Issue link: https://iconnect007.uberflip.com/i/1288481
SEPTEMBER 2020 I PCB007 MAGAZINE 19 have the right product characteristics, then you have a selling price and the difference is your margin. That's how to be profitable. Johnson: Even if you're not ready to increase your sophistication, there is still room to work on waste management on your manufactur- ing equipment roadmap. There's a profit there. Once the effluent portion of your manufactur- ing process has been optimized, then turn your attention to increasing your sophistication, in- cluding your technologies and capabilities. But everybody should ensure that they're paying attention to their effluent management and op- timizing that. Holden: You can build anywhere in the world you want because you don't make any waste. They're all connected, and some pillars may be stronger than others. Johnson: A technology roadmap is a blueprint for how you build each individual pillar and make your pillars work together. Holden: Most of these concepts are highlight- ed in key Harvard Business Review articles on competitive advantage, roadmap, and quality function deployment. Sometimes, you go to smart people and find that the answer may lie in another industry in terms of how they ap- proached it. Johnson: Very true. Thanks for your time, Happy. Holden: Thank you. PCB007 "We show that you can build abstraction into an AI sys- tem to perform ordinary visual reasoning tasks close to a human level," says the study's senior author Aude Oliva, a senior research scientist at MIT, co-director of the MIT Quest for Intelligence and MIT director of the MIT-IBM Wat- son AI Lab. "A model that can recognize abstract events will give more accurate, logical predictions and be more useful for decision-making." (Source: MIT News) Organizing the world into abstract categories does not come easily to computers, but in recent years, re- searchers have inched closer by training machine learn- ing models on words and images infused with structur- al information about the world and how objects, animals, and actions relate. In a new study at the European Con- ference on Computer Vision, researchers unveiled a hy- brid language-vision model that can compare and con- trast a set of dynamic events captured on video to tease out the high-level concepts connecting them. Their model did as well as or better than humans at two types of visual reasoning tasks, picking the video that conceptually best completes the set and the video that doesn't fit. Shown videos of a dog barking and a man howling beside his dog, for ex- ample, the model completed the set by pick- ing the crying baby from a set of five videos. Researchers replicated their results on two datasets for training AI systems in action recognition: MIT's Multi-Moments in Time and DeepMind's Kinetics. Toward a Machine Learning Model That Can Reason About Everyday Actions