NewCo Shift Forum 2018 — Signal P&G Talks
Jana Eggers on why AI shouldn’t be feared, and must be understood.
Jana Eggers, CEO of AI startup Nara Logics, wants everyone to understand AI. In her five minute Signal P&G talk at Shift Forum, she explains that in the end, it’s just math and a lot of data — and that data increasingly will be coming from businesses of all stripes, not just the big tech platforms.
Jana Eggers: Wow! What a day it has been. I’m nervous to stand up here in front of you all because of all the greatness that has been shared on the stage. Thank you for taking a few minutes with me.
I love the title of the conference, which is, Business Must Lead. I decided to take that lead for leading with AI when you aren’t one of the big guys in tech. [laughs]
Just a quick introduction, I’ve actually been working in and around AI for about 30 years. I was a research scientist out at Los Alamos at the beginning of my career, working with neural nets and genetic algorithms for some optimization functions. That’s on your left-hand side. It’s about the perception, categorization, and optimization about what’s going on.
What we’re doing now with Nara is working on the contextualization of relationships between things and the prediction to understand that cause and effect. Most of AI — as this is from Yann LeCun from Facebook points out — is focused in that blue stage.
We’re starting to move. The next area of AI is into that yellow stage. That just gives you a little idea of what we at Nara Logics do and what I’m excited about.
I want to give you a few tips about how you can lead as AI. The first is to stop thinking AI is magic. [laughs] That was in a plant earlier, that AI was called “magic.” Stop thinking of it as killer robots, please.
You can just start thinking of it as gobbledygook, as phrased by our Supreme Court, if you remember that. Recently, they said math is too much of gobbledygook.
What’s interesting here is that cause function is logistic regression. I’m not going to ask anyone afterwards to explain these equations. You’ll notice a lot of similarity between a traditional statistical approach right there on logistic regression and a neural network. It’s just more computation. That’s it.
It sounds like gobbledygook. Simply put, it’s math with more equations and computation going on. That’s all it is. We’ve been doing this for a long time. Logistic regression is way older than me, and I’m old. It’s just adding more layers to that.
I like to use the analogy of thinking of AI like artificial light. Artificial light didn’t replace the sun, in case you didn’t realize. It did enable a lot of great things. It enabled us to explore new areas. It enabled us to optimize others. It also caused us to work late at night, which isn’t always a good thing.
We need to enter this realizing, think about all the great things that can be done, and what are the bad things that could happen. We need to go into this thinking about those things. Go into that with the light behind us. I hope you can use that analogy for yourself as you think about AI, not magic but like artificial light.
The second thing is look beyond the hype. Everybody knows that Google and Facebook are way ahead of everybody else on AI. Everybody knows because that’s the hype. I’m going to show you — and some of this has been shown before — they’re also behind some of the problems. This can be seen today on Amazon. They’re advertising a baseball mat and a mask.
There was an article after the last terrorist bombing in London showing, if you went on and searched for some of the materials, you would find the other materials that were used for that. These are some of the best people in AI. It’s not about them. Just understand, there’s a lot of hype about how far ahead they are.
That’s a lot of hype. They have problems, too. Microsoft 8 is another example. Google, by the way, that’s from two years ago, the gorilla example, two years ago. It’s still “Wired” just covered a few weeks ago that that’s still not fixed.
This, on the other hand, was brought to you by a brick and mortar company, Starbucks. Look at the results of their bottom line. This is amazing, their rewards application that they built. It allows them to deliver real-time personalization. It’s now over 13 million. At the time of those results, it was about 10 million.
They went from having a two-week lag in what someone bought to personalize for them and only having 30 email templates for personalization, to doing real-time in their app for over 10 million users. They did that in a year. Don’t tell me you can’t do it. You can.
Those 18 percent that are in their app deliver 36 percent of the revenue. They’re worth double anyone else. That’s amazing. That was done, by the way, by a brick and mortar company, an old brick-and-mortar company. Don’t get caught up in the hype.
Third is decide what’s core to you. This was done by an even older company, P&G, 180 years old. Amazing company I’m proud to work with.
This is Olay Skin Advisor that Mark mentioned. What was exciting when they came to us to talk about it? They actually came to me and said, “Hey, we’ve decided that we need some AI, but we’ve decided that some of it is core.” That core is the skin diagnostics.
“We need someone to be able to take a picture of their face, guess their age, and understand where the focus areas are — the place in red — and their best areas because we want to highlight that to them as well.” That’s core to us. Knowing how to diagnose someone’s skin from a picture is core. Knowing how to make the best recommendations off of that diagnostics isn’t core.
That’s where we worked with them. We get over 90 percent unique recommendations on tens of thousands of people every single week. That’s, by the way, on about 100 products. 93 percent on tens of thousands of people still get a unique recommendation to them. There’s a lot you can do with small data as well as big data.
There’s a whole bunch of variants to this. We’re worldwide with them now. They’ve got over 20 variants of the Olay Skin Advisor tailored to the market that they’re going into. It’s increased their engagement with their target market. I can’t release all of that, they can. Ask Mark about it.
Your core is unlikely to be AI research. When people tell me, “You know what? We have to hire the best and the brightest in AI,” I tell them, “That’s actually not true.” I tell them, “You just have to hire great software engineers.”
There are things that are different about great software engineers, but I got to go faster forth. It takes a company. Unilever was brought up. They came out in October of last year saying they built an insights engine. I was excited about that because I thought, “Oh, they’re going to talk about technology here.” It wasn’t at all about technology. It was all about their organization.
You’ve heard, “Software eats the world.” AI actually feeds software, not eats it. Data feeds AI. Your organization feeds that data. Don’t forget that. That’s how you create your wow. It’s from your organization because that’s the virtuous cycle in AI. I see some pictures being taken.
Lastly, get ready to learn. I knew I’d be out of time, so I’m just going to give you two books as reference. Marty Cagan is the best at how to build tech products. That’s what you’re all going to be doing now. It’s a very old book. I shouldn’t say that old.
Jana: It’s younger than me. Peter Senge and “The Fifth Discipline” is all about becoming a learning organization. I believe that’s what AI is going to give all of us, is that ability to be learning organizations.
Ram Charan says, “Algorithms are the single greatest instrument of change.” This is the same reason why Adobe is excited about AI. I’m excited about it. You have the ability to use an instrument of change to increase your company value and increase your longevity. I’m going to ask you to become involved in AI.
Right now, it’s still a lot of us computer scientists. I’m a mathematician, so I have to throw mathematicians up there, too, in dealing with our data and compute power. We need more people to get involved. That includes you and, by the way, people not like you. That’s my request of you.
Thank you for the time.