Artificial Intelligence: Angel, Demon or Opportunity?

By March 29, 2018Conferences, Stories

A Marketeer view upon the Artificial Intelligence

Recently I was asked to prepare a speech for a conference, regarding the Artificial Intelligence. I am not an engineer or mathematician, however I am a super curious Marketeer excited by the new things.

When a new tech revolution comes up, the first people that get rich and successful are the engineers that invent the new machines. In the same time, in the following waves, even more successful become the ones that know how to use better the machines for human benefits. Think about combustion engines, electricity, electric bulb, telephone, personal computers, telecommunications and so on. On the longer run, the ones that make the difference are not the pure tech guys. But the commercial people that understand the everyday customer and the business potential.

This time will be about a new super-powerful animal: the Artificial Intelligence.

I. What is the Artificial Intelligence?

Well, its roots trace back to 1956, in the dawns of computing. It is not a new invention after all. The objective is to perceive and adapt to the environment and take actions that maximise the success. How to do that? By replicating the Human Cognition, by learning, reasoning, achieving knowledge, planning, understanding natural language, perceiving the environment and controlling objects. Staya Nadella, the Google CEO, just said that the AI is more important than the discovery of electricity. Let’s see.

II. How does an AI learn?

Machine Learning has ties with mathematical optimisation and overlaps statistics. Is dependent to Big Data, clustering and creating algorithms to recognise patterns in mountains of data. Creates complex models of analysis that can lead to predictions, through regression curves. Its powers rely in constant optimisation between of algorithm, based on the fulfilment level of its predictions.

Based on training data and experience, learns itself how to become better. Here lies its weak point: if the training data are not available or have poor quality being biased, the machine will fail (which happens most often). 

III. Artificial Intelligence main public successes:

1. Playing strategic games better than any human. In 1997 IBM Deep Blue managed to defeat chess world champion Gary Kasparov.

Kasparov vs. Deep Blue

AI took a even more complicated game: Go. In 2016, Google AlphaGo defeated the go 18 times world champion Lee Sedol.

Before the games, Lee did not believe that a machine could stand a chance. After the games, Lee recognised that the AI made original moves and made a particular unusual move that no human could do. There is a Netflix documentary about this clash.

Lee Sedol vs. AlphaGo

2. Recognising Images and Human Speech. Microsoft AI error rate just dropped slightly below human rate. In 2014, a chat bot won the Alan Turing Challenge. The Challenge had been set 60 years ago, when Touring bet that a robot could be intelligent enough to trick at least 30% of the humans into thinking that they talk with another person. A bot named Eugen Goostman managed to trick 50% of the jurors.

AI Speech Recognition

AI Speech and Image Recognition

3. Mimics Human Creativity. More than that just understanding the language, an AI can write. In 2015, an AI managed to compose a poem that also passed the Alan Turing Challenge. Other Bots can also paint in different styles, in ways that cannot be distinguished from human counterparts.

AI original Paintings

In 2017, Nutella was the first important commercial brand that used Artificial Intelligence in creating 7 Millions unique labels.

AI Nutella Labels

AI Nutella Labels

4. Uses speech recognition to improve Mental Health. IBM Watson has an interesting approach on speech recognition. Uses its algorithms to analyse speech and written words. They asses the likelihood of developing Parkinson or psychotic events into the future, at a level of confidence equal with trained psychologists. You could not like the idea of a robot being able to track your Social Media posts in order to highlight your schizoid correlation. However, this tools would help doctors detect, prevent and treat Mental Health.

5. Uses image recognition to detect skin cancerStanford University trained an AI, based on 129,000 clinical images. Was able to reach the same confidence level as trained doctors.

6. Self driving cars. Self driving cars are on the streets. Recent Uber accident was not provoked by technology issues. But by the cost cutting decisions to remove some of the car LIDARs.

Uber SELF DRIVING

Uber Self Driving Car, before accident

7. Sophia, first robot citizen. In a global PR stunt, Sophia a robot able to speak and express feelings and sentiments. Bcame a citizen of Saudi Arabia in October 2017.

Sophia

Sophia, first Robo-Citizen

IV. What are the Risks of Artificial Intelligence?

AI is incredible efficient to spot human vulnerabilities. It mimics human brain, but only the Neocortex/ logical part.

An AI does not have Reptilian Brain, which is the centre of fears. Does not either have the Mammalian brain, the source of emotions, values and long term human memory. When a human makes a decision, it is based on the System 1 (intuitive), based on emotions (Mammalian) and minimising risk (Reptilian). The Neocortex just give a veto and validates the automated human response.

That’s why human decision making is biased, doesn’t optimise itself, being anchored into past. The Artificial Intelligence can spot very accurately its issues and capitalise on that.

Human Triune Brain

Nir Eyal, developed the Hook Model, which explains how casinos work, but represent also the model of Facebook and Social Media.

Tracking the needs (e.g. being connected) -> leads into action (e.g. posting) -> have a variable reward that gives adrenaline or doppamine (e.g. likes or appreciations) -> asks for an additional investment (more time spent).

Nir Eyal Hook Model

In this way, Social Media became such an efficient tool for propaganda. Propaganda, as defined in 1930, uses mainly negative emotions, avoids ration and creates rapid movement. A bargain.

AI clusters people into homogenous segments based on deep tensions and fears. Than optimise the message learning how people react. Each new person is assigned to an existing segment. Know how to use that, understand blue collar people deep tensions, start advertise nationalistic messages to this target group which happens to be the largest one, and you get a Brexit or a Trump president.

There is an excellent Ted Talk on AI Dystopia made by Zeynep Tufekci, a techno-sociologist professor at Harvard. She gives the example of how You Tube will get you down the rabbit hole or how Facebook can influence ellections. 

Zeynep Tufekci AI Distopia

Zeynep Tufekci on AI Distopia, TedX

V. Key Ethical Questions (Based on World Economic Forum)

  1. Unemployment. Based on McKinsey, 78% of predictable work and 25% of unpredictable work will end. What will happen than?
  2. Inequality. Even the quality of life improved, the inequality jumped even higher. Humans look at inequality rather than absolute quality of living.
  3. How do machines influence behaviour? An average American spends 40 minutes a day on Facebook.
  4. Artificial Stupidity. Who will be responsible when an AI will make a mistake? Uber just did.
  5. Biased Robots. AI is as good as the training data. When trying to implement an AI to predict criminal behaviour, the AI showed a bias against black people. It inherited the stereotypes of its creators.
  6. Staying in control. AI is virtually a Black Box. Even the engineers cannot understand how it learns and how its optimise its algorithm, being too complicated for human mind. Two Facebook AIs were linked together and they were shut down after they altered the communication language.
  7. What will happen when will be linked to Quantum Computing? AI will be the main client. Once the AI power will grow exponentially, how complex will become?

VI. Commercial Opportunities 

  1. Customer interactions/ conversational interfaces. CRM & Voice Ordering chat bots, can respond personalised, without waiting time. Imagine that you wouldn’t have to wait minutes in line to order your pizza or to get customer help
  2. Individualised Content.  In the giant depots like Amazon, YT, Facebook or Netflix, you would have your unique store, empowered by dynamic shelves. This comes also with a “revenue opportunity” already used by Amazon: dynamic pricing.
  3. Empower doctors to have more efficient health tools. Can lead to cost reduction and improve quality and accuracy of medical services.
  4. Better understand human behaviour. Remain the most important current opportunity. If an organisation understands the needs of its customers, than can create a more efficient message. Later will lead to a customised offer. Up to this moment, political marketing was the best one, given the high emotional involvement.
  5. Improve Decision Making. I see huge opportunities to improve Decision Making for Companies Management or for Governments Macro laws. For example, an AI could predict the market reaction with more accuracy, based on a new laws proposed by Government or changing the interest level at National Bank.

We should keep an opened eye on Artificial Intelligence opportunities. The first wave of change will be held by engineers and the big 5 companies (Google, Microsoft, Amazon, Facebook, IBM). They all invest heavily, betting on their future. However the long term success will belong to entrepreneurs that understand the commercial potential of the new technologies.

The entire presentation on Artificial Intelligence is here: Artificial Intelligence.

Mihai Bonca

*****

Leave a Reply