“Aiera, Why Do You Like Amazon Shares?”

Discussion in 'Wall St. News' started by ajacobson, Dec 2, 2017.

  1. AI stock picking - Interesting.

    [​IMG]
    PHOTO: PATRICK JAMES MILLER FOR BARRON'S
    Ken Sena has become quite popular with the robot crowd since September, when the Wells Fargo Securities analyst introduced an artificial-intelligence system called Aiera that issues Buy and Sell calls on nearly 550 stocks. Aiera uses the machine-learning technology known as a neural network, which also makes our electronics talkative and our cars autonomous. With AI busting out all over, we caught up with Aiera and Sena to find out how machine learning is working on Wall Street and at the companies he recommends, like Amazon.com, Alphabet, and Alibaba.

    Barron’s: What led you to machine learning?

    Sena: A few years ago, we started inviting people from industry to help us understand AI. I met great people, including Bryan Healey, who is the head of AI for the online travel business called Lola. Bryan had worked on Amazon’s digital assistant, Alexa, as one of the original team members to get it off the ground. I asked him if he could automate what I do. He said I’d have to explain to him what I do.

    I told him a little bit about my role, and he explained some of the products and services that I would use to try to replicate it. Just a thought experiment, without a clear idea where it could go.

    Next thing you knew, you were talking to Aiera, the “artificially intelligent equity research analyst.” How does Aiera work?

    Very, very broad stroke—it is a function of algorithms and data. Algorithms are rules for how the software interprets and acts on the data. But the “algos” in machine learning act differently than the algos we’re used to—these algos are not necessarily determined by human coders. We feed these algorithms the data and the algorithms start to act differently. You are letting your hands off the wheel and seeing where it will go.


    What data sources does Aiera use?

    Over 1,000 data sources get pulled into Aiera. So she watches articles written by Barron’s and others. Also Twitter feeds and Facebook. Ultimately, she starts to understand what sentences, what articles, can move a stock. Using this type of technology, a certain amount of media bias can be eliminated.

    Media bias?

    Aiera will learn from experience that, perhaps, a story in Barron’s has additional credibility with investors, because the stock moves accordingly. But she decides. It is not something that we tell her. Aiera has insights on some media that you or I might recognize, for example: “Almost all the stories they write about this particular stock or industry are negative!”

    How current are the feeds from Twitter, Facebook, and such?

    It is continuous. It’s 24/7. You have thousands of media sources and a half-million pieces of media information fed into Aiera on any given day.

    What investment was required to build Aiera?

    The cost in both time and money is surprisingly little, given the capability. That’s an important takeaway for investors to get their heads around.

    [​IMG]
    Wells Fargo says that Aiera’s reports aren’t investment recommendations and may differ from the opinions of the firm’s research analysts. ILLUSTRATION: AIERA, WELLS FARGO SECURITIES
    We’ve been working on it for roughly six months, and we’ve been learning about it over that time, too. My goal in the beginning was simply to be able to explain to investors how you might automate the service we provide. So I was surprised that, by the time I wrote about the project in September, Aiera was already writing her own reports and providing her own predictions.

    Then it became a function of how good were her predictions, and how did we see that improving. The validity tests show that she continues to get better. So, I’d say, stay tuned.

    Does she have designs on your job?

    Every week there is some new development, where Aiera can do something that she couldn’t the previous week. She is starting to recognize the investment relevance when she sees mention of measures like “beta” or “operating margin,” for instance. At what point does Aiera start to provide better services than I do? I don’t see that coming anytime soon.

    To the extent that I’m making predictions about stocks that go up or down over a short time period, Aiera might be in a better position to assess whether or not a particular stock is oversold or overbought. But I would still say that my ability to lay out a fundamental idea to a client and look further over the horizon will continue to have an advantage over Aiera for some time.

    It is not necessarily about wholesale automation or replacement. It is really much more about enhancement and how do you use data science to do the job better than was done in the past.

    You report Aiera’s Buy and Sell recommendations, but Wells Fargo makes sure you surround what she says with disclaimers that say it’s just for education. Ken Sena’s recommendations are the official recommendations, right?

    Correct. We report Aiera’s predictions alongside of ours because they give people an interesting sense of whether or not her calls are improving. At some point, my work and Aiera’s will dovetail. Until we have conviction that Aiera is providing advice that somehow trumps mine, my calls remain the established ones. Ultimately, we hope this technology will enhance our work as analysts.

    Will machine learning affect any industries that your colleagues cover? Is it going to transform Detroit or insurance underwriting?

    Yeah, it will have a role in pretty much all of the industries we follow. It is harder to think about industries that it won’t have some effect on. We’ve heard from Google and from major universities that AI could represent an opportunity that’s bigger than mobile and maybe bigger than the internet itself.

    Thinking about which companies have a leg up, it will be a question of: A) Are they quick to apply and embrace this new capability? B) Do they have data sets that are useful and proprietary? And C) Do they have access to computing?

    It also raises social questions, because we are talking about a technology that has a tremendous capability for automation. That will affect society.

    Unemployment?

    Unemployment, wage growth…

    You obviously think that machine learning and AI will help the likes of Amazon, but how does your study of the technology change your discounted cash flow forecasts and target prices for the stocks?

    It falls into two buckets. One bucket is what they do themselves. A faster pace of innovation and growth will become possible for scaled players who have the data–particularly data close to the consumer–and an ability to drive those data signals into their operations and supply chains for greater efficiency.

    The second bucket is the services they offer to others. Companies like Amazon can turn their leadership in cloud services into leadership in AI services like speech recognition and translation.

    So which companies do you talk to your clients about?

    We usually lead with Amazon [ticker: AMZN], Alphabet [GOOGL] and Alibaba Group Holding [BABA]. Oh, and Tencent Holdings [700.Hong Kong].

    OK, why Amazon?

    Because when you are looking at data collection, computing, and an understanding of data science, Amazon is at the top of the list. Amazon is applying AI across its retail experience and cloud-services platform.

    [​IMG]
    Their cloud business is a bridge to many industries that will find themselves changed by AI, so you could see Amazon’s addressable market opening into new industries—allowing them to sustain their growth rate over a long period of time. If we go back five years and look at our Amazon models, what we were expecting was much lower than where they are growing today. You could say the same about Google and Alibaba.

    With Amazon trading around $1,176 today, your price target is what?

    It is $1,430 over a 12- to 18-month period.

    And at Alphabet’s Google, what are the market opportunities?

    Google was very early in understanding the science. If you are searching for a tennis racket from your desktop at work, they are going to send you to pages with that research. But if you are searching from your mobile device on the weekend, they are going to send you to a store where you can purchase it and tell you the quickest way to get there.

    Google arguably leads both industry and academia in understanding machine learning, from the algorithms to the infrastructure and computer hardware. You could see Google extending AI into other businesses, outside of search.

    What other kind of opportunities?

    Autonomous driving is one. Health care— that’s another industry where Google has made some serious investments. If we size the amount that gets spent annually on advertising within those two industries, we get somewhere in the neighborhood of about $12 billion to $14 billion. But if you look at the size of those two industries themselves, you are talking about several trillions of dollars.

    If Google can use this science to come up with solutions for industries that are more efficient than what exists currently—even if they don’t necessarily have the same share of these markets that they do in search—you can see there is an opportunity for them that might actually be bigger than what they could get from just continuing to focus on the ads themselves. Investors don’t credit Google for those opportunities and are maybe even discounting Google for its research investments.

    With Alphabet shares at $1,036, how high could they go?

    My target is $1,275.

    And how is Alibaba exploiting these technologies?

    What is so interesting about Alibaba is they take in data signals through search—like Google—but also from transactions—like Amazon. And with Ant Financial, they have the largest mobile payment platform in China, which is like PayPal. Then they also have China’s largest cloud-services platform and the leading business-to-business trading platform. So they have nearly a half-billion shoppers and one million cloud customers. Data scale and compute efficiency are the two critical ingredients in this neural-network-backed computer era, so we see Alibaba as potentially best positioned among our coverage universe. China will represent nearly half of the estimated $15 trillion impact that artificial intelligence is expected to have over 10 to 15 years.

    Alibaba stock is now $177. What do AI opportunities make it worth?

    I’m targeting $225.

    You mentioned Tencent.

    Mobile data are important drivers of how AI platforms understand the individual. In China, about 60% of mobile time spent is on Tencent. The opportunity to provide a bridge between businesses and customers, then use a platform like Tencent’s for customer support—is great for them. They also have the largest game platform in the world. AI technology can help learn where the value exists within a particular game content. That will ultimately make Tencent more efficient in how they go after the gaming opportunity and extend those franchises globally.

    So, with Tencent now at 398 Hong Kong dollars, what is your price target?

    HK$470.

    Our thanks to you, and to Aiera.