Adelyn, Marlene, and I had the honor of being invited to speak at CES 2018 about our book, Applied Artificial Intelligence: A Handbook For Business Leaders. Other authors featured in our cohort this year included John Grisham and Stephen Wolfram. In the past, Deepak Chopra and Alexis Ohanian have spoken on stage about their work.
Geez, no pressure…
We were extremely lucky to have Cindy Stevens, Senior Director of Publications for the Consumer Technology Association (CTA), be our interviewer. She and the rest of the CTA team made what could have been a high-pressure situation feel streamlined and relaxed. We’re excited to present our CES 2018 book interview along with the transcript of the event below:
CS (Cindy Stevens): Welcome to Gary’s book club at CES 2018. I’m Cindy Stevens, the Senior Director for CTA. Today, we’re going to talk about artificial intelligence and how it is being integrated into product across the CES show floor. I know you have already seen a lot of that already. AI promises to improve our lives tremendously but many of you may be wondering what AI means for your business.
The book Applied Artificial Intelligence is a guide for business leaders who want to leverage machine intelligence to enhance the productivity of their organizations. This book is designed to help business executives drive innovation by combining data, technology, design and people to solve problems.
The two women who wrote this practical guide work at TOPBOTS, a strategy and development firm specializing in applying artificial intelligence and machine learning to enterprise challenges. Adelyn Zhou is the Chief Marketing Officer and Mariya Yao is the Chief Technology Officer and Head of R&D. TOPBOTS guides Fortune 500 companies and executives on AI trends, strategy and adoption. Please welcome, Adelyn and Mariya.
MY (Mariya Yao): Thank you so much for the introduction, Cindy.
CS: Sure, my pleasure. Let’s just dive right in. So, why did you write this book?
AZ (Adelyn Zhou): We found through our work that there’s a lot of misconceptions about artificial intelligence. First of all, there are either books that are super technical that detail natural language processing algorithms, and other books that are very futurist, focused on making people feel that the world’s coming to an end. There is nothing practical out there about how you, as a business executive, can actually implement artificial intelligence and machine learning into your business processes. We’ve worked with many businesses to do exactly this, and we distilled out what we’ve learned into this handbook with frameworks and structures. We want to help business executives take advantage of all the powers of machine learning – to increase revenue, decrease cost, drive productivity and make their businesses better.
MY: We cut the noise. There’s a lot of junk out there about artificial intelligence, but our book is focused on being a down-to-earth, practical how-to guide on actually applying modern machine learning techniques to your organization, to solve real problems.
CS: And how would you define artificial intelligence for the group here?
MY: AI is a huge umbrella term, much abused in the media, and that’s why we spend two entire chapters of our book just to clarify what is artificial intelligence. On a high-level, AI refers to a set of techniques within computer science that are used to either replicate or even exceed human decision-making and human perception. These can include techniques like data science, data mining, symbolic and expert systems, machine learning, deep learning, evolutionary strategies, and the list goes on. We dedicate an entire chapter creating a friendly non-technical introduction so that a business executive can understand how these technologies differ and what they’re good for and what they’re bad for.
AZ: And there’s so much misinformation out there. As Mariya mentioned, we estimate that 90% of the media headlines out there about AI are actually fake news.
CS: Right. There are scare tactics with artificial intelligence.
AZ: A lot of clickbait trying to get people with sensational headlines. For example, they talk about how there will be no need for doctors anymore. We don’t believe that since we still definitely need dermatologists, cardiologists, and other medical experts to interpret the outputs of artificial intelligence. We absolutely still need a lot of the jobs and people that we talk about today.
CS: Right. So some of the misconceptions might be like robots are going to be embedded with AI, and they’re going to take over.
MY: Oh, the whole killer-robots thing. You don’t need to worry about it despite what Elon Musk might be saying.
CS: Right. That’s good to know. And so, what are some of the practical steps an executive might take to incorporate AI into their business model?
MY: There are two layers that need to happen if you want to transform your organization into a fully AI ready organization. The first layer is leadership. Things like educating your executives, pulling together budgets, political buy-ins from all your stakeholders. The stakeholders can be anywhere from your frontline employees to a lot of middle managers who might be resistant to the idea of AI “taking over their jobs.” We spend time in the book discussing, from a strategic level, how to prepare your organization from a leadership standpoint.
The second layer is the technical layer. What we find is that most organizations that are not traditionally technology companies often lack the requisite proprietary data to train useful models. Preparing and collecting these data can be a months-long, years-long procedure so you should get started now. Our book helps you figure out what are the right steps to get you started.
CS: Are there certain businesses that are better suited for incorporating AI?
MY: Absolutely. One of the reasons why technology companies have been at the forefront of AI is when we use Facebook, they track every single thing you do. If you are selling consumer package to your company, you aren’t puting sensors in your shampoo. You don’t actually know what the user is actually doing at that granular level. Companies that are naturally collecting a lot of data can then feed this data into unique machine learning models. Many companies, however, are even missing the data capture and the data collection procedures to get the kind of data they need. Therefore, those types of companies tend to be a little bit further behind.
AZ: It’s not too late for them to start. If you are aware of this now, then you can start the processes to collect the data, process the data in the right ways, and to store it in the right methods. Data is the new oil. You need the data in order to do machine learning. Furthermore, today, companies can start hiring a team, building a team around that skillset, and really honing expertise within their business and bringing the strategic focus on why they’re actually using AI to solve.
CS: Right. I’ve also heard the term, data is the new currency. Do you have any case studies of businesses that have successfully incorporated AI that you could talk about?
AZ: So one of the most common areas of application of AI is in customer service. I’m sure everyone here has had a very frustrating time when they called customer service and got stuck in and endless loop. With artificial intelligence, you’re able to now create virtual agents that are better suited to answering questions. Virtual agents that can alleviate cost center pressures, especially when you have huge volume spikes and and calls. This can help you achieve higher customer satisfaction and better customer service results.
MY: There are AI applications for every enterprise function as long as you have the data and the technology infrastructure. There’s probably some way to apply machine intelligence to improve whatever your current business process is. This falls into three different categories:
The first is automation. Tasks that are really low-level that take a human one second to do can be automated in the near term. The second is augmentation. For example, there are some AI systems that can give very accurate medical diagnoses and can support doctors and augment their decision making capabilities for disease identification and treatment planning. The third is completely new functionality.
We made a couple of major breakthroughs in AI in the last few years. We can train machine learning systems that can identify images and classify objects and images to about human parity. We can do the same for speech recognition and also text-to-speech. That enables a whole frontier of new functions that we really weren’t able to do before.
CS: So you see AI as augmenting what humans can do and just expanding creativity and efficiencies.
MY: The biggest misconception is that we’re really close to superhuman AI. We are really far from superhuman AI and so people are very often afraid – “Am I going to lose my job if we adopt this AI system?” Don’t worry about it. Systems that we can use right now can improve your predictive abilities but today, we simply don’t have artificial intelligence that’s at the human level, that can actually replace the kind of strategic and creative thinking that you need an executive or an expert for.
AZ: Many of the executives we work with start using AI for augmentation for different business processes. They don’t replace their workers, but rather re-align workers to do more higher level, interesting types of work. For example, someone in a customer care center, instead of resetting passwords all day, they can help customers with their banking finance or give them more higher-level strategic advice.
CS: How expensive is it for a relatively small business to get involved with AI?
MY: That depends. I mean, how expensive is a car? It can be anywhere from a few thousand dollars to very, very expensive. Especially some of the cars that you see here at CES.
There are a couple of ways to adopt AI. If you want to do AI at the level of Google, Facebook, some of these major technology companies, that’s a very expensive multi-billion dollar procedure to transform your organization from a leadership and technical level to do original AI research and original AI implementation. A small business is very likely to find a third-party vendor that has an AI solution that targets a very specific use case. It’s very unlikely that, as a small to medium-sized business, you’d be able to hire the requisite talent to design your organization and implement all of the processes that major tech companies use.
Depending on the organization, we recommend a blend of techniques. Perhaps when you are doing pilots, find the vendor closest to the use case that you want to experiment with and see if you can deliver ROI using a strategic partner. In the long run, we strongly recommend that if you want to stay competitive that you have to become a leading technology company. There’s just no way around it because automobile companies, financial services companies, even consumer companies are all fighting to be more technologically capable in order to stay competitive with Facebook, Google and leading technology companies.
CS: Right. So it’s critical that the companies start using their data as effectively as possible.
Audience Question: Where are we today with the data?
MY: Current model machine learning approaches require you to have a lot of data and what organizations need to do, prior to implementing AI, is to have big data infrastructure in place like a centralized data lake, centralized data processes, or often a center of excellence that manages data as well as provides leadership. This means having a Chief Data Officer or Chief Analytics Officer. Those roles typically need to be in place before an organization can use AI to any meaningful level.
What usually happens in an organization is that you have these data silos with little pieces of data sitting everywhere. It’s a giant mess, and no one even knows what the data means or how to process it. There’s a gap between people such as the business owners who want to extrapolate business insights and the actual data stewards which would be the engineers that are actually coding the data pipelines. It’s a multi-year process that many enterprises have to go through before they can really implement AI at a strategic level. Once you have a centralized data repository, you can start implementing machine learning at an enterprise scale.
Audience Question: Who are the companies doing interesting work with AI outside of the Googles, Facebooks?
AZ: One of the things with artificial intelligence is you need data. If you’re not one of the large companies such as Google, Facebook, or Alibaba, you’ll need a data source. That naturally forces companies to really be focused in certain niche areas where you can get data. That’s why you have in healthcare companies that are focused only on that industry, where they’re able to get data for their specific niche.
MY: It really depends on the use case. On our website, we do an analysis of all the different AI companies that tackle different enterprise function so HR, customer service, marketing. If you go to TOPBOTS, there’s hundreds of examples. Every year, we publish a review of all the AI companies that service each enterprise function.
We try to shortlist the ones that we think are the most functional, but again, it’s hard to say whether a company is good or bad for you. It really is very personalized and tailored to your business’ particular use cases. If you are looking at a computer vision solution, we really like the company Clarifai. We think they are doing probably the best computer vision outside of the Google, Amazon. Again, it really goes back to your use case. There are other computer vision companies that maybe aren’t as prominent, but really service your use case in a way that will deliver ROI for your business.
It’s hard to say who is the leading AI company overall. The question you should ask is: who is the leading AI company and the right strategic partner for you, your business, your team and your problem.
CS: What advice would you give to a small business that’s looking to get involved with this as sort of a first step? Like, what would they do? Would they hire an expert or what do you recommend?
MY: This is ironic to say, as a small business, sometime it’s cheaper to just hire a person to do that task than to solve it with AI. But I will say that there are just a couple of areas that are no-brainers.
For example, back office accounting. Who wants to do taxes? Nobody. Nobody wants to do taxes, nobody wants to do expense reporting, but everybody has to do it, right? There are a couple companies that are obvious use cases. If you want to automate some of your expense reporting, automate some of your taxes, automate some of your accounting, there are companies for those tasks. For the most part, they service large enterprises, but over time, you’re going to see more of them serve the broad market of small medium-sized businesses.
But there are just some obvious use cases, where it’s a no-brainer to offload them. Some of the these are within marketing. Adelyn’s an expert when it comes AI for marketing.
AZ: If you are starting a business, figure out where you are spending money and resources. Where you are spending a disproportionate amount of time and not getting enough ROI. Often this is in your back office or customer support. That’s why you hear the term chatbot or bots a lot. It’s because they are usually relatively easy to set-up. They may not always be the best, but they are one way to kind of dip your toes into trying to use an automated system. You can even start automating things like your Twitter responses or Facebook Messenger bots.
As Mariya has mentioned, keep an eye on external partners. As a small business, you don’t have all the resources to dedicate to do your own R&D. Many times, it is probably just better, easier and more cost effective to find a third-party vendor that supports and answers the question you are trying to solve.
MY: Just to add on to what Adelyn has mentioned. Conversational AI and conversational interfaces: we’re really excited about them. Every technology you use has graphical user interface. That means buttons, sliders, windows. All of those can be replaced or augmented with a natural user interface, where you use your voice, your text, your body language, your emotions, and other natural behaviors interact with technology, the same way you interact with a very smart and competent human. We see that a very accessible way for some people to understand AI and interact with AI. When you don’t have the technical aptitude to evaluate more complicated data driven approaches, we find even non-technical executives still understand conversational AI. “Oh I see the value of the system.” Or “I can see how I might design a better customer experience by using an automated bot, by using some conversational AI.” That’s a very exciting area.
CS: Where do you see the industry in 5 years?
MY: If you ask experts any questions about AI prediction, the guesses are going to be all over the place. If you even ask technical experts, we don’t really know. I would say from a business standpoint, is that if you don’t start adopting, if you don’t already have a big data strategy, you don’t already have centralized data infrastructure, you’re already behind the curve. And what I can see is that in 5 to 10 years, companies that have not made a sufficient technical transformation at the enterprise level to start adopting machine intelligence and more modern techniques, they’re going to be behind.
AZ: As an organization, regardless if you are B2B or B2C, we’re already experiencing a lot of the benefits of machine learning and AI. Many of the things that are happening behind the scenes that we don’t even think about, everything from our spam filters to our Spotify music playlist, is all being done using machine intelligence. If you’re another music streaming service, and you don’t have that fundamental artificial intelligence, you’re going to be behind. As a business, you absolutely have to start incorporating these into all aspects of your business and it can be anywhere as simple as HR, customer service, to your core functional products.
MY: We’re lucky right now that a lot of leading AI research is being productized, commoditized and made a lot easier to learn. There’s a lot of education to help software engineers become machine learning engineers, to democratize the access to AI both from an executive education standpoint and also from a technological innovation standpoint. We find a lot of the machine learning APIs are available from the big players and from third-party vendors, and many open frameworks allow you to build your own software. They’re becoming easier and easier to use so that some of the AI breakthroughs you read about in research papers are actually accessible to enterprises and, in some cases, small and medium-sized business.
CS: Right. I know a lot of people are using Amazon, you know Alexa, at home. Do you see that interface becoming something that would be used at work, to where people are using digital voice assistance to communicate?
AZ: Absolutely. Even in work, we’re starting to see companies use voice assistants to book conference rooms, to schedule meetings, to do office ordering. The office is a unique space because you’re in an open, especially in an open environment, so it’s a little bit harder because you don’t want to necessarily be talking and distract your colleague next to you. But we see those initial use cases very commonly, and if you have your own individual office, you will see similar types of use cases as in the home.
CS: So we’re getting ready to wrap up. Is there anything else that you all would like to add? Or you know, talk about for the audience?
MY: Right now, AI innovation is concentrated in a very small number of companies. When you see people doing cutting-edge AI research, it’s going be Google and your usual suspects. But if we want AI to benefit all members of society, then we need executives at all industries, at all levels to understand AI and what is it. For them not to fall for media hype and to understand where can they actually apply it to their business and also their communities.
There is so much low hanging fruit on how to use data and use algorithms in a way that’s beneficial for society. But a lot of people who know the problem, who understand the domain, unfortunately don’t know how to use technology to solve these problems. That’s why we wrote this book, to hopefully make it more accessible to non-technical people, to give you an understanding of what does it actually take to implement AI at an enterprise level including the unsexy things we mentioned before. They’re important from an organizational level and we believe that if we do educate artists, lawyers, politicians, business executives, people from all walks of life who want to solve problems with the ability to leverage machine intelligence, we should really be able to transform our society.
CS: That’s fantastic. Is there anything else, Adelyn?
AZ: Don’t be afraid of it. And you are here at CES, so that is one of the things, embrace it, learn about it. There’s just so many opportunities, I think, in the next 5 years, it’s going to be absolutely amazing. There’s so much that’s going to happen.
CS: Oh my gosh. Well, thank you both very much. That was very inspiring and very educational, so I appreciate you all.
AZ and MY: Thank you for having us.
[cp_modal display=”inline” id=”cp_id_fdd73″][/cp_modal]