On November 2, 3, and 4, I attended ODSC West in San Francisco. What is ODSC, you ask? It’s the Open Data Science Conference. With over 3500 attendees, it’s one of the largest applied data science conferences in the world, with focus areas on machinelearningdeeplearning , predictive analytics, natural language processing, data visualization, cognitive computing, data wrangling, AI research, and other data science-related topics. I realize that this could be seen as a pretty strange thing for someone in this industry to attend; I can see you scratching your head from here (yeah, you should probably turn off that camera on your laptop). But with the advent of inexpensive computing and the explosion of data that has now become possible, we have ushered in the era of AI—and that is a topic near and dear to my heart. AI is the New Electricity Sumit Gupta, a VP at IBM (formerly of Nvidia) was one of the speakers on the first day of the conference. After making the prediction that the money sent on cognitive IT will grow 3X in the next two years, he said that “AI is the new electricity.” What he meant was this: It was at the beginning of the 1900s when electricity became ubiquitous. It was very expensive to build the infrastructure to accommodate this new resource—but now it has become practically free. (Another friend of mine likes to point out that in the early 1900s when the prediction was that eventually only the rich would use candles for mood lighting, the prediction seemed ludicrous. Now look at the state of illumination—what do we use when we want to create a special atmosphere? Candles!) So it is the same is with AI. Not only will all emerging and existing companies use some form of AI in their product offerings, but society will become accustomed to it—in fact, we will come to expect it. In fact, I would argue that I already do! For example, I can’t stand it when Waze can’t seem to figure out that I prefer to take one route over another, and wish that it would figure out which parking lot I regularly use at work, and would realize that I would rather drive Highway 17 than the crazy mountain roads that it seems to think I would be better suited to take every day. As an example, Sumit talked about the systems set in place already using AI to make trailers for films and sports highlights reels. Using AI to identify the exciting bits is actually not all that difficult; train the system to listen to audience cheers or tense movie music (a relevant post: Is it Amper, or is it Music? , in which I wrote about AI creating film music!), and voilà, you have a sports highlight or movie trailer. How else can this technology be used? You name the doo-dad, and AI will become a part of it, just as electricity has become part of practically every other thing that is sold, nowadays. Big Head in Real Life Another talk was given by Lukas Biewald, CEO of Crowdflower. Although visually he reminded me of the character “Big Head” in the show Silicon Valley , the similarity ended there. Lukas was a fantastic presenter; funny, engaging, humble, and presented some very interesting ideas. I liked what he had to say about the future of work. He said: Artificial intelligence seems like it might … create jobs for artificial intelligence researchers and slowly displace all other kinds of knowledge work. And while this might be where we end up a century from now, the path to get there won’t quite look the way people think. We can see where we’re going from AI design patterns used at Google, Facebook and other companies investing heavily in artificial intelligence. In the most common design patterns, AI can actually increase demand for exactly the kind of work that it is automating. —Lukas Biewald (This text was taken from his blog , but he said similar things in discussion at ODSC.) The big fear nowadays is that AI will put everyone out of a job. What we’re not considering, though, is that AI will make more jobs that are different, not get rid of them altogether. Check out his blog post; it’s a bit of interesting reading and more than relevant. Lukas also talked about the importance of using “clean” training data to train your AI system, and how much more beneficial it is to use scrubbed data than it is to simply gather more and still-muddy data. It’s no big surprise that increasing training data improves accuracy more than using “better” algorithms, but cleaning the training data improves it even more—with the same effect as doubling the amount of training data. This is what Crowdflower, Lukas’ company, specializes in: training data, machine learning, and human-in-the-loop interfaces. Cool. The Point The main thing that I kept thinking about, through the entire conference, though, was this: AI is great for semiconductors. No matter what happens in the future, all of this vast amount of computing power will be done on silicon, and that is what Cadence does. No matter the industry, no matter how AI is used, we will still have to design chips, boards, and systems. Cadence is poised to be successful in the age of machine learning. There were more interesting talks at ODSC; stay tuned for more! —Meera![]()
↧