Research Snapshot: Insights abound at the O’Reilly AI Conference
Last month, I was one of the lucky ones attending the O’Reilly Artificial Intelligence (AI) Conference in NYC. Billed as a “rare opportunity to bypass the hype and discover how emerging developments can be applied into practical and profitable AI you can implement in your business today,” the line-up of speakers was impressive, and the content did not fail to provide attendees with some great first-hand looks at what is happening in the world of AI today.
Presented in conjunction with Intel AI, the event featured such powerhouse speakers as Stanford University’s Chris Ré; Olga Troyanskaya, professor of computer science at Princeton University and deputy director for genomics at the Flatiron Institute; Ruchir Puri, CTO and chief architect, IBM Watson; and Danielle Dean, principal data scientist lead at Microsoft (pictured above). One of the biggest takeaways from the conference was the sheer magnitude of impact that AI is having on the industry group that was present. Even more, the presentations and discussions made it clear that AI’s ability to transform is continuing to accelerate faster than ever.
Also featured were more than a dozen business leaders who presented their own enterprise use cases and demonstrated the very real value of using AI to plow through their data to drive revenue growth and to gain greater intelligence from every customer interaction. I was surprised to find that so many organizations already have clearly defined plans for using AI capabilities to achieve their business goals. What seems to have shifted in the past 18 months or so is that enterprises are finally recognizing the need for a new playbook to take advantage of the emerging business models that are being driven by AI.
AI’s trajectory—and competitive threat
Much of the content of the conference was aimed at educating business leaders about the current trajectory of AI and its potential impact on the enterprise. The results of the latest survey of enterprise spending intention by O’Reilly Media left no doubt that companies who fail to invest now will be at risk in the years to come. While it was clear that the level of spending is dependent on the maturity of an organization, the survey showed that 43% of organizations that are relatively established plan to invest at least 20% of their IT budget in AI during the next 12 months. Ben Lorica, Chief Data Scientist at O’Reilly Media, stated that he believes the gap between leaders and laggards will widen further due to lack of investment in AI by the less mature companies.
Another interesting survey finding was that the “lack of skilled people” remains one of the key factors that may inhibit AI adoption within many organizations. 57% of all survey respondents signaled that their companies were in need of machine learning experts and data scientists. Ben pointed out that this is likely not the only major skills gap. As with any new technology, he said that companies also need people who are able to identify use cases that lend themselves to AI solutions.
Some of the key innovations that were touched on included:
- Deep learning (deep neural networks). AI is evolving at an amazingly rapid pace. Perhaps the most dramatic example of this is the creation of deep neural networks that can mimic the function of a human brain. These computer-simulated networks of neurons are able to power through massive troves of data to highlight anomalies and analyze patterns that, in the past, would have taken humans days, weeks, or months to detect.
- Generative adversarial networks (GAN). One of the hottest topics at the conference, GANs are today’s most talked about machine learning models. Using two competing neural networks—a generator network and a discriminator network—GANs simplify a variety of processes by “tricking” each other to drive improvements. (It’s no wonder Facebook’s AI research director Yann LeCun called adversarial training “the most interesting idea in the last 10 years in machine learning.” It’s fascinating!) In industry, GANs can be used to manipulate data to, for instance, allow an ecommerce website to personalize the customer experience by changing the color of a shirt or even changing the face and body type of the model. GANs adversarial training capabilities are being used to increase the safety of fully autonomous vehicles, and to advance the applications of gene editing and gene therapy. (This article paints a clear picture of how GAN works and its implications in the art world and beyond.)
- Computer vision. According to the O’Reilly team, half of all AI-related patents today are in computer vision, with a growth of 24% CAGR in the last 3 years. Advancements in computer vision, including higher resolution and stronger video capabilities, are a key driver for AI’s adoption in the healthcare and automotive industries.
- Natural language processing (NLP). More accurate than ever before, NLP is now able to translate multiple languages in mere seconds—a shift that has transformed the abilities of virtual assistant platforms such as Siri and Alexa.
- Open platforms. For developers, free deep learning tools such as TensorFlow and Pytorch that are available to anyone via the web have accelerated their ability to optimize computer resources to automate processes. As the open ecosystem for interchangeable AI models continues to explode, so do the number of libraries being created and use cases being shared, with reinforcement learning considered to be the next major paradigm shift.
- Hardware of training and inference. The AI community is widely anticipating a next-generation of training solutions from chip providers Intel and Nvidia that are expected later this year. These solutions are designed to be more energy efficient and support edge devices, including routers, routing switches, integrated access devices (IADs), multiplexers, and MAN and WAN access devices. As Ben Lorica reiterated, while much of AI today requires a large amount of training data, processing that data requires both engineers to train these large models and infrastructure networks to enable them.
Case studies in advanced AI
The enterprise use cases were an exciting part of the conference, as some of today’s leading AI thought leaders painted vivid pictures of the near-term future of artificial intelligence:
- Intuit. Desi Gosby, vice president of identity at Intuit, highlighted the results of years of work by her and her team with advanced machine learning, including the ability to translate images and characters into an easy-to-use digital experience. The technical challenges they faced when applying computer vision to seeing and reading tax forms was amusing, but the real thrust of her presentation was on the significant progress of next-generation computer vision and how its new high-resolution images will change myriad industries.
- Intel. Gadi Singer, Intel’s vice president of the AI products group and general manager of architecture, shared the company’s excitement about working with cloud providers to grow segments of its business using deep learning. Inference engine, which enables AI systems to deduce new information from specific data sets using minimal computing power, is a key focus, and Intel is seeing the most concentrated use at the device level. With the explosive growth of connected devices, as well as latency issues and bandwidth constraints, Gadi sees inference as the key to expanding the accessibility of valuable AI models.
- IBM. Ruchir Puri shared his insight that CEOs are now driving the AI strategy for most enterprises, which means that discussions with technology vendors around how AI can support enterprise workflows are getting bigger and more complex. He also highlighted AI’s ability to optimize through trial and error; as the AI world converges with Cloud and Big Data, AI is getting more intelligent every day.
- Samsung. Adam Cheyer, now VP of R&D at Samsung, was a co-founder of Siri and Viv Labs. He shared that the way we interact with computers changes every decade. As we approach the eleventh year of having apps on our smartphone devices, he predicts that the new wave is upon us, with AI-powered and highly personalized “smart” assistants that are connected to any and every device. The decade of virtual assistants will bring us a world in which we will no longer need commands, and where our needs will be met by AI-driven suggestions and recommendations based on our behavior and past needs.
- Netflix. Tony Jebara, director of machine learning, explained how Netflix is using AI to personalize and optimize images shown to its more than 145 million subscribers. Each user has a customized experience—with the images of the movie changing depending on the viewer’s demographics and viewing history. Through intensive data analysis and trial and error, Netflix’s machine learning algorithm chooses the image that is most likely to increase viewing.
- Flatiron Institute. In healthcare, Olga Troyanskaya and her team at the Flatiron Institute are focused on developing machine learning methods to address cutting-edge problems in genomics and precision medicine. They developed a computational framework to interpret genomic data from large public datasets to predict cell abnormalities for viral infections related to autism. Olga shared that advancements in computer technology are the key to handling increasing workloads. In her research lab, Olga believes that biggest roadblock to using AI for genomic analysis is the ability to obtain high quality data. She and her research team at the Flatiron Institute believe that the latest techniques in machine learning and AI represent a fundamentally new way to combat this challenge—whether you are researching genomic analysis or discovering a new chemical compound.
Every presentation at the conference made it clear that the latest AI advancements are pushing the boundaries of everything we dreamed AI would be capable of achieving. New AI-based systems are already combing through data 24/7 to complete the tedious analysis and other data-driven tasks that are required to fuel the next wave of innovation. Healthcare. Autonomous vehicles. Precision medicine. Consumer goods. Manufacturing and logistics. The O’Reilly Artificial Intelligence Conference did a spectacular job of not only showing us all what has been achieved in these and myriad other industries to date, but also of illustrating the incredible potential for AI to transform the world of tomorrow. For business leaders and for investors, it was a worthy wake-up call to take action now.
By Lisa Chai, Senior Research Analyst, ROBO Global
Recent White Papers
Apr 1, 2020
In this exclusive report, ROBO Global advisor Louis-Vincent Gave, the ...
Mar 20, 2020
Several trends emerge as we review the close of 2019 and earnings for ...
- 3D Printing
- Additive Manufacturing
- Advisor Spotlight
- Advisory Committee
- Companion Robots
- Deep Learning
- ESG investing
- ESG policy
- Investment strategy
- life sciences
- Machine Learning
- Machine Vision
- Market Commentary
- Precision Agriculture
- Robo Global Index News
- Robotics & AI Investing
- Spotlight Article