The general public said Vogels still has a perception that “cloud is a few things that sit out there, somewhere,” he said. “And, maybe in the early days, when we launched the first two regions of AWS, that was the case.” “But now, there’s 24 regions around the world, 84 availability zones, hundreds of points of presence, pushing things to the edge in 5G access points, up in space – you know, cloud’s getting closer and closer to every individual,” said Vogels, referring to the overall scope of AWS infrastructure. Cloud is literally around the corner, in the sense that when you’re walking and talking and texting and app’ing, there’s some AWS there. “If you have your 5G phone, I wouldn’t be surprised if you are accessing AWS services of your carrier – actually running in an AWS access point close to you. AWS’s Local Zones might best symbolize the notion of being closer and closer to each person, an option where compute, and storage and additional services are deployed close to large population centers. “Anyone with streaming or gaming who wants to have really low latency toward their customers, they don’t run those access points in the main regions of AWS,” said Vogels, “they run it in the Local Zones to get closer – and that has been a very successful product.” Cloud getting closer and closer is already showing up in interesting ways. Take embedded technology in the built surroundings, smart technology and sensors. “One of the things I am seeing is that the clear distinction between what is digital and what is physical, or what is analog, has disappeared,” said Vogels. “The whole notion of ambient placement of digital sensors and interacting with them is definitely on the rise.” “But if they can talk, this is something they have done for years.” Some older users describe Alexa as “their savior,” he has observed. There is an instrumental lesson in that, perhaps. In an ageing global population, says Vogels, the tension between who needs care and who can be helped may, to some extent, fall back on smart technology to make self-care an important component of healthcare. “I definitely see a variety of new innovations happening in the space of people doing self-care,” said Vogels. “And one of the areas of that which I think are extremely important is elderly care; we’re all getting older.” He said that one challenge for society is “what kinds of innovations, first of all, to make sure that people can stay home, longer,” in old age, rather than having to go to a facility. Via sensors and automatic alarms and such, “there’s lots of innovation, especially around elderly care and health care in general, we will see major improvements in convenience,” said Vogels. Also: Why AWS sent a Snowcone into orbit Asked what things he would have emphasized had he done the talk, Vogels told ZDNet that he would have liked to have told the audience about the bad old days of archival radiology storage. “In those days, 40, even 30 years ago, “the cost of digital storage was too much for hospitals to pay,” and so to meet medical records needs, “instead of being able to capture all the MRI data in digital form, they basically printed it on films, and archived like the way that they had always done with X-rays.” The result of the cumbersome archival systems of film in vaults, said Vogels, is that “when the patient came back, we wouldn’t be able to compare the digital information with each, you would actually look with film.” For the hard-working radiologist, poring over films, “their eyesight was not as good at night as it was a 9 o’clock in the morning,” he noted. In one of the many details of how technological change can shape practice, that greater access to the information in digital form starts to demand AI and ML as an enhancement to what the radiologist does, said Vogels. “There’s still plenty of people who believe that AI and machine learning will take jobs away,” observed Vogels. “But talking to radiologists, they are very excited about all this.” With digital, “those comparisons become quite microscopic,” said Vogels, meaning looking at something such as s tumor. “Your eyes are not good enough for that, but AI and machine learning can discover these kinds of patterns and, sort of, guide a radiologist into areas they should be looking deeper at. “And so, in essence, they [radiologists] see AI in healthcare as augmentation of themselves, and maybe a chance to get home earlier at night.” The radiology example is just one instance of how “There is a major shift happening in healthcare because of the ability of machine learning and because we can keep data around,” said Vogels. “Nowadays, they look to measure, say, growth of tumors at almost the microscopic level versus centimeter,” explained Vogels. “That’s a major shift.” He said that microscopic differences that can be seen for the first time with machine learning in tumors can be major differences. “There are certain patterns, especially with MRI scans, where you can already see at microscopic levels the differences between malignant and benign” tumors, said Vogels. The radiologist still makes the decision, he said, but ML can “surface information” that couldn’t be before. Other advances include the digitization that is underway of the world’s medical note-taking, most of which is still in hand-written form. Getting those records into digital data will allow for AI to sort and sift it all, “where you can have analysis of patterns and cross-sections, is now possible where that wasn’t possible at all even five years ago.” Vogels said that techniques for such as tumor analysis of MRIs will demand innovations in the most advanced forms of AI, such as deep learning. “It’s not just ploughing through data or just writing a SQL query and thinking you can get the result out of your data,” he said. “It’s really pattern matching and deep learning,” and “many of these models are still evolving,” meaning the AI models used. Machine learning will deliver benefits in regulated industries, such as energy, opined Vogels. In liquified propane gas industry in Australia, for example, they are moving to predictive maintenance, he said. Where in the past, the facilities had alarms on their giant refrigerators only if something went wrong, “They are using newer sensors they to do predictive maintenance, where, instead of running into disasters, they can prevent it. “That’s major.” Manufacturing is set to get a big lift from machine learning. “In the U.S., the typical manufacturing equipment is 26 years old,” he observed. “That’s before machine learning, and so modernizing the manufacturing environments, and using the data that comes from those, for safety and security, and preventive maintenance, there’s many different areas where this can be applied to.” And what of the deeper question of ethics in AI? “There’s a lot to think about,” said Vogels. Part of what can be thought about is for those building things to be, in a sense, self-supervising, he implied. “On the one hand, we should look at technology to be able to do to see what we can do to really get insight into what these models actually are – I even expect regulatory requirements will come down that way.” At the same time, “ethics is a difficult topic; it’s very societally driven,” he said. “Some technology that may be very much applicable in one country, whereas in others making use of image recognition or other tools may not be appropriate.” Vogels offered the example of London being replete with closed-circuit cameras for surveillance. “If you did that in Amsterdam, you would get a revolution,” he observed. As a “very sensitive topic,” said Vogels, to understand the responsible use of technology means getting to a fundamental fact: “On the one hand, technology can be used for good, and we’ve seen some other cases where technology can support less – not ethical uses.” “That’s a decision that will be made by regulators in combination with those who make technology.”
