designs for the custom-built hardware server that drives its deep learning work. And in December a group led by Tesla Motors founder Elon Musk and Y Combinator president Sam Altman unveiled a $1 billion non-profit organization called OpenAI that will be overseen by Ilya Sutskever, formerly one of Google’s top AI researchers. It vows to share all of its AI research and technology with the world.
Start-ups the world over are also expected to play an important role in moving things forward.
“As Kevin Kelly [founding executive editor of Wired Magazine] provocatively put it, ‘the business plans of the next 10,000 start-ups are easy to forecast: Take X and add AI,’” Shivon Zilis, a San Francisco-based investor at Bloomberg Beta, wrote in a recent blog post. “In many cases you don’t even need the X — machine intelligence will certainly transform existing industries, but will also likely create entirely new ones.”