CES-G 2020 Keynote | Artificial Intelligence: Machine Learning and Mission | Chad Steelberg, CEO

Thank you, Chad. The cloud has transformed several things. Solutions now move from a world where you back up in case it goes down to a world where it's distributed and replicated across multiple data centers so it doesn't go down. Second, the cloud has transformed the capabilities of those solutions. Hardware and software are now bound together in a way that didn't happen before to support the computing power. I'll provide Microsoft's perspective. You can see the stats here: it's reach and depth at the same time. There's a data center complex in the US for Microsoft that isn't so much a building as a small town where all the buildings are data centers. That gives you a sense of the elastic capacity to support all those high-end capabilities Chad's talking about. But with that comes a real responsibility to get security and compliance right. We have controls and processes to support more than 90 compliance standards worldwide. In the US, critical ones include DoD DISA up to level five, FedRAMP, ITAR, and the senior security policy from the FBI. This is critically important in our work with Veritone, especially regarding criminal justice information. Public safety agencies and federal agencies get trust among everyone because it's a shared accountability. That brings us together: unlocking value in data that already exists in systems for public safety, defense, and intelligence agencies—data that is in some ways laying dormant. How do you unlock the value by applying cognitive capabilities like Veritone's? It goes from locking data away to unlocking it. But I want to reinforce what Chad said: it's not to take the person out; it's to make the person more productive, more effective, more efficient. Chad put it as from in the loop where they're doing the work to on the loop where they're guiding the work of the AI, and then... privacy, accountability, transparency, reliability, inclusiveness, fairness—these underpin what gets done in AI. The power of bringing together Veritone and Microsoft is this idea: the cloud started as hardware running your solutions in somebody else's data center. That's a great first place. But what happened after that unlocks all this power: the cloud is now a neutral common platform to bring stakeholders together. It is now bringing its own unique capabilities that can only be best delivered through the cloud at scale: artificial intelligence, machine learning. When you bring those together, you get something very powerful, as you see on the screen. Instead of having data locked in a digital drawer, it's now a shared asset driving better outcomes more quickly. The result of unlocking the power of the cloud, delivering new capabilities with security and compliance, is safer communities, citizens, and responders, and a nation.

Thanks a lot. It's always a pleasure being at this conference with this constituency. There are only three things—and this is a huge topic we could spend days on. What we're trying to communicate is it's very difficult to imagine AI being brought into production for the military-industrial complex by one particular company or one particular offering. You see some incredible corporations on stage. The three things I want to talk about: we take our custodianship of that data for the government incredibly seriously. So I want to talk about data, security, and cloud. If you're going to bring AI to a meaningful place in this marketplace, you need to be able to feed the cognitive engine with a vast amount of data. We're entering the third decade of the 21st century; we no longer measure data latency in milliseconds; we are in the single-digit... decades learning the domain. We have systematically provided the data that analysts need, giving it to them because they have a need to know. It's taken them a long time to build their experience set. We are now able to support analysts by having a cognitive engine run across the vastness of data in any format: spatial, relational, financial, human intelligence—you name it. The cognitive engine can go through that, and we just have to worry about the integrity of that data. Bad actors are rapidly evolving; they would do nothing but change our algorithms so that we make the wrong decisions in a decision support structure. We see that a lot right now coming out of Russia—they are manipulating the data so we have a different decision support matrix, disrupting a lot of what we're doing. The security of that data, where it resides, and how to expose it to AI engines like the powerful one Veritone has developed means it may not be economically feasible to put all that data into a single cloud and then try to cross... There's no one who doubts that AI will be able to run on-prem in a legacy contained environment, on a TS Enclave where only 15 people have access to the data. That's not going to work. But if we're going to get to actionable artificial intelligence—the term we use on my team, A2I—that's where I'm really interested in getting to. So the data to feed the engine needs to be vast, super fast, incredibly secure, and needs access to the cloud. Cloud-adjacent strategies allow us with our technology base to take all of the data... every day is to do nothing but improve the decision support for the military-industrial complex that I'm personally responsible for from the infrastructure side. Our interconnect with Azure means that Veritone running in Azure, connected to a cloud-adjacent architecture where we put petabytes of data—structured, unstructured, Hadoop data—means customers will get a lot faster answers. Thank you very much for the opportunity. I look forward to working with any and all of you as we move forward. Chad, this has been a real pleasure. Thank you.

Thank you, Chad. Good morning. I'm Patrick Clark from Deloitte. We focus on helping our clients solve their toughest problems, and we do that by partnering with elite technology companies, many if not all represented in this room. We've heard several instances this morning: not only is data growing in size, but it's growing in complexity. There is simply not the capacity within the investigations and discovery world I'm supporting our clients in to work with that data in the traditional manner. We need to call through that data in a traditional manner. For instance, we were supporting a financial crime investigation and were presented with over 20,000 foreign language emails. As was referenced this morning, you have no idea if those emails are birthday invitations or critical mission information. In the traditional environment, you would have had to either put those emails aside and risk losing important information, or go through the natural foreign language. Instead, we partnered with Veritone in this instance and within hours had that information translated, then were able to run analytics. Supporting the Department of Justice, we partnered with Chad and the Veritone team. We recently rolled out support to all 94 districts of the EOUSA attorneys, allowing them access to this technology. Previously, in discovery and investigations, we were worried much more about the written word—structured and unstructured emails. Today, audio, video, social media data are all coming into play, and you simply cannot work through that in the traditional manner. The final point I want to make is that all this requires a secure environment. Everyone here probably knows that the FedRAMP process... the first thing we got to do is get FedRAMP certified. I was impressed that your team made that investment, Chad. I know it's going to serve government clients. Thank you. So the Department of Justice is now connected to a FedRAMP instance of AI in partnership with Deloitte, and we're actually processing cases for them now.