Alan Turing is often referred to as the Father of AI. As a WWII codebreaker, he was responsible for the design and build of a machine that could solve mathematical problems beyond the calculation capability of humans through using manual computational methods. In the 2014 movie, Turing was played by Benedict Cumberbatch, seen here in front of his code-breaking machine. Turing’s invention is said to have shortened the war by two to four years, saving millions of lives.
In cracking the German Enigma Code using the machine, a dilemma was exposed: do you save lives now by scuttling a German operation that you know is taking place, but thereby risking the encryption codes being changed again by the Germans, which would mean starting over. Or do you allow that operation to go ahead unimpeded, to preserve the decryption advantage for something bigger and more decisive, effectively letting many of your people die as a consequence?
This dilemma highlights a fundamental paradox in the use of computing, machine learning and AI, for any human decision-making problem. The application of the answers will have very human consequences. Essentially Artificial Intelligence reflects the human mind and society - with all the hidden and unknown scripts that determine human behaviour.
Take this example published in WIRED from 2018:
IN 2015, a black software developer embarrassed Google by tweeting that the company’s Photos service had labelled photos of him with a black friend as “gorillas.” Google declared itself “appalled and genuinely sorry.” An engineer who became the public face of the clean-up operation said the label gorilla would no longer be applied to groups of images, and that Google was “working on longer-term fixes.”
More than two years later, one of those fixes is erasing gorillas, and some other primates, from the service’s lexicon. The awkward workaround illustrates the difficulties Google and other tech companies face in advancing image-recognition technology, which the companies hope to use in self-driving cars, personal assistants, and other products.
So perplexing is this challenge that even this year Facebook made the same error in a video.
NY Times, Sep 3 2021. Ryan Mac
The complexities of applying AI to image recognition go beyond exposing unconscious culture and gender bias, it extends to something comparatively less ambiguous such as skin cancer detection. Take the example of a neural network which was been trained to detect malignant skin lesion:
“Because medical images of cancerous lesions include a ruler for scale, the model learned to identify the presence of a ruler as a marker of malignancy, because that’s much easier than telling the difference between different kinds of lesions.”
VentureBeat: When AI flags the ruler, not the tumor — and other arguments for abolishing the black box
Like Turing’s machine, AI represents a way to solve problems at a scale humans are not capable of. However, an algorithm has no sense of what it is to be alive, inhabit the world, and be sentient. It is prone to making decisions that do not make sense in the context of being human.
What is becoming increasingly apparent is the need for transparency in the conclusions that AI algorithms reach – so that we can make sense of how it arrived at any given answer.
In the first Industrial Revolution we overcame the limitations of human muscle power, with the use of steam. In the context of the Fourth Industrial Revolution, we are overcoming the limitations of the human mind to conceive and solve complex problems using Artificial Intelligence.
The value of AI is unlocked as the mechanism that leverages the power of hyper-scalable digital infrastructure, cyber-physical systems, and vast amounts of data, to create new ways to generate business value, and solve social and environmental problems.
One-minute watch: Tomorrow Starts Here: 2014 commercial by Cisco
Consider the value chain in the video above and how it is being transformed because of the evolution of digital business - specifically the use of AI, IoT and hyper-scalable digital infrastructure. Driverless cars may be some time away, but they will transform entire sectors of the economy – from personal transportation to the need and demand for parking and car ownership models. The use of AI has far-reaching consequences in terms of the creation of new forms of business value.
The mainstay of Artificial Intelligence is to solve big, complex problems – from optimising courier delivery routes to the use of driverless trucks to haul freight from one hub to another.
Recommended watch - FRONTLINE Dec 2019: The Age of AI. Embark is a start-up that is already operating a fleet of autonomous trucks in the South West USA.
Upstart companies like Embark Trucks are predicting substantial adoption in the next five years of autonomous freight haulage, with significant efficiencies along the way. Advances in autonomous vehicle use are already in operation and no longer science fiction, and are incrementally becoming science fact.
“I really do think the world has entered a new era.
Artificial intelligence holds so much promise, but it's going to reshape every aspect of the economy, so many aspects of our lives.
Because A.I. is a little bit like electricity.
Everybody's going to use it.
Every company is going to be incorporating A.I.,
integrating it into what they do, governments are going to be using it,
non-profit organizations are going to be using it.
It's going to create all kinds of benefits in ways large and small,
and challenges for us, as well."
The promise of AI is to change the way we do things, in the same way that steam power transformed society, economics and business in the first Industrial Revolution. What it means at a practical level in your business today is that it can be applied to substantially transform how value gets created.
Far more than business analytics (the application of analytics to support human understanding and decision making as it pertains to your business), the application of Neural Networks, Machine Learning and Artificial Intelligence can enable a decision-making process to be independent of a human and performed in a millisecond.
In the case of the autonomous truck, the human operator is making millions of decisions during operation. One substantial advantage a human has is that it is sentient and therefore understands the human impact of making a bad decision – such as avoiding a child chasing a ball on to the road vs running over an animal of some kind – each has a quite different moral hazard associated with it.
It is fair to say that an algorithm may have a better reaction time than a human, and so could outperform the reflexes of a human operator. However, the absence of an ethical code associated with being a sentient being may override the reflexive advantages.
The big opportunity for AI is to deliver significant efficiency gains and new forms of value through automated decision making, leveraging substantially more data points than a human can possibly reference. The flipside is, that without transparent models to interrogate the decisions being made, an algorithm could easily misdiagnose a skin cancer, simply because a ruler is present in the data set. Or classify someone of a different skin colour as being a gorilla or primate, and therefore having a lower moral hazard score, when having to choose between hitting or avoiding a human vs an animal.
When considering the adoption and use of AI, the West has taken its cues from the advances driven by Silicon Valley entrepreneurship. However, over the last decade or so, China has become the other major global force in AI. Substantial government funding has been granted to build smart cities from the ground up, using AI in all sorts of different and innovative ways – all unconstrained by Western norms. The use of facial recognition in China is pervasive for everything from making fast-food purchases, to policing jaywalking.
In China, Facial Recognition Tech is Watching You. FORTUNE, 2018
This heavy investment by China is fuelling a different sort of arms race, one where superiority in the application of AI is giving rise to a new type of superpower state. The Silicon Curtain, similar to the Iron Curtain, is the divide between the application of AI for competitive advantage between competing nation-states. Underlying this AI arms race is the cultural norms of two very different cultures of China and the USA.
For Artificial Intelligence to work well, it needs vast data sets. China with an entirely different relationship to privacy and a one-party state has an enormous advantage in source data. Unfettered by Western concerns for individual freedom and privacy, China collects data on its population of nearly 1.5 billion. AI is being applied for the collective good, but potentially at the cost of individual liberty.
The ethical norms of a society have a direct bearing what the beneficial or detrimental impacts there will be of using AI.
Yuval Noah Harari – Author of 21 Lessons for the 21st Century
Consider this point by Yuval Noah Harari, in a recent interview on Channel 4 News entitled: Will artificial intelligence create a useless class of people?
“…there is also the dangerous potential of the new technologies and as I said that the biggest danger is the rise of new totalitarian regimes worse than anything we've seen before. Because for the first time in human history it's now possible to follow everybody all the time and you now have AI algorithms, which unlike human beings these algorithms are able to process enormous amounts of information very fast. And therefore we are in a position that you can follow everybody all the time. You can get to know people better than they know themselves. If you collect enough information on me and you have enough computing power you, can get to know me better than I know myself, better than my husband knows me, better than my mother knows me. If this is not managed wisely this is a recipe for the worst totalitarian regime that ever existed - there is no privacy.”
These concepts may seem abstract and removed from our day-to-day professional lives, but the decisions we make around applying AI will have far reaching consequences on jobs, livelihoods, and personal freedoms. In an industrial setting, a decision to allow AI to control a failsafe valve could have far reaching consequences without appropriate human oversight. The Boeing 737 Max MCAS software malfunction serves as a cautionary tale in relation to using an algorithm without appropriate transparency, pilot training and adequate systems design. Boeing will pay more than $2.5 billion to settle a criminal charge related to the two 737 Max plane crashes in 2018 and 2019 that killed 346 people.
As leaders in our industry, we have a moral obligation to ensure there are sufficient guide rails on the use of AI in our enterprises.
While substantial leaps forward and entirely new forms of value will come from Digital Business innovation using AI, it comes with a moral hazard that could be as dangerous as using nuclear power. In Turing’s case the moral hazard of breaking the German Enigma code, which was beyond the reach of human computers, meant that lives were both spared and sacrificed.
As we navigate the new possibilities of Digital Business, long term advantage will come not just with the use of the most advanced technology, but with the ethical application of the technology with the right levels of transparency and oversight.
For a deeper dive into the Cognitive Paradox and the implications of AI use, read The Alignment Problem by Brian Christian. Available in both audio, print and ebook.
