Indian team backs IBM's Telum chip aimed to stem $30 bn consumer fraud loss

IBM Telum Chip: The team owned a significant portion of the logic design, verification and validation, in addition to the complete physical design of the new dedicated on-chip accelerator

Kailash Gopalakrishnan, 43, a top technologist at IBM, enjoys travelling across the world. He also likes to visit beaches in the Caribbean and Mexico, which are warmer spots for him, especially during winters in New York City, where he lives. But one of his most favourite places is in Kerala, which he visited in early 2019. Due to the coronavirus pandemic, Gopalakrishnan has not been able to travel outside the US or visit his family in India.

But the pandemic has not stopped Gopalakrishnan from pursuing his other passion, which is artificial intelligence and collaborate with his team in India and build innovation remotely. In fact, the India Systems Development Lab (ISDL) of IBM has contributed significantly to the Telum chip, details of which were recently unveiled by the Big Blue. The new IBM Telum Processor is designed to bring deep learning inference to enterprise workloads to help address fraud in real-time. 

Telum is IBM's first processor that contains on-chip acceleration for AI inferencing while a transaction is taking place. Three years in development, the breakthrough of this new on-chip hardware acceleration is designed to help customers achieve business insights at scale across banking, finance, trading, insurance applications and customer interactions. A Telum-based system is planned for the first half of 2022.

The Telum processor is an enterprise CPU chip developed with technology that came from IBM Research. The consumer loss from fraud is growing dramatically and is expected to reach in excess of $30 billion by 2023. IBM said this solution can help its clients minimise losses from fraudulent transactions.

“India (team) was a very critical pillar of our overall work that has happened, not just in the last couple of years but over the last 5-6 years,” said Kailash Gopalakrishnan, IBM fellow and senior manager, accelerator architectures and machine learning, IBM Research. “There are many contributions related to AI from the IBM Research Lab in India. As we were designing and building this chip, we were able to collaborate very closely with the design teams from IBM India. We had lots of members from IBM India who worked really hand in hand with us.”

For instance, the India Systems Development Lab (ISDL) hardware team has played a vital role in the design, verification, hardware validation, and power management of the 8-processor core IBM Telum processor. The team drove several innovations to translate the micro-architectural blueprint into a functional design.

The team’s traditional strength and expertise in verification ensured the functional correctness of the core components. The circuit and layout design team improved area efficiency and methodology while meeting the high-frequency requirements across numerous units. This included the memory controller. 

The team also owned a significant portion of the logic design, verification, and validation, in addition to the complete physical design of the new dedicated on-chip accelerator. This enables real-time AI-based embedded directly into enterprise transactional workloads in a secure environment.

Finally, a team of experts has enabled hardware validation and power management, while working remotely in these constrained times. With such extensive contributions to all aspects of the design, IBM said it is fair to say the fingerprint of the India hardware team can be seen across the Telum processor.

This chip enables clients to conduct AI-driven workloads at scale with low latency. The 7 nm microprocessor is engineered to meet the clients' demands for gaining AI-based insights from their data without compromising response time for high volume transactional workloads. IBM Telum is designed with a new dedicated on-chip accelerator for AI inference to enable real-time AI embedded directly in transactional workloads, alongside improvements for performance, security, and availability:

“The technology allows our clients to do AI at scale,” said Gopalakrishnan, an alumnus of IIT-Bombay and Stanford University.

In the digital economy, data is the new natural resource. It offers tremendous opportunities for enterprises to harvest value. Yet extracting insight from real-time enterprise transactions can present an elusive goal. Running deep learning models on high volume transactional data is difficult. There are issues related to latency, variability and security concerns. This can make those models impractical in response-time sensitive applications. IBM is addressing this challenge through recent innovations.

“This chip allows you to really enable AI very closely to the data on the platform itself,” said Gopalakrishnan.

This is important at a time when businesses typically apply detection techniques to catch fraud after it occurs. This process can be time-consuming and compute-intensive due to the limitations of today's technology. This happens particularly when fraud analysis and detection is conducted far away from mission-critical transactions and data. Due to latency requirements, complex fraud detection often cannot be completed in real-time. This means a bad actor could have already successfully purchased goods with a stolen credit card before the retailer is aware fraud has taken place.

Gopalakrishnan said what happens today is when one uses a credit card or  ATM card and if there is a fraud, it gets detected, typically a few hours or a day after the event has occurred. “What we can enable with Telum is to be able to detect the fraud in the window of a transaction,” said Gopalakrishnan. “The transaction itself happens within  10s of milliseconds. Within that window of the transaction, even before you get authorization when you swipe your card, (the technology) allows you to detect that fraud. We can move from fraud detection posture to a fraud prevention posture.”

According to the US government agency Federal Trade Commission's 2020 Consumer Sentinel Network Databook, consumers reported losing more than $3.3 billion to fraud in 2020, up from $1.8 billion in 2019. IBM said Telum would enable evolving from catching many cases of fraud today, to a potentially new era of prevention of fraud at scale. This can be done without impacting service level agreements (SLAs), before the transaction is completed.

The new chip features an innovative centralized design, which allows enterprises to leverage the full power of the AI processor for AI-specific workloads. This makes it ideal for financial services workloads like fraud detection, loan processing, clearing and settlement of trades, anti-money laundering and risk analysis. With these new innovations,  IBM said clients will be positioned to enhance existing rules-based fraud detection or use machine learning. They would be able to accelerate credit approval processes, improve customer service and profitability. They can identify which trades or transactions may fail, and propose solutions to create a more efficient settlement process.

The chip contains 8 processor cores. The dual-chip module design contains 22 billion transistors and 19 miles of wire on 17 metal layers. Gopalakrishnan said India is an important market for IBM. He said the technology would have a huge impact on any market where financial transactions take place and it includes India as well.