How Do We Create Artificial Intelligence That Is More Human?
LIKU baby humanoid robots are demonstrated on the Torooc Inc. stand on the opening day of the MWC Barcelona in Barcelona, Spain, on Monday, Feb. 25, 2019. At the wireless industry’s biggest conference, over 100,000 people are set to see the latest innovations in smartphones, artificial intelligence devices and autonomous drones exhibited by more than 2,400 companies. Photographer: Stefan Wermuth/Bloomberg photo credit: © 2019 Bloomberg Finance LP
By Jennifer Kite-Powell
On February 11, 2019, President Trump signed an executive order on Maintaining American Leadership in Artificial Intelligence and in February 2019, a survey called Artificial Intelligence and Machine Learning indicated that only 16% of business leaders surveyed are getting significant value from advanced artificial intelligence (AI) in their companies. The report also found that companies of all sizes and across industries are investing heavily in advanced AI with an average of $36M spent in the fiscal year 2018.
Of those same companies surveyed, 10% plan to increase their budgets over the next two years. The survey also found that businesses in Asia-Pacific have adopted advanced AI faster than the rest of the world. Twenty-two percent of Asia-Pacific companies in 2018 were at the advanced stages of machine learning compared to only 11% in North America and seven percent in Europe.
On March 4, 2019, UNESCO held a conference on creating core principles around AI with an emphasis on a more humanist approach to AI. The goal was to foster dialogue between the public and private sectors, technical community, media and academia, civil society, international and regional organizations and look at how much machines should be allowed to decide for us as a society including who writes what values and priorities into the algorithms of machines.
In a statement, Audrey Azoulay, the Director-General of UNESCO said that AI is humanity’s new frontier.
“The guiding principle of AI is not to become autonomous or replace human intelligence. But we must ensure that AI is developed through a humanist approach,” said Azoulay.
Ajay Bhalla, President, Cyber & Intelligence Solutions, Mastercard, says AI is already impacting most of the industry conversations today.
“Over the next few years, there won’t be a single sector of the economy untouched by AI. It will massively increase the value and speed of data transactions, interactions, and decisions,” said Bhalla.
“AI is therefore critical to the business of the future – building systems through AI will be essential for remaining competitive,” said Bhalla.
H.E. Younus Al Nasser, Assistant Director-General of Smart Dubai and CEO, Smart Dubai Data agrees.
“Artificial intelligence has become a sector of its own [ ..] and as an increasingly growing part of our everyday life gets automated, we can only look forward to many more AI-powered breakthroughs and services, and this calls for setting principles to regulate and guide the sector,” said Nasser
“That said, the field is not mature enough yet to draft laws to govern it [..], so we developed an Ethical AI Toolkit to set clear guidelines on the ethical use of the technology, and prevent having a fragmented, incoherent approach to ethics, where every entity sets its own rules,” said Nasser. “We look forward to seeing the Toolkit evolve into a universal framework that determines ethical requirements for AI design and use, offering informed solutions to help stakeholders adhere to ethical principles, and, eventually, serving as a blueprint for governments to draft pragmatic AI laws and regulations.”
When it comes to trust, Mastercard’s Bhalla does believe there’s a lot of talk about trust in technology today – whether it’s the use of consumer data, privacy or the proliferation of new platforms and technologies in the sharing economy.
“The payments and financial services industries have tended to lead the way in building more secure systems as people’s money is at stake,” said Bhalla. “There’s a growing understanding that ethics encompasses more than regulation and compliance, with conversation moving from ‘we are compliant’ to ‘we are doing the right thing,’” said Bhalla.
“Ethical considerations provide competitive differentiation, and this is a principle we follow [..],” said Bhalla. “We talk about privacy by design and security by design – a value proposition based on trust. If we consider AI, we need to be thoughtful about how AI technology can be used and create standards that define where AI can be used, when it can be used and what it can access.”
According to Rajan Sethuraman, CEO, LatentView Analytics, ethics in AI is a wide sweeping topic because it is an attempt to apply what are often contextual and cultural norms and guidelines to virtually endless theoretical applications.
“When AI is used to automate the most mundane repeatable human tasks, then ethical considerations may not run that deep,” said Sethuraman. “But the more we train and ultimately expect machines to emulate us as social beings and attempt to integrate fundamentally human ideas like judgment, empathy, or fairness into an AI equation, the more we’re faced with the same ethical issues that accompany human interactions.”
The ‘rights’ of sentient automatons is a theme we’ve seen over and over in popular sci-fi movies – and while we’re not yet at the level of a HAL 9000 – ethics are something we must deeply consider as we develop more and more advanced AI with less and less predictability,” added Sethuraman.
Shekhar Vemuri, CTO, Clairvoyant believes that ethics in AI needs to be viewed as a coin with two equally important sides.
“Anyone developing AI technology – from big tech to small startups – needs to ask themselves if they have permission to use whatever data they have to power the AI. If not, everything else that follows will be tainted with that initial ethics breach,” said Vemuri.
“On the other side of the coin is transparency. A lot of companies just focus on how powerful and scalable their AI is. That’s important, but AI is a bit like a government – or anything people turn power over to – its authority and ability to grow must be checked by transparency,” said Vemuri. “Is the machine doing the correct thing? And more importantly – if the outcome isn’t right (e.g., biased bank lending or hiring), then is the process and plumbing open to inspection? Can the algorithm be audited to uncover the flaw, or is it all a black box?”
Vemuri believes that not addressing and solving these privacy and transparency questions will result in a backlash which will translate into regulatory brakes being applied to the adoption of AI.
“It’s our social and ethical responsibilities as AI practitioners to ensure our natural human biases aren’t amplified in the solutions we build – and if they do creep in and tip the scales – that there’s a transparent process to root out and fix the problem,” added Vemuri.
But Bhalla believes that AI can have a valuable role in fighting fraud which affects consumers.
“Mastercard has been investing in AI technology for the past decade, and AI is proving invaluable in fighting fraud, particularly anti-money laundering,” says Bhalla. “Seventy billion transactions a year pass through our AI-enabled fraud and decision platforms have helped prevent more than $50 billion of fraud in the last year.”
The company acquired AI specialists Brighterion in 2017. Bhalla says this allowed Mastercard to push the limits of what they can do with detection towards prediction.
“There’s simply not a single part of our business that will remain untouched by AI, from loyalty and reward schemes through fraud prevention and user experience to new channels,” said Bhalla. “In the future combining AI with biometrics will be an important way of transferring the intelligent recognition we use in the physical world to the digital realm. Identifying people and their devices online is currently a major challenge we’re addressing.”
Sethuraman returns to the ethics of AI.
“Ethics in AI must increasingly consider potential outcomes and attempt to keep humans as involved and in touch with the input and design as possible,” said Sethuraman. “This is humanist intelligence – or the augmentation of humans by machines and machines by humans, and is in fact how most AI will most likely play out over time.”