Joe Curley is Senior Director HPC Platforms & Ecosystem Enabling at Intel, and he opened the agenda of Intel’s AI Day in Munich recently with his keynote on “Powering the Experiences of Tomorrow with Intel AI”, introducing Intel’s strategy for AI as he pronounced them “more than just a CPU company” in the age of artificial intelligence. AI Business reported on the event, and at lunchtime got the chance to speak to Joe to hear more about his thoughts on AI in the enterprise, Intel’s specific AI focuses and what is coming up in 2017 and beyond.
“Massive – but also massively unpredictable”
This was Joe’s initial response when we asked him what the impact of AI will be on the business world in the next three to five years. He expanded on this point, citing a number of key industries such as energy, finance, healthcare and pharmaceuticals which will be transformed by the implementation of AI. Joe imagines that healthcare – in which Intel are heavily involved – could move from the realms of science fiction to “near-science”, for instance by “using AI simulation to classify cancer type from sparse data or saving large amounts of time in understanding the molecular assemblies that lead to cancers.”
Joe also believes energy and oil will undergo a paradigm shift: “currently seismic scans require massive amounts of data taken from a seismic survey. So, surveys have to be carried out hundreds of times to get projections – imagine if you could do have AI learn from the on the fly as the data is coming in. Finding the oil first is the key to winning the ball game; coming second isn’t much use.”
When the Intel AI Day audience were asked to raise their hands if they had any experience with AI, just over an eighth of the room’s hands went up. Joe draws upon this survey, explaining that he sees AI technology still in its nascent stage – but not for much longer. “Algorithms that we have understood for a long time are becoming practical and being reduced into science, and becoming available and becoming optimised and can plug into infrastructure; that’s just happening now and it’s happening everywhere.”
So how long does Joe think it will be till everyone’s hands in the room go up to declare their involvement in AI? “The interesting thing is that when everyone’s hand goes up it will probably be the machine that’s answering for you,” Joe says. “When technologies become pervasive they begin by solving very specific problems, but then start to enter other platforms as they are optimised and become more cost-effective. Artificial intelligence will become pervasive, and the vast majority of people won’t even know that it is there”. While he couldn’t put a date on AI becoming pervasive, Joe pointed out the projections of many self-driving car manufacturers, who are setting their sights on the release of self-driving cars at the turn of the decade, as a telling sign that revolutionary possibilities are just around the corner.
Collaborative steps forward
But the path to success for such bold projections is not necessarily a smooth one. Joe cites “availability and uniqueness” of the technology as two key factors required for increased implementation of AI. So what steps are Intel taking to make AI-powered computing and an AI-enabled business world a reality? Joe and his team at Intel believe in a collaborative approach, and therefore are working with open-source communities, academia as well as the enterprise. “It really does take an ecosystem”, he says, “and one thing we have learnt is that we all rise together.” He expands on this to explain their strategy within this ecosystem: “We are working on taking the early versions of these open-source frameworks such as Caffe and TensorFlow and making them run at high performance on general purpose infrastructure. Intel can engage the community and adopt it, and build it out from its initial experimental stage to larger scale”.
The fourth industrial revolution
We move on to discuss the implications of the AI progression on the workforce – how will employees be affected by and respond to the increased prevalence of AI systems? For his answer, Joe refers to Jevons paradox, which he defines in the context of IT: “As technologies create efficiencies that in theory reduce the need for a commodity, new uses actually create demand that actually increased the need for that commodity. Lately in IT people have looked at it in the terms of how cloud computing should reduce the need some servers, but in fact it’s grown as new efficiencies create new uses.” He adds that “by building better technology people can do more work in much more efficient and cost-effective way.” Joe concludes our conversation by looking to the past in order to look to the future: “there is always fear and uncertainty and doubt around new technology, but in past industrial revolutions we have seen this happen again and again and again.”
By approaching the fourth industrial revolution with their 360-degree approach to open-source communities, academia and the enterprise, it certainly looks like 2017 is going to be a big year for AI at Intel.
Feature image credit: https://newsroom.intel.com/press-kits/intel-headquarters-video-b-roll/