Publications
Partners
Applied Artificial Intelligence: A Handbook for Business Leaders by Mariya Yao, Adelyn Zhou, and Marlene Jia
Who is the most intelligent person you know? Why do you consider him or her so intelligent? Is it because she is a quick thinker who can absorb and process new knowledge immediately? Is it because he is astonishingly creative and seems able to generate a constant flow of new ideas?
People have a wide spectrum of what can be described as intelligence. Computers, on the other hand, are best known for the one specific area in which they are far superior to people: large-scale computational tasks. Aside from this narrow area of expertise, currently machine capability lags behind human intelligence.
This book provides an indispensable entre to AI for business leaders, and aspirant business leaders. They can ill afford not to have a working understanding of an area that will have a profound impact on their success.
"To help business executives disentangle the functional differences between different AI approaches, we've segmented applications along our Machine Intelligence Continuum," the authors explain.
Below are listed the key elements of each.
Rule-based automatons
At the lowest level of machine intelligence are "Systems That Act", which the authors describe as rule-based automatons. These systems are built to perform as they are scripted to, such as 'if this happens, then do that'.
An example is the cruise control in your car. It uses a motor to adjust the throttle position to keep the car at the speed you selected. It is incapable of dynamic actions or decisions. Most companies claiming to have AI are really using only "Systems that Act".
From known to unknown
The next level is "Systems That Predict". This system has the capability of analysing data and producing probabilistic predictions. It moves from known information to unknown information.
The retailer, Target, was able to identify 25 products, including unscented lotion and calcium supplements, that can predict the likelihood of a shopper being pregnant - and even the stage of her pregnancy. They used this information to target the lady and sell specific products. Of course, these predictions are only as good as the data. Flawed data will produce erroneous results.
Deep learning
To produce "Systems That Learn" requires both machine learning and deep learning. What makes this level different to the previous one is that they can perform tasks without being explicitly programmed for this. The result is that these machines can perform at human or better-than-human levels.
The system acquires data from which a prediction about the world is created. This prediction is then combined with higher-level judgement and an action, to produce a particular result. Information in the form of feedback and measurements from the result, can inform earlier decision points and improve the task performance thereafter.
A self-driving car, for example, must control the whole driving task. It must turn video and sensor feeds into accurate predictions of what is happening around it and take the correct action.
Are humans needed?
It is often asserted that with all the AI we still need humans, because only we are capable of creativity. This is not true. Computers have been used for creating design and art for decades. The recent breakthroughs in neural network models have led to a revival of computational creativity. Computers are now capable of producing original writing, imagery, music, industrial designs, and even AI software!
Sony's Flow Machines used AI trained on Beatles' songs to generate their own hit, Daddy's Car, which eerily resembles the musical style of the Beatles. Ditto for Bach music which human evaluators often couldn't differentiate from the real Bach.
Human employees collaborate more with AI tools at work and use digital assistants like Apple's Siri and Amazon Echo's Alexa. As such, machines will also need to be emotionally intelligent, and so we will increasingly need "Systems That Relate".
Sentiment analysis or emotional AI, extracts and organises emotional states from our text, voice, facial expressions, and body language. This data will allow computers to respond empathically, just as the most sensitive people do. Amazon is already prioritising emotional recognition for the Echo.
Towards mastery
A human toddler can see a tiger only once and develop a mental construct of the animal, and recognise other tigers. If people couldn't recognise different tigers, we would have been killed by them. A deep-learning algorithm needs to process thousands of tiger images to be able to recognise them in pictures and videos, but this doesn't transfer to other abstractions of tigers such as cartoons or a person in a tiger costume.
People are the ultimate "Systems That Master", intelligent agents capable of constructing abstract concepts from sparse data. We can create representations of the world and transfer knowledge from one domain to another.
There is a material difference between AI and AGI – artificial general intelligence. To date no AI system has AGI, only humans do.
The authors final category is "Systems That Evolve". These are systems that exhibit superhuman intelligence and capabilities. Such systems can change their own architecture and design, to adapt to environmental needs.
As humans, we're limited by our biological brains, our "wetware". We evolved through generations and cannot simply re-architect our own biology in our lifetime.
Computers are currently limited by both hardware and software: we are limited by our wetware. Some futurists suggest that we may be able to achieve superhuman intelligence by augmenting biological brains with synthesised technologies, but currently this is more science fiction than science.
If we could achieve this
synthesis we would achieve "singularity", when machine intelligence
surpasses human intelligence, leading to an intelligence explosion and the emergence
of superintelligence.
We are already seeing the promise of AI in action in the fight against social injustice and crime; in addressing health and humanitarian crises; in the solving of pressing community problems; and in the dramatic improvement in the quality of life for everyone.
Consider these examples.
When Sahil Singla joined FarmGuide, a social impact startup, he discovered that thousands of rural farmers in India commit suicide every year. When harvests fail, desperate farmers are forced to borrow from microfinance loan sharks at crippling rates, and unable to pay back, kill themselves.
Using FarmGuide which analyses satellite imagery to predict crop yields for individual farms, Singla and his team built better actuarial models for lending and insurance. Armed with this information they can provide farmers with loans at lower, fairer interest rates.
UNICEF's U-Report, is a social reporting bot that enables young people in developing countries to report social injustice in their communities via SMS and other messaging platforms.
When U-Report polled 13 000 users in Liberia to ask whether teachers at their schools were exchanging grades for sex, an outrageous 86% said yes! Within a week of the U-Report on the "Sex Grades" epidemic, hotlines around the country were inundated with reports of child abuse, effectively exposing a pervasive taboo, and empowering victims to speak up and reach out for help.
AI can dramatically streamline and improve medical care. In both pathology and radiology there is a reliance on people to spot anomalies. Studies reveal that two pathologists assessing the same slide of biopsied tissue will only agree about 60% of the time.
AI systems for diagnosing breast cancer utilises computer-vision techniques that have been optimised for medical image recognition. They were able to interpret patient records with 99% accuracy.
But AI is not without its own challenges. Poor or incorrect data leads to poor or incorrect results. Biases of the technology creators trickle down to their creations, which can have serious consequences for identifying terrorists, predicting criminal recidivism, and triaging medical cases, and so on.
Perhaps the biggest challenge is that while AI will transform the economy on a large scale, not everyone will benefit equally from the opportunities that will become available. New jobs in this economy will require greater technical competency than is currently available in the workforce.
"Business leaders have an ethical responsibility to workers to minimise and ameliorate the potential disruptions that AI may bring to the workplace," note the authors.
Investments in one's workforce could increase the likelihood that advanced automation increases productivity, while ensuring high levels of employment and shared prosperity.
Readability Light ---+- Serious
Insights High -+--- Low
Practical High ---+- Low
Ian Mann of Gateways consults internationally on strategy and implementation and is the author of 'Strategy that Works' and 'The Executive Update.' Views expressed are his own.
