How Artificial Intelligence will Drive Semiconductor Markets

IBM’s Watson and why it’s a vision of the future of silicon: It was exciting to get the call to fly out to New York for a deep dive with IBM’s Watson and some of the team who have brought it to life. I say ‘with Watson’ because it’s like no machine that I’ve ever encountered. That said, I must warn that until mid-2016, I had not done any active research in AI since the nineties. I was called out to have the perspective of an experienced skeptic looking at the space with fresh eyes. So while excited, I did have to check my skepticism at the gate, as it was bigger than FAA regulations allow. After all, the history of Artificial Intelligence (AI) delivering on the promise of its name is pretty poor. AI has seldom been intelligent and mostly artificial. There is good reason for this. AI, in its classical form, was oversold. It crashed and burned some time ago and every time it’s reappeared it’s been a disappointment. The original proponents of what’s now called GOFAI (Good Old Fashioned AI) promised too much and then failed to deliver miserably. GOFAI was a long-term vision with no map or road to get there (for more see, Artificial Intelligence: It’s not new … below). What I found at IBM was a more realistic focus on deliverables. Plus there’s a deep understanding of the limitations of the compute power that’s available and what needs to be developed to move the Semiconductor industry forward. A short history: AI was virtually dead until IBM won Jeopardy in 2012. Research had moved forward, taking the form of neural networks, machine learning, and deep learning. But it was a very low profile. Winning Jeopardy revived AI in the public’s eye, while making it the Next-Big-Thing. And like most NBT’s, AI quickly jumped to HYPE-er-space at light speed to rise back into the pantheon of buzzwords. Research funding rose rapidly and moonshot programs were launched to catch up. As the 2010’s came to a close, AI’s had lots of companies back working in the state of ‘beat-the-human-at-games.’ Deep learning has made great leaps in natural language processing and translation. And few companies were successfully monetizing it. Watson is a tool that began to help doctors bring the most comprehensive and advanced healthcare possible to their patients. It had done so well, that it was being deployed in Medical Centers around the world and some doctors were saying it would become the standard-of-care. By 2017, Watson was being applied on a much broader scale. IBM had partnered with Quest Diagnostics to take Watson to 70% of America’s Oncologists. They were working with Medtronic to help people with diabetes. And it didn’t stop with medicine. The takeaways: When you spend time walking through it, one can see many applications far beyond medicine. Applications will be anywhere professionals need to make decisions based on years of education and experience — especially in organizations that seek consistent levels of quality throughout their organizations. Watson can play a key role in mistake-proofing the most elite professionals, while raising the capability of entry-level people and for every expertise level in-between. It does this by distributing the knowledge base of what has been traditionally held by an elite few throughout an organization and across the world. Think of Watson as like taking the lean out of manufacturing and dropping it into knowledge work. To be continued. Stay tuned to VLSI’s app and web site, where semiconductor market updates appear daily in ChipChirps™ and VLSI Releases, with more for subscribers in other sections. Refreshed from The Chip Insider®.