Andrew Ng, a prominent Silicon Valley executive and investor who previously led some of the biggest A.I. projects at Google and its Chinese rival Baidu, says the next wave of A.I. will be in industries in which the tech giants aren’t firmly rooted. Think manufacturing, agriculture, and healthcare.
Ng is a bit biased considering that his latest venture, Landing AI, helps traditional companies adopt A.I. But he makes a compelling argument that established companies still have a chance.
Speaking at TechCrunch’s business technology conference in San Francisco last week, Ng likened the current state of A.I. to the Internet's rise in the 1990s. Companies like Apple, Microsoft, and FedEx were not Internet natives, he explained, but they were able to become “Internet companies” by creating new businesses that depended on the Web.
For instance, Apple was primarily a computer maker, but it eventually created a huge business out of its Internet-driven app store. These companies did more than merely create websites and apps and then call it a day.
Likewise, traditional companies that haven’t embedded A.I. into their businesses still have time to do so. It just won’t be as simple as buying a cloud software service “where you swipe your credit card and you use it and now your company is A.I.-enabled,” Ng said.
Instead, executives at traditional businesses must think hard about how they can apply deep learning, a key component of artificial intelligence, to their specific needs. For them, Ng has a few tips. For instance, agriculture companies could affix sensors to their farming equipment to collect data about their fields and then use A.I. techniques to analyze that data to obtain better crop yields.
The challenge is that current deep learning techniques, many of which were created by the tech giants, only work well with enormous quantities of data. Non-tech companies, like agricultural businesses, may have to develop their own A.I. techniques that rely on only small amounts of farm data, Ng said.
But if agricultural companies create neural networks—the foundational software used for data training— that learn from small amounts of data, it would be a huge breakthrough. This could level the playing field between the A.I.-powered tech giants and conventional businesses.
“One of the myths we tell in Silicon Valley is that whenever there is disruptive technology, the startups always win,” Ng says. “That’s not true.”
A.I. IN THE NEWS
The Pentagon and A.I. ethics. The Pentagon is looking to hire an ethics expert who can help the Defense Department navigate some of A.I.’s most pressing ethical concerns, trade publication Defense Systems reported. The news comes amid employee protest at companies like Google over the potential military uses of A.I. and the company’s role in selling the government powerful, data-crunching technology.
The Department of Energy’s A.I. office. The U.S. Department of Energy created the DOE Artificial Intelligence and Technology Office, which is intended to coordinate the department’s A.I. projects as part of the White House’s national A.I. strategy. Energy Sec. Rick Perry said in statement that the new office would “concentrate our existing efforts while also facilitating partnerships and access to federal data, models and high performance computing resources for America’s AI researchers.”
Singapore’s A.I. dreams. Singapore is trying to cultivate an A.I. technology scene and remain a neutral A.I. player between China and the U.S., Bloomberg News reported. The island city-state’s government is investing $500 million on A.I.-related projects through 2020, and the nation is now home to A.I. research offices of Alibaba and Salesforce.
Academic A.I. brain drain. The New York Times reported on a study showing the impact on universities and the startups they produce when A.I. professors leave their full-time academic positions to work at corporations. The study “focused on the start-up economy, showing that departures led to fewer student start-ups,” the Times reported, noting that “experts are split on whether a decline in the start-up economy will harm the progress of A.I.”
学术人工智能人才流失。《纽/约/时/报》(The New York Times)报道了一项研究，该研究显示了当人工智能教授离开全职学术岗位去公司工作时，他们对大学和初创企业的影响。《纽/约/时/报》报道称，这项研究“关注的是创业经济，研究显示离职导致学生创业数量减少”，并指出“对于创业经济下滑是否会损害人工智能的进步，专家们意见不一。”
DEEP LEARNING DOUBTS
Computer science experts Gary Marcus and Ernest Davis write in The New York Times about the limitations of deep learning technologies and why other A.I. approaches are important. The two write: “In particular, we need to stop building computer systems that merely get better and better at detecting statistical patterns in data sets — often using an approach known as deep learning — and start building computer systems that from the moment of their assembly innately grasp three basic concepts: time, space and causality.”
计算机科学专家加里·马库斯(Gary Marcus)和欧内斯特·戴维斯(Ernest Davis)在《纽/约/时/报》撰文，阐述了深度学习技术的局限性，以及为什么其他人工智能方法很重要。两人写道:“特别是，我们需要停止构建那些仅仅越来越擅长检测数据集中的统计模式的计算机系统——通常使用一种被称为深度学习的方法——并开始构建那些从组装时就天生掌握时间、空间与因果关系三个基本概念的计算机系统。”
EYE ON A.I. TALENT
Online music service Spotify hired Tony Jebara as president of engineering for personalization and to lead its machine-learning strategies. Jebara, also a Columbia University computer science professor, was previously a machine learning director at Netflix.
在线音乐服务Spotify聘请托尼•杰巴拉(Tony Jebara)担任个性化工程总裁，并负责其机器学习策略。杰巴拉也是哥伦比亚大学(Columbia University)计算机科学教授，曾在Netflix担任机器学习主管。
EYE ON A.I. RESEARCH
Deep learning’s gender problem. The Pew Research Center published a study about the difficulties deep-learning systems have identifying people’s genders based on their photos. The study showed that gender-classification systems generally work better when they are trained with a diverse set of photos representing multiple age-groups and ethnicities.
深度学习的性别问题。皮尤研究中心(Pew Research Center)发表了一项关于深度学习系统根据人们的照片识别性别存在困难的研究。研究表明，当他们接受一组代表不同年龄层和种族的不同照片训练时，性别分类系统通常会工作得更好。
In some cases, however, the researchers found that gender-classification systems can occasionally work well when trained on less diverse datasets, which the Pew Research team found surprising and confusing.
A.I.-aided drug discovery. Researchers from biotechnology firm Insilico Medicine published a paper in the Nature Biotechnology journal about using A.I. techniques to significantly decrease the amount of time it takes to create molecules useful for drug discovery. The researchers’ technology used a combination of reinforcement learning—a type of A.I. that learns through many trials—and so-called generative adversarial networks, which can be used to create realistic, but fake photos, among other tasks.
FORTUNE ON A.I.
Alarmed By Deepfake Videos, Facebook Creates Contest to Detect Them – By Jeremy Kahn
Most Americans Distrust Companies Using Facial Recognition Technology – By Jonathan Vanian
大多数美国人不信任使用面部识别技术的公司 – 乔纳森·多尼
Deepfake App Zao Makes You a Movie Star. But It Also Raises Big Privacy Concerns – By Alyssa Newcomb
深度山寨应用Zao让你成为电影明星。但这也引发了人们对隐私的巨大担忧 – 艾丽莎· 纽科姆
Making A.I. safe for the U.S. and the rest of the world. Researchers in the U.S. and China must work together on A.I. to ensure that the technology is safe, writes Matt Sheehan, a fellow at the Paulson Institute’s MacroPolo think tank, in Bloomberg. Sheehan is concerned that competition between the two countries in A.I. could lead to U.S. lawmakers severing ties between U.S. and Chinese A.I. researchers, who sometimes collaborate and communicate with each other during A.I. conferences and on research projects. Doing so, he writes, “threatens to create a dangerous knowledge vacuum on AI safety precisely when we need smart, strategic cooperation between scientists to mitigate these risks. In this case, engagement will make the U.S. far safer than isolation.”