Computer change | MIT Technology Review

Wing himself did not want to study computer science. In the mid-1970s, he joined MIT to do electrical engineering, inspired by his father, a professor in the field. Realizing that he was interested in computer science, he called her to ask if it was a passing fashion. Besides, the field did not even have textbooks. He assured her that it was not. The wings have been shown solely to give a sense of proportion.

Former vice president of Microsoft Research and now vice president of research at Columbia University, Wing is a leader in promoting data science in a number of areas.

Anil Ananthaswamy recently asked Wing about what he wants to promote “reliable AI,” one of the 10 difficult research was known for trying to make AI systems more fair and impartial.

Q: Can you say that there is a change going on in the calculation method?

A: Absolutely. Moore’s law took us a long way. We knew we were going to get to the roof of Moore’s Law, [so] Similar computing became popular. But the phase change was cloud computing. The original file-sharing system was like a computer computer, whereas your files were not local with your machine; was somewhere on the server. Cloud computing takes it very seriously, when data is not near you; The computer is right next to you.

The next change is about data. For a very long time, we designed the circuit, making things faster – processors, CPUs, GPUs, and other similar servers. We did not ignore the data section. Now we need to fix it on the data.

Q: That’s part of the science of data. How would you explain? What are the challenges of using data?

A: I have a very brief definition. Data science is the study of obtaining value from data.

You can’t just give me a pile of raw data and push a button and the price comes out. It starts with collecting, organizing, storing, monitoring, analyzing, and viewing data, and interpreting results. I call it data life. Everything to do around there is a lot of work.

Q: When using big data, anxiety often develops over privacy, security, fairness, and bias. How does one solve these problems, especially in AI?

A: I have these new research that I recommend. I call it a reliable AI, inspired by many years of progress that we have made on a reliable computer. By loyalty, we often mean security, reliability, availability, privacy, and usability. Over the past two decades, we have made significant progress. We have standardized methods that can ensure the accuracy of the code; we have security protocols that add security to a particular system. And we have some secret ideas that are legitimate.

Reliable AI raises the ante in two ways. Suddenly, we are talking about resilience and justice — meaning resilience if you disrupt what you are introducing, the results are not marred by too much. And we’re talking about interpretation. These are things that we didn’t like to talk about when we talked about computers.

Source link

Leave a Reply

Your email address will not be published.