- About us
- Alliances and Partnerships
- Global Offices
- Engagement Models
- Experienced Teams
- Flexible Services
- Service Excellence
- Data-Informed Decisions
- Tools and Techniques
- Data Science
- Consumer Packaged Goods
- Financial Services
- Engagement Models
- Experienced Teams
- Service Excellence
- Flexible Services
- Data Informed Decisions
- Tools And Techniques
- Data Science
- Business Intelligence
- Global Analytics Partnership
- Customer Insights
- Marketing Effectiveness
- Pricing Analytics
- Products and Solutions
- Fractal Sciences Lab
- Data Transformation
- Insights And Predictive Analytics
Fractal Analytics Blog
Morality, Humanity & Possibility of AI – A discoveryBy Sagar Shah
March 24, 2017
The world is changing in all walks of life – business, politics, economies, societies, individuals. The cycles of AI (decades) when AI failed to progress because of lack of data & computing power, are almost over. Today, we have a lot of data, and also computing power to run, train & model algorithms to solve our business problems using machine learning. Also, these algorithms will become usable in a human world with the current biases in the data as well. For instance, Microsoft’s chatbot Tay used Twitter data & interactions with users. Unfortunately, Tay became an aggressive racist and had to be closed down in 24 hours. It’s still exciting, uncertain, optimistic times ahead.
Artificial intelligence, machine learning and now deep learning in businesses has made Amazon, Uber, Airbnb, Tesla, Google, Facebook, Instagram, Snapchat, LinkedIn the new-age companies with maximum daily active users compared to many others. Cutting through all industries – automation, algorithms, robots have cut down repetitive jobs making better products. Recently launched Amazon-Go Concept Store promising cashier-free shopping experience is just the beginning of disruption of CPG-Retail landscape using deep learning algorithms. DeepMind beat a human in the complex game of Go using Deep Learning. (a previous outstanding challenge). Blippar demonstrated an augmented reality app to recognize objects in real-time using deep learning. This year, Blippar received Series D Funding by Khazanah (the same firm invested in Alibaba, and recently Fractal Analytics).
An AI Assistant app can now help you buy the perfect shoes. Siri, Google Assistant, Alexa, Cortana have reached good user experience levels now. It’s predicted that AI bots will power 85% of customer service interactions by 2020. Shareholder reports, legal documents, market reports, press releases, white papers may be written in part by AI. Finance companies are trying to predict stock markets. Matchmakers are using algorithms with a mindboggling 1/3rd of all marriages being carried out by these apps in today’s world. Android operating system’s speech recognition system, being powered by Google Brain uses artificial neural networks using deep learning algorithms on 16,000 computers.
Chatbots are using Artificial Intelligence, natural language processing, speech recognition, flow optimization. The fundamental idea of a chatbot is to allow only enough user interface for customers to interact with, whether conversational and/or widgets, to delight them by a service/brand with immediate access without installing a new app. (Octane AI, Assist, Gupshup, Textit, api.ai, IBM Watson, Chatfuel, converse.ai are some examples of services for chatbots. This can immensely impact call center jobs since people would start preferring typing on their mobiles to query instead of calling up. Messaging has surpassed social networking already. Speech processing, virtual assistants, Chat Bots are just some examples which can scare many insecure of their jobs affected by these technologies) – driverless cars, call centers and so on. Bill Gates has called AI the holy grail of Computer Science. Elon Musk on the other hand calls it ‘summoning the demon’.
There has been some criticism in academic circles about neural networks (deep learning) not being science but technology and data scientists are ‘only engineers’. Technology writer Roger Bridgman suggests “the bunch of numbers that captures its behavior would in all probability be “an opaque, unreadable table…valueless as a scientific resource… An unreadable table that a useful machine could read would still be well worth having”
“The greatest achievement of our technology may well be the creation of tools that allow us to go beyond engineering – that allow us to create more than we can understand” – ‘The Pattern on the Stone’ by Danny Hallis
Artificial Intelligence is far from perfect and has advanced issues to be resolved.
There is a fear that it will overtake jobs. There is a fear that it will overpower humans. There are issues in implementation that it will be unable to understand the complexity of human emotion and decision making and make only the rational decisions it is programmed for.
How does an AI machine take a moral decision? How do you code one? Is morality timeless; is it universal, or objective?
The philosopher John Stewart Mill’s Utilitarianism assumes intrinsic value & all other values are believed to derive their worth from their relation to this intrinsic good as a means to an end. This hedonic calculus i.e. analysis of happiness as a balance of pleasure over pain assuming that it is possible to compare the intrinsic values produced by two alternative actions and to estimate which would have better consequences.
Let us consider the famous trolley thought experiment where a trolley running down the tracks with 5 immobile people. There is an alternate track with 1 immobile person. Which would the train driver choose? Kill 1 to save 5 or kill 5? Mill’s Classic Act Utilitarian view suggests it’s imperative to kill 1 person to save 5 for the greater good for the greater number. However, this is immoral as per Immanuel Kant’s philosophy of reason and morality, which absolutely abhors killing in any situation (that is why Batman never kills Joker in-spite of catching him 100s of times and puts him in Jail, in-spite of the fact that Joker gets out every time and kills hundreds of people every-time before being captured. Batman follows Kant’s absolute morality here).
Utilitarian rule suggests that if this creates less utility in the long-term by setting a precedence and people fearing for their lives, you shouldn’t take the decision to save 5 for 1’s life.
A similar but a bit different experiment is as follows: If you have 5 patients in a hospital in need of a kidney, heart, lungs, brain and eyes (and all 5 will die if they don’t get it right away) and assuming all can be transplanted from a single person, and this person is happily sitting in a nearby coffee shop. Would you, as a doctor visit the coffee shop to kill that person to save the 5 patients? This puts higher pressure than the trolley problem on some people. This would create a huge fear in public mind for long-term and might start killing a lot of people for similar causes and hence this is a clear answer that the person in the coffee shop lives since greater good for greater number is guaranteed for the longer term.
Self-driving cars, robo-doctors in the future and similar artificially intelligent machines in charge of similar decisions will face similar issues. If you are in a self-driven Uber and the car has to decide similar to the trolley problem whether it kills 5 people to save you (customer) in the Uber versus kill the customer to save 5 others; which would you want Uber to choose? Yourself, most probably. This goes against the Utilitarian principle.
Now imagine that out of the other 5, two are your parents. Would you still want Uber to take the same decision? Probably not.
The answer is not so simple and often will create dilemmas difficult to resolve since there is no universal right answer which everyone would be ready to accept. Philosophy, Psychology & Artificial Intelligence will intersect every-time in such situations. A human-like AI is tough to program. While humans are mostly irrational, we will soon have moral, rational robots & machines – Artificial ‘Human’ Machines.
Scary or Exciting?
- Book: ‘The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World’ by Pedro Domingos
- Ted talk on Deep Learning
- Google Talk by Pedro Domingos (on youtube)
About the author:
Sagar Shah is a Principal Consultant at Fractal Analytics. He has close to a decade of experience in Analytics Consulting and has partnered with CXOs to institutionalize analytics in Fortune 100 clients for building roadmaps & execution. He has led multi-cultural teams (120+), managed multi-million dollar relationships, engaging for several global solutions – AI-ML, Deep learning, Big data, digital analytics, marketing optimization & pricing analytics, social listening, initiative assessments, immersive story-telling, BI-Visualization, rationalization & consolidation. He has also conducted 100+ workshops across North America, Latin America, Asia, Europe and led business development with 200+ proposals/RFPs. Sagar is also passionate about writing and travelling.
- Advanced Analytics (12)
- Artificial Intelligence & Machine Learning (4)
- Big Data & Visualization (4)
- Customer and Loyalty Analytics (3)
- Customer Insights and Analytics (26)
- Digital Marketing (4)
- Forecasting and Business Intelligence (34)
- Institutionalizing Analytics (6)
- Integrated Marketing Effectiveness (1)
- Marketing Analytics (29)
- Marketing Effectiveness (5)
- Others (20)
- Pricing Analytics (11)
- Sports Analytics (2)
- @raradhyula @fractalites will share insights on "Building careers in the age of #AI" at @iitbombay on August 21. - 4 days ago
- #HappyIndependenceDay! Let's go back in time and hear it from those who witnessed it on 15th August, 1947 /t.co/OFvCwBTttp - 6 days ago
- Tejas S. to take a session on "Uses of #MachineLEarning to solve business problems" at IIM Nagpur on August 13 /t.co/3kuGHMpdq5 - 10 days ago