The Commercial Future: Artificial Intelligence

spark-of-lifeIn this blog, I am interviewing guest Nikolai Pereira, an entrepreneur and founding partner of Leova, an AI system that voice-enables mobile apps.  He is an algorithm junkie, computer science and electronics engineer expert with direct knowledge on the future of Big Data and Artificial Intelligence. Let’s begin:

SR: Many of us have a notion of Artificial Intelligence as a giant leap into Science Fiction, a cold, ominous voice echoing in the chambers of a satellite station, omnipresent, omniscient.  What is it exactly? Are we there yet?

NP: I’ve gotta admit, the movies are pretty frightening. The thought of a HAL-like entity watching me coldly, from my ‘smart washing machine’ haunted my younger years. But reality is pretty far removed. This fear originates from our tendency to project our human selves onto machines. As humans, it is our genetic programming that orients and drives us.

I’d like to subject you to an example: Ralph and Macy are co-workers in a company where dating your co-worker is prohibited. Irrespective of that, they may still choose to go out with each other. That’s because office directives are ‘rules’, whereas their genetic directives, to procreate – is part of guiding programming that’s wired into them to ensure survival of species. Unfortunately, this survival of species has led us to where we stand in terms of unmanageable population levels in various parts of the world, uncontrolled exploitation of our planet’s resources as well as climate change.

Machines, however, are driven differently. Just like your iPhone, every program, every algorithm is given a purpose by its designer. I’ve discussed this in great detail in a post on my blog: . The power steering in your car is driven by a singular purpose – to make your steering experience more convenient, within the constraints of safety, et al. At no point in time can it change its ‘mind’ about what it does and become a music system (if that example sounds absurd, it is – but it’s also extremely analogous to the possibility of a machine behaving any way other than the way it was programmed).

So what we’re projecting onto machines is the fear that they might behave like us, and then compete with us for natural resources. If you’re worried about machines spontaneously deciding to do this, then have no fear. If you’re worried about a human being programming them to procreate like us and compete for our resources, well, then you’re onto something. If you’re worried about some hacker breaking into the control system of your car and disabling the brakes, you’re not paranoid. These are real – but they’re not caused by the AI going nuts. But then again, computer viruses have behaved in that exact same way for the past 20 years or so; and we worry less about them every day than we ever have before. Remember, as a species, we’re supremely qualified to survive almost everything that might come our way.

SR: How different is it than Big Data which is a concept that has floated around for a long time?

NP:  Big Data is completely something else; and is related to AI in only a supplementary way. Big Data is the ability to process large amounts of data and then derive some insight from them. Companies have been collecting data about their customers for a long long long time; and it’s only recently that the tools to process and sort this data has emerged into affordability and accessibility. While Big Data offers amazing possibilities, it also offers a disappointing insight; that maybe we’re not as unique as we’d like to think we are; and that our preferences resemble that of many other people. And that a lot of our actions are predictable based on the previous actions of people who could be grouped into our demographic.

For example, when you search for X movie on NetFlix and NetFlix offers you recommendations to watch Y and Z, based on what you’ve watched, that’s big data processing in action. And those recommendations come from NetFlix seeing that other people who’ve watched X also tend to watch Y and Z. And more often than not, NetFlix is usually right – I do want to watch movies Y and Z.

Big Data is being used to empower decision making everywhere. Companies like GoodData and Mu Sigma crunch data to provide their customers with insights from this data. An example of this as a value add to a customer like Target could be to identify buying habits of a particular demographic (such as college students) and then cluster their frequently bought products contiguously in a store so that they don’t forget to buy something. This makes the customer more likely to have higher dollar totals at checkout. Big Data is used to make commercial decisions about many things – and as the ability to crunch big data becomes more commonly available, it will be used by ordinary mortals to make better decisions regarding their daily lives and well-being.

SR: What are commercial applications that we can see in use of AI today?  The other day I was joking that buying a phone will be analogous to interviewing for a personal assistant. Voice, tone, readiness will be as important as accuracy of data.  Right now Siri, Google Now and Cordova lead the pack.  What am I missing?

NP: Actually, right from your washing machine, which makes decisions about how much water to fill in its drum, all the way to ‘smart’ thermostats in the office place to Google Ads being served to you – AI is ubiquitous. Associating it with an interface (such as voice or a blinking red light in the case of HAL9000) is limiting and takes away from how omnipresent ‘smart decisions being made by machines’ are.

As time goes by, and computing power becomes cheaper and more affordable to inventors and scientists, you will see more visible manifestations of AI in our world. A fantastic case of AI showing off its computing chops is in Google’s self-driving car. This car (actually a set of roughly 25 cars) has driven over 1 million autonomous miles across the country, without a single accident that wasn’t caused by external human error. I could argue that if all cars were switched to self-driving technology, with zero human interference, our grand-children may never know a road accident in all their lives.

Right now, the most obvious and in-your-face use of AI would be Google Now, which uses AI, often coupled with big data, to try to predict the rest of your data and show you information related to your schedule. Siri isn’t really AI, except in the most rudimentary sense of the term.

SR: It seems technology is seldom consumed in one serving.  Different parts come together over time to solve a problem.  For example the internet was the marriage of PC, telephone, and browser technology advances over a period of 30 years.  What needs to come together for AI to accelerate user adoption?

NP:  I’d actually say that we’re all already using AI in one way or another. If you’re not consuming it directly, that’s because AI-output, in its current state and form is error-prone, and best consumed and ‘cleaned’ before being sent to the end-user.

SR: Talk to me about your product, Leova and how it can crack the AI riddle?

NP:  Well, Leova is an algorithm that is able to understand naturally spoken language. This is a big step forward, because Leova doesn’t need you to use or memorize keywords; and its ability to manage a conversation is a big big deal. To explain what this means: Leova’s travel implementation allows you to ask for a flight in your first sentence; and then in the second sentence you can change the date of the flight; and then you can change the destination city; and then you can tack on a return flight. We spent a lot of time studying human interactions to arrive at an understanding of how human beings talk to each other before attempting to build a system that is able to handle interactions of that nature. All our work can be boiled down to hard science and our ability to understand spoken human interaction.

SR: It was great talking with you! Good luck.  I am excited about your new algorithm!

Change The Channel

montehallSetting up a new sales or distribution channel is like the Monte Hall problem on the game show, “Let’s Make A Deal.”

There are 3 doors: one of which has a large prize like a Ferrari, two of which have a stubborn goat.

Monte Hall, the host, knows which door has which item and after the contestant selects one, he discloses a goat behind one of the two remaining doors.

The question: Do you stick with the gut of your first choice or do you switch?

The answer: Unless you have fantasies of goat-herding in the mountains, you should always switch doors.  Even if you have been shown one of the doors behind which exists the car,  the probability of success jumps from 1/3 to 2/3 if you switch. (Please research the proof I will not explain it in this piece)

Many technology incumbents struggle with the Monte Hall dilemma and do not change their sales channel even when the odds are favorable for them to do so.

In the case of consumer packaged goods, brands are challenged by retailer private labels and the multiplicity of brands on the counter, blurring options for the consumer.  It seems obvious that brands need to switch from the first door of “retail” they picked and select the other door of “digital”, setting up a new direct channel strategy to get closer to the voice of customer.  A digital add-on to an analog enterprise.  The internet has low overhead plus global distribution.  But many CPG brands fail to commit due to fear of cannibalization, where their traditional dealer network and retailers may reduce order volumes to protect their own revenue streams.

On the other side of the deal, with digital brands like Amazon, setting up a new channel or “picking another door” means the opposite.  They have to front high fixed retail expenses and sacrifice financial agility.  According eMarketer, by the end of 2015, about 7% of sales will come from online shoppers, and 1% will stem from people shopping on mobile devices.  The e-commerce era was an intoxicating time when you didn’t have to visit a store to get what you wanted– but let’s face it– most transactions like cars, clothes and groceries require a human touch.  If a firm is not leveraging the four walls and layout of retail, it is leaving a sizeable chunk of the market on the table.  Interestingly enough, it is rumored Amazon is setting up a physical retail store as a new channel for the 2015 holiday season.

For the longest I have been bullish on digital, calling all businesses emerging and established,  to change the channel to digital.  I have an all-in approach, but Digital is not enough.  Although Digital channels are dominating the conversation of channel thinking–  ‘human-assisted’ channels will always be relevant. When a sales cycle is long and  complex or the majority of consumers prefer call centers and face to face storefronts, the approach has to be Omni-inspired, seamless.

Thus the question most of often is not just which Door has the car behind it, it is where can the car take us? We are scratching the surface in regards to sorting the differences between and within channels.   For example, online customers tend to have different buying habits and pathology than those who are motivated by a print catalog to order online, by phone, or by mail.  A telecom provider located in the heart of a metropolis with a customer base that is interested in on-the-go data products cannot be supported similarly to a telecom provider in a sparsely trafficked rural town interested in money transfers. As mobile becomes more ubiquitous, brands have start to use segmentation to customize their channels to match local needs better, more intelligently.   Otherwise no matter the door they pick or how many times they switch, the prize will be a goat. Epic fail….