NYU Professor Q&A
NYU Professor Q&A
The Futurist Perspective: Risk and AI in a Hyper Connected World Interview with Ina Wanca (Futurist), NYU Professor, and Cyber Risk & Governance in Smart Cities Expert
Given your focus on the connected world, what do you feel (not feel but think) is the greatest innovation to come out of 2018/2019?
There are many innovative ideas and prototypes which are being developed. However, the best way to approach this question is to think what problems we are trying to solve with the use of AI technologies in new ways? Since, the invention of the computer, engineers and researchers have dreamed of developing an AI system which can mimic human intelligence. Even though the word “intelligence” does not have a universal definition, it entails the ability of humans to think abstract and apply this knowledge into different domains. Today, AI systems do not have this capacity. They are good in executing specialized tasks such as planning, driving cars, translating but cannot make abstract claims to apply them beyond their field. The exciting part is that AI researchers are working towards creating a General AI system which can generalize concepts and mimic the human brain. Some great progress has been made with the development of the General Adversarial Networks which require less human supervision in training and can make machines learn more effectively. Another development is the Google BERT system that can complete the missing parts of sentences almost as well as a human.
Where do you see the greatest opportunity for artificial intelligence?
The greatest opportunity is to create a General AI system which can help people in many ways and not just with specific tasks. But, how do we get to creating a General AI remains an open question and there are many different approaches. These approaches are based on the study of the human brain. For example, Google DeepMind uses a biology approach and the study of the visual cortex in their neural network methods to create better AI systems. But, I think Dr. Ben Goertzel and his OpenCog AGI project is on the right path of creating a truly AI system which can think like humans. Ben’s team is using an integrative approach. You cannot create a General AI with one algorithm, instead there are multi-level algorithms which can compute what the operational human-level intelligence is doing. These algorithms are inspired by the theories of evolution, logic, and neuroscience among others and help researchers create a computational version of the brain processes. Today, we have an unfair advantage because data is everywhere, and computers are faster than years ago. So, the opportunity of creating a General AI is now.
How do you see these technological advances interacting/ impacting the financial industry?
Today, financial companies are investing heavily in AI solutions because they deal with a lot of data. On one hand, companies use matured AI solutions to automate certain predictable and simple activities related to administrative, compliance or data management tasks. On the other hand, companies are also experimenting with new AI techniques such as deep learning to gain insights of their data and create personalized solutions for their customers. For example, with the advances of natural language progressing, banks are investing in developing chatbots which can understand clients’ needs in regard to investment and can advise clients about loan rates, savings and even fully resolve clients queries without the intervention of a human rep.
How do you envisage (envision) the future of AI and ML?
ML is a subset of AI and it entails the ability of machines to learn like humans. But these are buzzwords and have been used in many different ways to describe a specialized type of AI algorithms such as deep neural nets. Majority of what we see today in regard to AI refers to supervised learning of computer systems - meaning the AI system is trained on a particular set of data.
I believe we will see a progress made in developing AI systems which can learn from each other and will be proliferated through blockchain technologies. I also envision AI to be used for voice and face payments over mobile devices. I believe there a few financial companies which are also investing through their R&D in developing facial recognition and bio-identification for check out payments. However, the AI is a broad term and MLor deep neural nets are only one tiny part of the AI field, which are good in one particular thing such as pattern, speech or image recognition. The AI literature is pretty long and deep and there are logic systems, evolutionary systems and other interesting AI methods that aid value. Each of these systems solve one particular area of what the human brain can do. So, we need them to network together and focus not only on ML or deep learning or reinforcement learning, but on all of them.
And how this can be applied to operational risk and financial risk?
The biggest opportunity is to use AI for fraud detection and improve risk rejected payments. Currently rejected payments cause more losses than actually a real fraud. AI can be used to understand more accurately clients’ behavior by gaining insights from clients' data streaming from multiple resources, including social media. The more personalized the AI solutions become, the more accurately they can identify if a customer is susceptible to make the payments in question. I believe in the future; banks will also create a central storage payment instruments tied to a particular client with ability to use AI to personalize the offers based on real-data about the customer. For example, if a customer is on the phone looking to buy a vehicle, an AI chat bot can immediately be able to advise the client on the rate of the loan and approve the client for a line of credit. Improving payment efficiency and time to access the funds will be a priority for the banks and AI can aid a value in these matters.