我在4月5日投稿的评论稿件《突发危机暴露AI的局限》在4月15日刊登。距离4月5日，各地人工智能专才可能已经开始掌握更多的data point和parameter来创建更具时效的人工智能模型。但是，我始终认为，以人工神经网络为主干的机器学习和深度学习，其实是有亟待解决的软硬体技术缺陷的。这两天，我尝试做更多的阅读，然后把当下AI算法在技术层面上的缺陷，摘录于我的部落格文章如下，以补充我之前的文章因篇幅所限，无法阐述所有技术细节的不足之处。
------------
世界经济论坛的一篇文章指出，人工智能有潜力协助人们解决由冠状病毒引发的燃眉危机。然而解决方案不会仅是科技，而是包括驾驭人工智能系统之人类的知识以及创意。
文章也不留情面的表示，冠状病毒危机很可能会暴露人工智能的重大短处。这是因为当前主流的人工智能形式 -- 机器学习，以发掘过往训练数据的规律为操作方式。然而这类的人工智能系统需要大量高度相关的针对性数据来探查规律。机器学习也假定今时今日输入的数据所处的条件与过往培训建模时所面对的条件没有不同。
换句话说，人工智能系统认为以往应验的方案在未来会同样收效。
冠状病毒疫情让我们面临着前所未有的时代。 我们在每一个当下所面临的情况，很大可能与几周前的状况截然不同。 我们今天需要尝试的一些以前从未试过的新事物。 同样的，过去有效的方法今天很可能已经失效。
由是观之，人类与AI所面临的局限性没有什么不同，这在一定程度上解释了为什么我们目前的形势如此严峻。 由于没有先例可循，我们无法确定最佳的行动方案。 传统上我们惯以认定的因果关系可能不再成立。
但是，人类还是比人工智能更有优势。 我们能够从一种环境中学习教训，将其应用于新的状况，并利用我们的抽象知识对可能起作用或可能发生的事做出最佳猜测。 反之，AI不具人类举一反三的能耐，每当设置或任务发生微小变化时，AI系统就必须从新学习来掌握具体的状况。
[QUOTE] Artificial intelligence (AI) has the potential to help us tackle the pressing issues raised by the COVID-19 pandemic. It is not the technology itself, though, that will make the difference but rather the knowledge and creativity of the humans who use it.
Indeed, the COVID-19 crisis will likely expose some of the key shortfalls of AI. Machine learning, the current form of AI, works by identifying patterns in historical training data. When used wisely, AI has the potential to exceed humans not only through speed but also by detecting patterns in that training data that humans have overlooked.
However, AI systems need a lot of data, with relevant examples in that data, in order to find these patterns. Machine learning also implicitly assumes that conditions today are the same as the conditions represented in the training data. In other words, AI systems implicitly assume that what has worked in the past will still work in the future.
What does this have to do with the current crisis? We are facing unprecedented times. Our situation is jarringly different from that of just a few weeks ago. Some of what we need to try today will have never been tried before. Similarly, what has worked in the past may very well not work today.
Humans are not that different from AI in these limitations, which partly explains why our current situation is so daunting. Without previous examples to draw on, we cannot know for sure the best course of action. Our traditional assumptions about cause and effect may no longer hold true.
Humans have an advantage over AI, though. We are able to learn lessons from one setting and apply them to novel situations, drawing on our abstract knowledge to make best guesses on what might work or what might happen. AI systems, in contrast, have to learn from scratch whenever the setting or task changes even slightly. [/QUOTE]
出处/Reference
World Economic Forum: AI can help with the COVID-19 crisis - but the right human input is key
https://www.weforum.org/agenda/2020/03/covid-19-crisis-artificial-intelligence-creativity/
------------
[引述] 另一篇于2016年刊载于《哈佛商业评论》的文章点出，算法（Algorithm）不懂得权衡取舍（trade off），只会一心一意的追逐目标（pursue objectives single-mindedly）。
算法能够做出较准确的预测，但也产生自身的风险，尤其是当我们不了解它们时。
举例来说，今天许多网站都部署算法来决定呈现给用户什么样的在线广告和链接。当这些算法过分专注、狭隘的追逐用户点击率的最大化时，网站充塞了低素质、只图垂钓用户点击的文章。结果是点击率提升，用户满意度一落千丈。
为了避免出错，负责管理算法系统的经理需要了解什么算法表现良好（what algorithms do well），包括它们回答什么问题（what questions they answer），不回答什么问题（what questions they do not）。
一个算法能够阅读每篇《纽约时报》的文章然后告诉你哪一篇文章最可能被推特用户转发，却不一定能解释为什么人们把文章发到推特上。一个算法能够告诉你哪一位员工最可能会成功，却无法探知哪一些最重要的特质促成了员工的成功。[/引述]
[QUOTE] Algorithms don’t understand trade-offs; they pursue objectives single-mindedly.
Algorithms make predictions more accurate—but they also create risks of their own, especially if we do not understand them.
For example, today many sites deploy algorithms to decide which ads and links to show users. When these algorithms focus too narrowly on maximizing user click-throughs, sites become choked with low-quality “click-bait” articles. Click-through rates rise, but overall customer satisfaction may plummet.
To avoid missteps, managers need to understand what algorithms do well—what questions they answer and what questions they do not.
An algorithm can read through every New York Times article and tell you which is most likely to be shared on Twitter without necessarily explaining why people will be moved to tweet about it. An algorithm can tell you which employees are most likely to succeed without identifying which attributes are most important for success. [/QUOTE]
出处/Reference
Harvard Business Review: Algorithms Need Managers, Too
https://hbr.org/2016/01/algorithms-need-managers-too
------------
[引述] 一篇于2018年刊载于《南华早报》的科技评论文章指出，人工智能的最大局限性在于它仅与所提供的数据集一样聪明。
实际的情况是，人工智能只会在特定业务的应用中产生最深远的影响。它主要的限制还是在于它得从给定的数据中学习。不同于人类的学习，它没有其他方法可以整合知识。这意味着数据中的任何错误都会反映在人工智能的运算结果中。今时今日的AI系统被训练来完成明确定义的任务。一个下棋的系统不能用来玩单人纸牌或扑克牌。[/QUOTE]
[QUOTE] The biggest limitation of artificial intelligence is it’s only as smart as the data sets served.
The reality is that the impact of AI will be most profound in specific business applications. The main limitation is that it learns from given data. There is no other way that knowledge can be integrated, unlike human learning. This means that any inaccuracies in the data will be reflected in the results. Today’s AI systems are trained to do a clearly defined task. The system that plays chess cannot play solitaire or poker. [/QUOTE]
参考资料/Reference
South China Morning Post: The biggest limitation of artificial intelligence is it’s only as smart as the data sets served
https://www.scmp.com/business/china-business/article/2131903/biggest-limitation-artificial-intelligence-its-only-smart
https://www.scmp.com/business/china-business/article/2131903/biggest-limitation-artificial-intelligence-its-only-smart
------------
麦肯锡全球研究院的专才则在一个播客对谈中，点出了当人们希冀应用人工智能来解决人类在判断过程中惯常出现、难以避免的偏见时，可能正在不自觉中让人工智能产生另一种意想不到的决策偏颇。
还有就是，许多世界一流团队逐年激烈竞争，先后宣称取得的重大人工智能科研突破过程中所采用的评比标杆数据，都是一样的公共开源数据。这些数据的象征意义也许大于实际作用。因为现实世界的数据往往更多元纷呈、瞬息万变，非一般年度科研比赛所采用的相对静态式的标杆数据所能比拟。
大多数人都以公共数据作为衡量人工智能建模运算准确度的基准参照点。因此，如果每个人都应用有内在偏颇缺陷的相同一组开放数据，我们其实在复制大规模的人工智能偏颇。
Title: McKinsey Podcast: The real-world potential and limitations of artificial intelligence
-- McKinsey Global Institute partner Michael Chui and MGI chairman and director James Manyika speak with McKinsey Publishing’s David Schwartz about the cutting edge of artificial intelligence
Source: https://www.mckinsey.com/featured-insights/artificial-intelligence/the-real-world-potential-and-limitations-of-artificial-intelligence
[QUOTE] David Schwartz: At some level, I’m hearing from the questions and from what the rejoinder might be that there’s a very human element. A question would be: Why is the answer such and such? And the answer could be, it’s the algorithm. But somebody built that algorithm, or somebody—or a team of somebodies—and machines built that algorithm. That brings us to a limitation that is not quite like the others: bias—human predilections. Could you speak a little bit more about what we’re up against, James?
It becomes very, very important to think through what might be the inherent biases in the data, in any direction.
James Manyika: The question of bias is a very important one. And I’d put it into two parts.
Clearly, these algorithms are, in some ways, a big improvement on human biases. This is the positive side of the bias conversation. We know that, for example, sometimes, when humans are interpreting data on CVs [curriculum vitae], they might gravitate to one set of attributes and ignore some other attributes because of whatever predilections that they bring. There’s a big part of this in which the application of these algorithms is, in fact, a significant improvement compared to human biases. In that sense, this is a good thing. We want those kinds of benefits.
But I think it’s worth having the second part of the conversation, which is, even when we are applying these algorithms, we do know that they are creatures of the data and the inputs you put in. If those inputs you put in have some inherent biases themselves, you may be introducing different kinds of biases at much larger scale.
The work of people like Julia Angwin and others has actually shown this if the data collected is already biased. If you take policing as an example, we know that there are some communities that are more heavily policed. There’s a much larger police presence. Therefore, the data we’ve got and that’s collected about those environments is much, much, much higher. If we then start to compare, say, two neighborhoods, one where it’s oversampled—meaning there’s lots and lots of data available for it because there’s a larger police presence—versus another one where there isn’t much policing so, therefore, there isn’t much data available, we may draw the wrong conclusions about the heavily policed observed environment, just simply because there’s more data available for it versus the other one.
The biases can go another way. For example, in the case of lending, the implications might go the other way. For populations or segments where we have lots and lots of financial data about them, we may actually make good decisions because the data is largely available, versus in another environment where we’re talking about a segment of the population we don’t know much about, and the little bit that we know sends the decision off in one way. And so, that’s another example where the undersampling creates a bias.
The point about this second part is that I think it becomes very, very important to make sure that we think through what might be the inherent biases in the data, in any direction, that might be in the data set itself—either in the actual way it’s constructed, or even the way it’s collected, or the degree of sampling of the data and the granularity of it. Can we debias that in some fundamental way?
This is why the question of bias, for leaders, is particularly important, because it runs a risk of opening companies up to all kinds of potential litigation and social concern, particularly when you get to using these algorithms in ways that have social implications. Again, lending is a good example. Criminal justice is another example. Provision of healthcare is another example. These become very, very important arenas to think about these questions of bias.
Michael Chui: Some of the difficult cases where there’s bias in the data, at least in the first instance, isn’t around, as a primary factor, people’s inherent biases about choosing either one or the other. It is around, in many cases, these ideas about sampling—sampling bias, data-collection bias, et cetera—which, again, is not necessarily about unconscious human bias but an artifact of where the data came from.
There’s a very famous case, less AI related, where an American city used an app in the early days of smartphones that determined where potholes were based on the accelerometer shaking when you drove over a pothole. Strangely, it discovered that if you looked at the data, it seemed that there were more potholes in affluent parts of the city. That had nothing to do with the fact there were actually more potholes in that part of the city, but you had more signals from that part of the city because more affluent people had more smartphones at the time. That’s one of those cases where it wasn’t because of any intention to not pay attention to certain parts of the city. Understanding the providence of data—understanding what’s being sampled—is incredibly important.
There’s another researcher who has a famous TED Talk, Joy Buolamwini at MIT Media Lab. She does a lot of work on facial recognition, and she’s a black woman. And she says, “Look, a lot of the other researchers are more male and more pale than I am. And as a result, the accuracy for certain populations in facial recognition is far higher than it is for me.” So again, it’s not necessarily because people are trying to exclude populations, although sometimes that happens, it really has to do with understanding the representativeness of the sample that you’re using in order to train your systems.
So, as a business leader, you need to understand, if you’re going to train machine-learning systems: How representative are the training sets there that you’re using?
People forget that one of the things in the AI machine-deep-learning world is that many researchers are using largely the same data sets that are shared—that are public.
James Manyika: It actually creates an interesting tension. That’s why I described the part one and the part two. Because in the first instance, when you look at the part-one problem, which is the inherent human biases in normal day-to-day hiring and similar decisions, you get very excited about using AI techniques. You say, “Wow, for the first time, we have a way to get past these human biases in everyday decisions.” But at the same time, we should be thoughtful about where that takes us to when you get to these part-two problems, where you now are using large data sets that have inherent biases.
I think people forget that one of the things in the AI machine-deep-learning world is that many researchers are using largely the same data sets that are shared—that are public. Unless you happen to be a company that has these large, proprietary data sets, people are using this famous CIFAR data set, which is often used for object recognition. It’s publicly available. Most people benchmark their performance on image recognition based on these publicly available data sets. So, if everybody’s using common data sets that may have these inherent biases in them, we’re kind of replicating large-scale biases. This tension between part one and part two and this bias question are very important ones to think through. The good news, though, is that in the last couple years, there’s been a growing recognition of the issues we just described. And I think there are now many places that are putting real research effort into these questions about how you think about bias. [/QUOTE]
------------
著名美国杂志《连线》曾在一篇报导中指出，深度学习的发展是模式识别技术的产物。人工神经网络以记忆不同组别的事物作为辨识数据规律的基础，所以在遇见“似曾相识”的事物时就能下大致上准确的结论。可是大多数有趣的认知性命题与分门别类无关。一位谷歌的研究员弗兰科伊斯.科勒就表示：“人们天真的以为如果你运用深度学习技术，增加一百倍的神经网络分层，再加上多一千倍的数据，人工神经网络就能做到任何人类可以达到的事。但，那是不正确的。”
根据纽约大学教授马可斯所述，深度学习是贪心、脆弱、不透明和短浅的。说深度学习系统“贪心”，是因为它们需要大量的训练数据；认为它们“脆弱”，是因为当神经网络被给予“转移测试”时（遭遇与训练阶段截然不同的方案之数据），它们无法准确掌握情境继而惯常出错；称它们“不透明”，是因为它们与具备正规、可调试代码的传统编程不同，神经网路系统的参数只能以数学几何的权重来诠释。这样的结果是，它们就如同运算结论无法被轻易解释的黑箱，让人们质疑它们的可靠性和担心它们的偏颇；最后，它们被认为“浅显”，是因为它们的编程方式赋予极少的内在知识。它们不具备这个世界，还有人类心理学的通识。
[QUOTE] Deep learning’s advances are the product of pattern recognition: neural networks memorize classes of things and more-or-less reliably know when they encounter them again. But almost all the interesting problems in cognition aren’t classification problems at all. “People naively believe that if you take deep learning and scale it 100 times more layers, and add 1000 times more data, a neural net will be able to do anything a human being can do,” says François Chollet, a researcher at Google. “But that’s just not true.”
According to skeptics like Marcus, deep learning is greedy, brittle, opaque, and shallow. The systems are greedy because they demand huge sets of training data. Brittle because when a neural net is given a “transfer test”—confronted with scenarios that differ from the examples used in training—it cannot contextualize the situation and frequently breaks. They are opaque because, unlike traditional programs with their formal, debuggable code, the parameters of neural networks can only be interpreted in terms of their weights within a mathematical geography. Consequently, they are black boxes, whose outputs cannot be explained, raising doubts about their reliability and biases. Finally, they are shallow because they are programmed with little innate knowledge and possess no common sense about the world or human psychology. [QUOTE]
出处/Reference
WIRED magazine: Greedy, Brittle, Opaque, and Shallow: The Downsides to Deep Learning
-- We've been promised a revolution in how and why nearly everything happens. But the limits of modern artificial intelligence are closer than we think.
https://www.wired.com/story/greedy-brittle-opaque-and-shallow-the-downsides-to-deep-learning/
Posts
0 Comments