In the second installment of its Rockstar Conversations series, C2C welcomed Andrew Moore to discuss with a global audience his thoughts on accelerating transformation with analytics, artificial intelligence (AI), and machine learning.
Those in attendance were in for a treat, as there isn’t anyone more passionate about this subject than Moore, who worked at Google from 2006 to 2014, took a break to enter academia, and ultimately came back to Google as the head of artificial intelligence (AI) at Google Cloud in 2018.
“I think for a lot of us who've gone through both academia as well as the business side of things, we've really spent our careers wanting to show how advanced automation can improve the human condition,” Moore said. “Those of us lucky enough to be in these roles transitioning from AI theory to actual deployed systems really enjoy that aspect of life.”
Here are five key takeaways from the very lively conversation that took place on the C2C Rockstar stage.
- Advanced Automation Can Improve Our Lives Professionally and Personally
With his advanced background, Moore has a very keen sense as to where AI technology can go in the future and what it is capable of accomplishing in the present. “One of the big AI trends right now is the ability to understand human utterances with 99.9% precision,” he said. This means that we can take industries where automation is encoded in unstructured documents or images and start to structure them so the big databases of structured information, which run traditional enterprises, now get enriched with much more additional information.
Moore noted that in the past, humans spent a lot of time and money analyzing all of these unstructured pieces of information. However, we currently live in an era where “it is now possible to instead turn that into something where all of the boring extraction stuff happens automatically.” With advanced automation, we can focus into the nuances of information, while also stepping back and looking at the complete picture, all along making better decisions.
- Know Your Data—All of It!
When I first spoke to Moore in preparation for his Rockstar Conversation with C2C, he discussed his passion around democratizing AI and making it a tool that anyone can help build. Some people in the audience, however, noted that the real challenge of AI isn’t the training or the model, but rather the data cleaning and filtering out of the bias that may go into it. They asked, “What is the next great evolution we can expect to help speed up these time-consuming tasks?”
Moore really enjoyed this question and warned that he may get a bit philosophical with his response. “We don't know exactly what's going on in the world right now; we’re all working off of partial information,” he said. “The same is true for the computers using AI as well. Data simply reduces the uncertainty—whether in the world or in a computer.”
Moore cautioned that you shouldn’t focus on retrieving data from one simple spreadsheet, but rather from as many sources as possible. “If you want to understand a problem, there's no single source or single beautiful spreadsheet to work with.” He added, “It is the job of a good data scientist or good AI system to combine all these partial pieces of information to reduce the uncertainty and to get a full picture of what's going on.”
- How to Start and Manage Your AI Project Like a Pro
During the Next OnAir Talks series on Cloud AI, we heard a lot of customers asking for insight into starting AI projects. Moore addressed this during the C2C Rockstar Conversation and noted that the very first thing you need to do is discuss whether the experience you're trying to accomplish is for your business or the customer. To answer this, you must be able to uncover what is most important to the customer and what business efficiencies you're trying to accomplish.
Beyond just looking at the outcome experience, Moore noted you need to also understand how AI affects everyone involved. “You got to ask who’s currently affected both in your organization as well as existing customers and third-party stakeholders. Begin with the problem, then look at how it would have to change an organization and be ready for that.”
With the above questions answered, Moore suggests being mindful of things like regulations in your industry as well as potential technology blockers. “Get a senior-level sponsor early on,” he said. “It’s really important to have the hearts and minds of high-level sponsors who can help move that change management forward.”
- How to Determine Useful vs. Wasteful AI
Moore is passionate about making sure technology is useful to not only businesses, but also to society at large, and he says there are ways of identifying what is useful versus what is wasteful AI. “You have to have measurements to know whether your AI system is succeeding or not,” he insisted. “Getting those metrics into place is the most important part of this.”
He summarized it by describing advertising in the early days of AI where most businesses used the technology to predict if a customer would click on a link that was presented to them. The intention should be different today as AI has evolved. He recommended training systems on more long-term goals now, focusing on whether your organization's audience will make a purchase and become repeat customers.
Moore describes the importance of solving for a specific business problem rather than trying to predict a specific action. “Whether metrics of success are simply the application accuracy or regression accuracy of the underlying machine learning tells me that we are just trying to be good at predicting a specific thing, rather than solving an end-to-end business problem.”
- When Done Right, AI Is Good for Everyone
Although Moore is widely known for his work in computer science, he is also very committed to addressing problems that affect everyone. The last time he and I talked, he told me that technologists need to do their part and actually help make our planet better for our children and grandchildren. During the C2C Rockstar Conversation, he noted his concern with the current carbon emissions and energy consumption in the world and explained how Google Cloud has been heavily investing in low-energy data centers to eventually become carbon neutral.
One way Google Cloud is accomplishing this is through deep reinforcement learning on Google Cloud. “With this technology, we learned how to make data centers autonomously control themselves to use less energy,” he said. “That worked out well enough that we now have a product on the Google Cloud Platform for building energy management, which in general is able to work outside the world of just data centers, but for other clusters of buildings, too.”
As we look to technology to help solve not only business problems, but societal problems as well, Andrew Moore and his team at Google Cloud are taking the steps necessary to make analytics, AI, and machine learning tools we can use. He’s approaching it from all angles, deploying his extensive knowledge in the space, and working with others for a better future.
Make sure to register for the next Rockstar Conversation with Urs Hölzle, taking place on Oct. 8. Register here. Don't delay—space is limited.