2018-07-03
目前企业对人工智能（AI）的采用比较缓慢，但是专家预计这一趋势将会迅速增长。目前，科技巨头们正在引领风骚。
尽管与人工智能有关的宣传和产品公告大量涌现，但是到目前为止很少有企业采用了这一技术。
Gartner公司副总裁兼杰出分析师惠特.安德鲁斯（Whit Andrews）能够为这一趋势提供一些硬性数据。他说：“企业正处于采用人工智能的最早阶段。具体来说，在我们最近一次针对首席信息官（CIO）的调查中，25位首席信息官中只有一人表示其企业已经在采用人工智能。”
认知计算联盟（Cognitive Computing Consortium）的董事总经理兼联合创始人哈德雷 雷纳兹（Hadley Reynolds）表示，科技巨头公司的人工智能技术最为成熟。他说，这些公司“把大部分业务放在各种机器学习和深度学习技术上”，因此他们在研究人工智能技术和招聘具备AI技能的人才方面投入了大量预算。
绝大多数其他公司——那些还没有将业务模式放在人工智能上的公司——则还未登上人工智能这艘大船。
但是，大多数专家预计，采用人工智能的企业将会迅速增加。根据安德鲁斯的说法，“在25位首席信息官中，6位表示正在试行或计划在短期内采用人工智能，5位表示计划在中期内采用人工智能。”
研究公司IDC的分析师对AI增长曲线给出了类似的结论。他们预测，2017年全球认知和人工智能系统的收入将达到125亿美元，比2016年增长59.3％。此外，他们预计到2020年，这一收入将以每年54.4％的增长率继续增长，届时将达到460亿美元。
专家指出，最终几乎所有的公司都会受到人工智能的影响，但目前大多数公司还不确定人工智能能为他们做些什么，或者他们应该如何使用人工智能。研究公司451 Research的联合创始人兼软件研究副总裁尼克 佩兴斯（Nick Patience）表示“人们对此有些困惑”。他补充说，大多数企业还处于“非常早期的阶段”。
有些困惑可能是因为人工智能有很多可能的用例。专家指出，AI技术一些可能的应用包括服务台、客户支持、推荐引擎、欺诈检测、聊天机器人、图像识别、语言处理和市场细分。安德鲁斯指出，人工智能甚至有助于提高大学毕业率或减少入狱人员再犯罪。
那么公司应该从何处着手呢？
研究人员建议从整个公司正在努力解决的核心问题开始，且公司在该问题上拥有大量的数据——数据太多，人类分析师自行分析这些数据需要花费很长的时间。
安德鲁斯说：“人工智能在帮助人类进行分类和预测方面非常出色。”
佩兴斯说，这些公司“把大部分业务放在各种机器学习和深度学习技术上”，机器学习和深度学习都是人工智能的子集，因此他们在研究人工智能技术和招聘具备AI技能的人才方面投入了大量预算。
但是，在企业看到将AI应用于这些用例的任何好处之前，他们需要克服一些棘手的挑战。
最大的问题是缺乏人才。据安德鲁斯说，“一半以上的企业认为，实施人工智能的最大障碍是他们没有这方面的技能。”
另一个问题是搞清楚你想用AI来做什么。佩兴斯说：“这个挑战在于理解某个领域是否是应用AI和机器学习的良好选择。”许多企业表示，他们不知道如何开始，或者他们没有实施AI项目所需的战略或资金。
向AI系统提供的数据也面临挑战。佩兴斯说：“收集和管理数据是一个关键的挑战。我们一次又一次地看到，企业或者没有足够的数据，或者拥有足够的数据却无法访问。”
技术本身也是一个问题。尽管近年来人工智能研究进展迅速，但我们仍然没有真正像人类一样思考和学习的通用人工智能。因此，人类与人工智能的交互有时不甚理想。“计算机本身的‘笨拙’也是一个严峻的挑战”，哈德利补充说，“从客户体验的角度来看，计算机更有可能出现错误和灾难。”
这导致了更大的问题：信任。“该领域发展的首要问题是：人们相信他们从机器获得的结果吗？”雷诺兹说，“每次人工智能交互中都存在某种信任问题。”
人工智能可能还需要一段时间才能真正获得普通消费者的信任。安德鲁斯指出，人们有两大担忧：人工智能失控或人工智能取代人类的工作。
安德鲁斯说：“我认为人工智能远远未达到自主或独立于人类社会的时空，而且任何人工智能都不会自主到引发风险的程度。”
然而，在就业方面，人们可能会有更多的担忧。安德鲁斯说：“Gartner公司曾经说过，AI创造的就业机会比消除的要多。但是我认为，假设所有工作被取代的人都能找到新工作或新工作与旧工作等同的想法是错误的。”
佩兴斯指出，人工智能可能会导致一些行业出现重大动荡。他说：“长期来看，一些市场将会发生巨大的变化。”他认为金融服务、零售和医疗行业会受到最严重的冲击。
有些人——和有些企业——可能难以适应这些变化。
尽管如此，分析师们仍然有理由对AI的未来潜力持乐观态度。
安德鲁斯说：“人工智能使人类能够更好地享受生活。我们这才刚刚揭开了序曲。”
哈德利 雷诺兹：专注于你真正想要解决的特定业务问题，并在评估AI技术方面寻求一些帮助。
尼克 佩兴斯：着眼一些业务流程已经被写入软件的领域，因为这个流程很可能是应用机器学习的不错选择。
惠特.安德鲁斯：确定企业希望你改进的关键绩效指标......如果你有足够的时间和空间来做实验，那么你应该尝试人工智能。
雷诺兹：由于存在大量的数据和大量的用户，AI系统的成功试点和大规模部署之间可能充满风险。
佩兴斯：确保您了解AI对用户（无论是内部用户还是外部用户）体验的影响。
安德鲁斯：找出对你的企业来说很可能是独特的或特殊的领域，然后扩大投入。
雷诺兹：如果你只有一个问题，那你没事了。如果不只有一个问题，你需要成立一个可以把AI扩展到其他问题的内部团队。
佩兴斯：下手够早，干得好！请寻找新的和更高级的方式来研究和使用AI。
安德鲁斯：创建一个在整个企业中具有一致性和稳定性的战略。
1/15/2018
08:00 AM
Enterprise adoption of AI is slow today, but experts expect it to increase very rapidly. So far, tech giants are leading the charge.
Despite a flood of publicity and product announcements related to artificial intelligence, it seems that few enterprises have adopted the technology so far.
Whit Andrews, vice president and distinguished analyst at Gartner, was able to put some hard numbers to the trend. "We are in the very earliest stages of enterprise adoption of artificial intelligence," he said. "Specifically, in our most recent CIO survey from 2017, one in 25 CIOs described themselves as having artificial intelligence in action in their organizations."
The companies farthest along with the technology tend to be technology giants, said Hadley Reynolds, managing director and co-founder of the Cognitive Computing Consortium. These companies are "basing much of their businesses on various kinds of machine learning and deep learning technologies," he said, so they have invested heavily in research and recruiting talent with AI skills.
The vast majority of other companies — those that are not basing their business models on AI today — haven't yet gotten on board.
More enterprises expected to jump in
However, most experts expect enterprise adoption of AI to ramp up quickly. According to Andrews, "Six in 25 are either piloting or have AI in their short-term plans. Five in 25 have it in middle-term plans."
IDC analysts have come to a similar conclusion regarding the AI growth curve. They are forecasting that worldwide revenues for cognitive and AI systems will reach $12.5 billion in 2017, an increase of 59.3% over 2016. And, they expect growth to continue at 54.4% per year through 2020, at which point revenues could top $46 billion.
Experts say that eventually nearly every company will be impacted by AI, but most companies aren't yet sure what AI can do for them or how they can use it. "There’s quite a bit of confusion," said Nick Patience, co-founder and research vice president of software at 451 Research. Most enterprises are in the "very, very, very early stages," he added.
Leading AI use cases
Some of that confusion may be because there are so many potential use cases for AI. Experts pointed to help desk, customer support, recommendation engines, fraud detection, chatbots, image recognition, language processing and market segmentation as some of the possible applications of the technology. Andrews pointed out that AI could even be helpful at tasks like improving graduation rates at universities or reducing recidivism at prisons.
So where should companies begin?
Observers recommend starting with a core problem that the entire company is trying to solve. To be a good candidate for AI, it needs to be a problem about which the organization has a lot of data — so much data that it would take too long for humans to analyze it on their own.
"AI is really good at making it possible for humans to be much better at classifying and predicting," Andrews said.
Patience of 451 Research said these companies are "basing much of their businesses on various kinds of machine learning and deep learning technologies," both of which are subsets of artificial intelligence, so they have invested heavily in research and in recruiting talent with AI skills.
Top AI challenges
But before organizations can see any benefit from applying AI to those use cases, they will need to overcome some thorny challenges.
The biggest problem is the lack of talent. According to Andrews, "More than half of organizations say the biggest obstacle to their implementing AI is that they don't have the skills to do it."
Another big issue is figuring out what you want to do with AI. "The simplest challenge is simply understanding whether it's a good use case for AI and machine learning," said Patience. Many organizations say they aren't sure how to get started, or they don't have the strategy or funding that they need to get a project off the ground.
The data that feeds AI systems can also present obstacles. "The gathering and curation of data is a key challenge," said Patience. "We see that over and over again, where either organizations don't have enough data, or they have it and can't get access to it.”
Then there are the problems with the technology itself. While AI research has advanced incredibly quickly in recent years, we still don't have a general artificial intelligence that truly thinks and learns the way humans do. As a result, human interactions with AI are sometimes less than satisfactory. "The 'klutziness,' if you will, of a computer itself is a serious challenge," Hadley said, adding, "The opportunities for mistakes and disasters from the point of view of the customer experience are much more likely."
That leads to a bigger issue: trust. "The overarching issue in the whole development of the field is do people trust the results that they get out of a machine?" Reynolds said. "There’s some sort of version of that trust problem in every AI interaction."
The future of AI
It may be some time before AI can truly gain the trust of the average consumer. Andrews pointed out that people tend to have two big concerns: AI getting out of control or AI taking over jobs.
"I don't think artificial intelligence is anywhere near the space or time where it establishes autonomy or independence from human society, and I don’t see a pathway in which any such intelligence would be sufficiently autonomous to present a risk," Andrews said.
However, on the jobs front, people might have more cause for concern. Andrews said, "Gartner has said that AI creates more jobs than it eliminates. But I think it would be a mistake to assume that everyone whose jobs are changed find their way to a new job or that the jobs would have parity with one another."
Patience pointed out that AI could result in major upheaval for some industries. "Long-term some markets are going to have massive changes," he said, citing financial services, retail and healthcare as those most primed for disruption.
Some people — and some enterprises — might find it difficult to adapt to those changes.
Still, the analysts see reason to be optimistic about the future potential of AI.
"AI makes it possible for humans to be better at being human," said Andrews. "We're just getting started on this chapter."
Just a Bit of Advice
From Getting Started With AI up to Production
If you are just getting started with AI:
Hadley Reynolds of the Cognitive Computing Consortium: "Focus on a specific business problem that you really want to solve," and get some help evaluating the technology.
Nick Patience, co-founder and research vice president of software at 451 Research: Look at areas where you already have some business process written into software, because that process is probably a good candidate for machine learning.
Whit Andrews, vice president and distinguished analyst at Gartner: "Identify the key performance indicators that your organization expects you to improve. . .Then if you have time and space to experiment, that's where you should try AI."
If you have an AI pilot project:
Reynolds: "With AI systems, the gap between a successful pilot and a large rollout can be fraught with risk" because of the large quantities of data and large number of users.
Patience: Make sure you understand the impact the AI will have on the user experience, whether they are internal users or external users.
Andrews: Figure out what you learned that is "likely to be unique or particular to your organization" and then expand the effort.
If you have AI in production:
Reynolds: "If you only have one problem, you're done." If not, you need to grow an in-house team that can expand AI to the other problems in your company.
Patience: "Well-done for being early." Look for new and more advanced ways of doing and using AI.
Andrews: "Move toward a strategy that has consistency and stability across the organization."
附件：
《AI Begins to Infiltrate the Enterprise》--原文.pdf
《AI Begins to Infiltrate the Enterprise》--译文.pdf
微信公众号