In our Zoom for Thought on July 6th, 2021, UCD Discovery director Prof. Patricia Maguire spoke with Dr. Prag Sharma, Senior Vice President at Citi Innovation Labs, on the subject of, “Technology-fuelled innovation”. In case you missed it, here are our Top Takeaway Thoughts and a link to the video.
Citibank is a global bank with over 200,000 staff and a presence in 160 countries, including Ireland. Prag is Global Head of Citi’s AI Centre of Excellence, a “one stop shop for ensuring that Citi approaches innovative technologies like machine learning and artificial intelligence consistently across the organisation, that they are using these technologies ethically and responsibly, and that they are providing products and services that are based on these technologies in a way that benefits customers and society”.
The pandemic has accelerated innovation and “technologies are also converging at the same time”. Cloud computing gives startups access to the same computing power as large organisations so they can compete directly without having to invest heavily in infrastructure. Meanwhile, with the Internet of Things (IoT), millions of different devices like our smartphones can connect to the internet. “Blockchain is another sort of technology that is really changing the world and challenging regulators and bankers to think fundamentally differently about money that we use every day.” Another technology that “cuts across all of these” is AI. “Artificial intelligence is promising a lot and it appears that in some of these promises it will deliver. When it delivers well, it will make a huge difference to how we do things in everyday life.”
There has been “a general realisation” across industries that you need multiple different skill sets to build something innovative that lasts. “You need designers, back end developers, front end developers, AI experts, other technologists and other skill sets to make a complete product or service.” The broad spectrum of expertise at Citi facilitates this design thinking. “If I wanted to talk to somebody about private banking for extremely high net worth individuals, I have experts in that space who know how to develop products there. If I was interested in environmental, social and governance related activities, we have experts in that space also… This multidisciplinary approach is essential.”
A Price Waterhouse Cooper survey published recently found that 40% of Irish workers believe that their jobs will be obsolete within five years, mostly because of automation. Prag does not believe these fears are justified. “We are overestimating that the jobs are going to go away… Every time new technology comes we think it is going to take over jobs, but actually jobs will just shift and form around that new technology.” The jobs won’t be lost soon but “may change a little”. While some jobs will eventually be lost to automation, “that’s the very long term” picture.
Machine Learning Bias
Machine learning bias comes about because “humans can be very biased” and “when we build something, that can also pick up on our biases”. If we input biased data into a machine learning algorithm the AI “simply amplifies our own biases, so we have to be careful when we use these technologies”. The western world generates a lot more data than other countries so data is “very one-sided” and therefore biased. “Because we have access to more technology, we generate more data. If we generate more data, then we have more data to analyse and build models on. Then those models will work better for us, as opposed to other people in other parts of the word.”
If an application is recommended to you through a machine learning algorithm, it is best practice to interrogate the fairness of the underlying data. “You have to take it with a pinch of salt. Your own gut feeling, instincts, what you have learned through life are there to guide you across this new AI-driven world of innovation and machine learning. So use it.”
A 2018 study claimed that Bitcoin, which uses blockchain encryption, uses as much energy as Ireland. “I don't think it's a great idea that this network consumes so much energy, and there needs to be better ways to mine for coins.” This is just one of the “grand challenges” we face today, along with climate change and the biodiversity crisis. “We are really impacting our environment and other living things. We need to adopt technologies to help us in that regard.”
Will machine learning always require human expertise? “It might not require humans-in-the loop if we can make conscious machines. But we don't know why we are conscious or how we are conscious so we might trip up on that or we might accidentally discover it.” In the short term we need humans-in-the-loop because “we really don't understand exactly how these trillion-parameter AI models work. But in the long term your guess is as good as mine”.
This article was brought to you by UCD Discovery, fuelling interdisciplinary collaboration.
Based in Dublin, Dr. Prag Sharma leads the Emerging Technologies Group within Citi’s Innovation Labs.
Data Analytics and Artificial Intelligence is a focus for the group, which includes research and development on Machine Learning,
Natural Language Processing and Graph Analytics approaches among others.
Distributed Ledger Technologies (i.e. Blockchain) is another area of research for the group.
Prag has significant experience in R&D Management, experimental and algorithmic design and development in both business and academic environments with start-ups and multinationals.
Prag holds an MBA and a PhD in Computer Vision.