Issue link: https://iconnect007.uberflip.com/i/1243344
MAY 2020 I SMT007 MAGAZINE 57 This is what we are doing at Manncorp. Hopefully, by hearing about our methods, other companies can develop their own solu- tions for remotely helping their customers. We urge companies in a position like ours to implement virtual or remote servicing meth- ods in an effort to meet government guide- lines. When this crisis has passed, businesses will open again, and it will be due to those who have done the vital work to keep things going in the meantime. SMT007 Editor's note: Written by Emmalee Gagnon with technical expertise from Chris Ellis. Chris Ellis is a sales manager/ engineer for Manncorp Inc. To read past columns or contact Ellis, click here. and we were able to conduct these services remotely. The results were so positive that the university has decided to order a second pro- duction line. Serving companies and organizations that are critical to our national infrastructure is crucial during this difficult time, and many customers are involved in government and military work. Most recently, we helped a con- tractor in Minnesota working for Naval Surface Warfare Center IHEODTD—a key component of the Defense Industrial Base. When working with military contractors, personnel are often unable to enter facilities, and the organizations require remote contact alternatives. Working with companies housing sensitive information or hazardous products is one way that remote service will continue to be necessitated even after the stay-at-home order is lifted. Driven by an innate curiosity, children pick up new skills as they explore the world and learn from their expe- riences. Computers, by contrast, often get stuck when thrown into new environments. To get around this, engineers have tried encoding simple forms of curiosity into their algorithms with the hope that an agent pushed to explore will learn about its environment more effectively. An agent with a child's curiosity might go from learning to pick up, manipulate, and throw objects to understanding the pull of gravity— a realization that could dramatically accelerate its ability to learn many other things. In recent years, the design of deep neural net- works—algorithms that search for solutions by adjusting numeric parameters—has been automated with software like Google's AutoML and auto-sklearn in Python. That's made it eas- ier for non-experts to develop AI applications. But while deep nets excel at specific tasks, they have trouble generalizing to new situations. Algo- rithms expressed in code in a high-level program- ming language, by contrast, have the capacity to transfer knowledge across different tasks and environments. MIT researchers created a "meta-learning" algorithm that generated 52,000 exploration algorithms. They found that the top two were entirely new—seemingly too obvious or counterintuitive for a human to have proposed. Both algorithms generated exploration behavior that substantially improved learning in a range of simulated tasks, from navigating a two-dimensional grid based on images to make a robotic ant walk. Because the meta- learning process generates high-level computer code as output, both algorithms can be dissected to peer inside their decision-making processes. (Source: MIT) Automating the Search for Entirely New 'Curiosity' Algorithms