| Hello, and welcome to this week’s edition of the Future in Five Questions. This week I interviewed Vinesh Sukumar, head of artificial intelligence and machine learning product management at chipmaker Qualcomm. Vinesh and I discussed Qualcomm’s efforts to create chips that can do the intensive computing required for modern AI tools locally on users’ devices, as well as his self-described machine learning geekery, the relative appropriateness of edge vs. cloud computing, and how intuitively his mother has taken to using AI tools. An edited and condensed version of the conversation follows: What’s one underrated big idea? Personalization. Working with machine learning and AI, I believe that you should build a system that is personally connected to the user. This personal connection with the user could include emotional and environmental intelligence. At Qualcomm we put a lot of emphasis on this across our stack on the hardware side, which might be in a position to extract both structured and unstructured, labeled and unlabeled data, which is private to the user but present in the system itself. We use this as much as possible to guide the use of AI models and systems to be much more predictive and much more relatable for the user. That is what’s missing in most [AI] behavioral patterns today, and we believe that with our investments moving forward we can accomplish it. What’s a technology that you think is overhyped? There’s been a lot of emphasis on distributed processing, where you can extract information about the user [from their data] that is distributed across multiple machines and then get a personalized experience, which I mentioned before. But one of the biggest challenges has always been with the compute investment on [non-cloud, consumer-facing] edge devices and whether it’s sufficient to give you that genuine specialized inference. There was an evolution where a lot of people talked about “hybrid AI,” where you could do portions of AI inference on the edge and portions on the cloud. We think it’s a great concept, and we love it. From the Qualcomm perspective we want to make sure that we support this and get it done. But I don't think the industry at this point is fully ready to enable hybrid data. What book most shaped your conception of the future? I'm a machine learning geek, for lack of a better word. So one of my favorite books is “The Alignment Problem: Machine Learning and Human Values” by Brian Christian. It emphasizes the social impact of machine learning, and how if you're trying to put a massive number of AI systems into production, wrongful access to training data can lead to wrong conclusions, and bias or lack of transparency can create problems for users. From my perspective as a product guy, this makes me think outside the box for much of my work at Qualcomm. Because it's not about building a system, but having the entire, end-to-end pipeline cleaned up in a way that the human experience with the product is much more impactful and much more meaningful. It’s a good book for product guys like me who work on AI, where you can see these [social] “what if” scenarios and then make sure we invest in the infrastructure to push it in the right direction. What could government be doing regarding technology that it isn’t? The Biden administration’s AI executive order was a really nice reflection on standards for AI safety and security. It’s not my place to speak on what the government should do, but I think it was an investment in the right direction to start to give guidelines around the deployment of generative AI models. You don't want to be in a position where personal assistants are encouraging violence. It’s really up to vendors like Qualcomm to integrate it and take it seriously. What surprised you most this year? The impact of generative AI across all age groups. One morning my mother called me at 5:00 a.m., which to me obviously meant there's something really, drastically wrong. But she tells me she has a problem with some of the images in her phone. And I said, okay, this is probably not the right time to talk about it. But after some grilling she tells me she takes a lot of selfies, then uses an app where she's able to remove wrinkles in her face, and suddenly it’s stopped working. And she says, “aren’t you working on AI?” So it made me at least happy that she realized after 20 years that this is the space I’m working in. And I was quite happy as well that she was using generative AI tools, which I never expected a 65-year-old person to do.
| 
