How Can Artificial Intelligence Fight Fraud with NVIDIA

How Can Artificial Intelligence Fight Fraud with NVIDIA

With all forms of fraud increasing in occurrence, Kevin Levitt, Director Industry Business Development, Financial Services at NVIDIA shares some of the ways AI is fighting fraud.

Tell me more about NVIDIA

NVIDIA has been a presence in the financial industry for more than 15 years. Financial institutions harness NVIDIA’s full stack accelerated computing platform to power AI and high-performance computing applications that utilise vast amounts of data to increase revenues, reduce costs, and mitigate risk across the enterprise. NVIDIA works with the largest banks, credit card issuers, and insurers across the globe helping financial institutions deliver AI-powered solutions.

What problem did you set out to solve

Financial institutions bring a range of problems and opportunities to NVIDIA to solve through AI. Ultimately, financial institutions, including fintechs, seek NVIDIA’s expertise in accelerated computing to AI-enable hundreds of applications to drive better insights, increase revenues and remove operational inefficiencies. Key use cases for AI in financial services include identity verification for Anti-Money Laundering and Know Your Customer (AML/KYC), fraud detection for payment transactions, enhanced customer service via virtual assistants and call centre AI, as well as recommendation systems for next best action and experience personalisation.

Financial institutions recognise the key role AI will play in maintaining a competitive advantage, as 83 percent of respondents of our AI in Financial Services study stated that AI is important to their company’s future success.

What makes you unique

NVIDIA provides an accelerated  computing platform, we deliver a full stack, including GPU hardware, accelerated software libraries, and application frameworks; all three layers of which are critical to deploying AI into production. The platform enables banks to successfully execute their enterprise AI strategy – whether on premises or in the cloud.

For example, we enable banks to build sophisticated fraud detection applications that run on a GPU-based accelerated platform that enables faster fraud detection running 24 hours, 7 days a week around the globe.

PayPal saw a 10% increase in fraud detection. This type of improvement is extremely valuable to financial institution customers, resulting in improved customer experience and better merchant satisfaction.

How severe is the fraud/cybercrime threat currently?

We work closely with Financial institution customers all over the world. Leveraging AI to mitigate fraud is one of the top conversations. We help migrate their models from less sophisticated machine learning models to deep learning, providing them with a platform that leverages more data and more sophisticated models. AI-enabled applications for fraud detection improve fraud detection accuracy and reduce false positives – running in less than 10 milliseconds on NVIDIA GPUs – delivering enhanced revenues to the bank and a better checkout experience to merchants and their customers.

 What are some of the trends in utilising AI you’re seeing currently? H

We are observing three key trends in utilising AI for fraud prevention today:

  1. Companies are migrating from rudimentary rules based and machine learning fraud detection models to more sophisticated deep learning/AI enabled models. Deep learning models incorporate more data and in turn are more accurate, resulting in better fraud prevention.
  2. Banks are moving beyond structured (tabular) data and now are including unstructured data (images, voice, etc.) into their fraud prevention models.  For example, Intelligent Voice has an AI-enabled solution that analyses a callers voice to flag identity fraud.
  3. New GPU-powered technologies are enabling federated learning and other data sharing models that protect consumer privacy and adhere to data residency requirements while allowing more data to be shared within and external to the company. This allows financial institutions to share data in an encrypted and privacy maintaining fashion so that more data may be incorporated into fraud prevention models. With more data to build better AI-enabled applications, data scientists at the banks are building more sophisticated models to fight financial crime.
  4. A financial institution can use the NVIDIA platform to execute a model during the transaction flow. For example, American Express uses NVIDIA Triton to run a deep learning-based model in under 2 milliseconds. This not only allows better fraud detection but reduced the number of false positives that could occur with less sophisticated models.

Why is it important for banks and other financial organisations to address the fraud issue?

Banks have seen the early results of AI investment as AI has moved out of the R&D labs and into enterprise-level production.

Fraud results in significant losses to the bank and ultimately to the consumer, either indirectly through potentially increased prices for products or directly by having funds stolen from their account. Banks must stay a step ahead of the bad actors and AI is the best tool to protect the broader financial ecosystem.

Is AI the key to solving the fraud issue in finance?

Financial criminals are getting smarter. From home insurance to banking to asset management, the key is better insights, faster.

AI helps financial institutions keep up with criminals by incorporating vast amounts of data into models that minimise latency, for faster, more accurate insights.