Michael Jordan, Artificial Intelligence — The Revolution Hasn’t Happened Yet, Harvard Data Science Review, Apr 19, 2018.
When my spouse was pregnant 14 years ago, we had an ultrasound. There was a geneticist in the room, and she pointed out some white spots around the heart of the fetus. “Those are markers for Down syndrome,” she noted, “and your risk has now gone up to one in 20.” She let us know that we could learn whether the fetus in fact had the genetic modification underlying Down syndrome via an amniocentesis, but amniocentesis was risky—the chance of killing the fetus during the procedure was roughly one in 300. Being a statistician, I was determined to find out where these numbers were coming from. In my research, I discovered that a statistical analysis had been done a decade previously in the UK in which these white spots, which reflect calcium buildup, were indeed established as a predictor of Down syndrome. I also noticed that the imaging machine used in our test had a few hundred more pixels per square inch than the machine used in the UK study. I returned to tell the geneticist that I believed that the white spots were likely false positives, literal white noise.
She said, “Ah, that explains why we started seeing an uptick in Down syndrome diagnoses a few years ago. That’s when the new machine arrived.”
We didn’t do the amniocentesis, and my wife delivered a healthy girl a few months later, but the episode troubled me, particularly after a back-of-the-envelope calculation convinced me that many thousands of people had gotten that diagnosis that same day worldwide, that many of them had opted for amniocentesis, and that a number of babies had died needlessly. The problem that this episode revealed wasn’t about my individual medical care; it was about a medical system that measured variables and outcomes in various places and times, conducted statistical analyses, and made use of the results in other situations. The problem had to do not just with data analysis per se, but with what database researchers call provenance—broadly, where did data arise, what inferences were drawn from the data, and how relevant are those inferences to the present situation? While a trained human might be able to work all of this out on a case-by-case basis, the issue was that of designing a planetary-scale medical system that could do this without the need for such detailed human oversight....
The developments now being called AI arose mostly in the engineering fields associated with low-level pattern recognition and movement control, as well as in the field of statistics, the discipline focused on finding patterns in data and on making well-founded predictions, tests of hypotheses, and decisions.
Indeed, the famous backpropagation algorithm that David Rumelhart rediscovered in the early 1980s, and which is now considered at the core of the so-called “AI revolution,” first arose in the field of control theory in the 1950s and 1960s. One of its early applications was to optimize the thrusts of the Apollo spaceships as they headed towards the moon....
While some view this challenge as subservient to the creation of artificial intelligence, another more prosaic, but no less reverent, viewpoint is that it is the creation of a new branch of engineering. Much like civil engineering and chemical engineering in decades past, this new discipline aims to corral the power of a few key ideas, bringing new resources and capabilities to people, and to do so safely. Whereas civil engineering and chemical engineering built upon physics and chemistry, this new engineering discipline will build on ideas that the preceding century gave substance to, such as information, algorithm, data, uncertainty, computing, inference, and optimization. Moreover, since much of the focus of the new discipline will be on data from and about humans, its development will require perspectives from the social sciences and humanities....
The past two decades have seen major progress—in industry and academia—in a complementary aspiration to human-imitative AI that is often referred to as “Intelligence Augmentation” (IA). Here computation and data are used to create services that augment human intelligence and creativity. A search engine can be viewed as an example of IA, as it augments human memory and factual knowledge, as can natural language translation, which augments the ability of a human to communicate. Computer-based generation of sounds and images serves as a palette and creativity enhancer for artists. While services of this kind could conceivably involve high-level reasoning and thought, currently they don’t; they mostly perform various kinds of string-matching and numerical operations that capture patterns that humans can make use of.
Hoping that the reader will tolerate one last acronym, let us conceive broadly of a discipline of “Intelligent Infrastructure” (II), whereby a web of computation, data, and physical entities exists that makes human environments more supportive, interesting, and safe. Such infrastructure is beginning to make its appearance in domains such as transportation, medicine, commerce, and finance, with implications for individual humans and societies. This emergence sometimes arises in conversations about an Internet of Things, but that effort generally refers to the mere problem of getting ‘things’ onto the Internet, not to the far grander set of challenges associated with building systems that analyze those data streams to discover facts about the world and permit ‘things’ to interact with humans at a far higher level of abstraction than mere bits....
We now come to a critical issue: is working on classical human-imitative AI the best or only way to focus on these larger challenges? Some of the most heralded recent success stories of ML have in fact been in areas associated with human-imitative AI—areas such as computer vision, speech recognition, game-playing, and robotics. Perhaps we should simply await further progress in domains such as these. There are two points to make here. First, although one would not know it from reading the newspapers, success in human-imitative AI has in fact been limited; we are very far from realizing human-imitative AI aspirations. The thrill (and fear) of making even limited progress on human-imitative AI gives rise to levels of over-exuberance and media attention that is not present in other areas of engineering.
Second, and more importantly, success in these domains is neither sufficient nor necessary to solve important IA and II problems. On the sufficiency side, consider self-driving cars. For such technology to be realized, a range of engineering problems will need to be solved that may have little relationship to human competencies (or human lack-of-competencies). The overall transportation system (an II system) will likely more closely resemble the current air-traffic control system than the current collection of loosely-coupled, forward-facing, inattentive human drivers. It will be vastly more complex than the current air-traffic control system, specifically in its use of massive amounts of data and adaptive statistical modeling to inform fine-grained decisions. Those challenges need to be in the forefront versus a potentially-distracting focus on human-imitative AI.
Michael I. Jordan, Dr. AI or: How I Learned to Stop Worrying and Love Economics, Jul 02, 2019
Indeed, classical recommendation systems can and do cause serious problems if they are rolled out in real-world domains where there is scarcity. Consider building an app that recommends routes to the airport. If few people in a city are using the app, then it is benign, and perhaps useful. When many people start to use the app, however, it will likely recommend the same route to large numbers of people and create congestion. The best way to mitigate such congestion is not to simply assign people to routes willy-nilly, but to take into account human preferences—on a given day some people may be in a hurry to get to the airport and others are not in such a hurry. An effective system would respect such preferences, letting those in a hurry opt to pay more for their faster route and allowing others to save for another day. But how can the app know the preferences of its users? It is here that major IT companies stumble, in my humble opinion. They assume that, as in the advertising domain, it is the computer's job to figure out human users' preferences, by gathering as much information as possible about their users, and by using AI. But this is absurd; in most real-world domains—where our preferences and decisions are fine-grained, contextual, and in-the-moment—there is no way that companies can collect enough data to know what we really want. Nor would we want them to collect such data—doing so would require getting uncomfortably close to prying into the private thoughts of individuals. A more appealing approach is to empower individuals by creating a two-way market where (say) street segments bid on drivers, and drivers can make in-the-moment decisions about how much of a hurry they are in, and how much they're willing to spend (in some currency) for a faster route....
Certainly market forces are not a panacea. But market forces are an important source of algorithmic ideas for constructing intelligent systems, and we ignore them at our peril. We are already seeing AI systems that create problems regarding fairness, congestion, and bias. We need to reconceptualize the problems in such a way that market mechanisms can be taken into account at the algorithmic level, as part and parcel of attempting to make the overall system be ‘intelligent.’ Ignoring market mechanisms in developing modern societal-scale information-technology systems is like trying to develop a field of civil engineering while ignoring gravity.
沒有留言:
張貼留言