Financial institutions in Asia are banking on generative AI, with the International Data Corp expecting the sector to spend US$4.3 billion ($5.87 billion) on generative AI by 2027. The technology will be used to enhance operations efficiency, automate repetitive tasks, and optimise back-office processes such as fraud detection and the creation of intricate documents.
Unlocking the vast potential of generative AI requires the support of a robust and resilient data infrastructure. “Generative AI models consume massive amounts of unstructured data, typically at the terabyte level. This is the consumption of information for training the models. By doing so, it increases the accuracy and ultimately intelligence [of the generative AI solution like ChatGPT]. And generative AI needs to generate those accurate results quickly,” says Brian Gallagher, executive consultant at Global Solutions. He was speaking at the Innovative Data Infrastructure Forum, which is part of the Huawei Digital and Intelligent Apac Congress in Bangkok, yesterday.
He also shares three major challenges banks will face as they adopt generative AI. The first is making huge datasets visible, manageable and available for the data collection and preparation phase of AI. “Data is the fundamental DNA of intelligence. Most financial institutions have geographically distributed data that are both on-premise and [in the cloud]. If you have accessibility or availability issues with your data, you will not create intelligence.”
Besides that, financial institutions might struggle to ensure AI cluster availability when they train their AI models and find it challenging to enhance their resilience against cyberattacks.
Three zeros for data resilience
Huawei is helping financial institutions address those challenges by strengthening their IT infrastructure to enable “three zeros”. “We believe there are three pillars of data resilience for financial institutions: Zero-touch operations, zero wait and zero trust,” says Steven Lee, senior solution architect of Huawei’s Asia Pacific Enterprise Data Center Marketing and Solution Sales department, at the same event.
See also: Alibaba anoints new chief in revamp of stalling commerce arm
Simplifying the data storage infrastructure is key to enabling zero-touch operations. Huawei does so by offering a unified portal that allows financial institutions to access data across their data centres and devices for the development of AI models.
Huawei’s storage solutions also provide automatic data tiering of hot and cold data to improve data scheduling efficiency. “When you do data modelling for AI, you’ll focus on hot data but once that’s done, storing cold data in fast, high-performance storage will be costly. So, our unified data scheduling technology can automatically move your cold data to an archive storage [for cost-efficiency],” states Lee.
Zero wait calls for AI cluster availability, and Huawei believes the checkpoint synchronous read/write feature in its storage solutions will deliver that capability. Lee explains: “There are different stages of AI/machine learning, including training, evaluation, export and deployment. By having checkpoints at various stages, you can restart the AI modelling from the point that experienced data disruption instead of restarting the entire process.”
See also: Break up Google? What’s at stake in antitrust action
With the rise of ransomware, embracing zero trust is vital for data and business resilience. “[Huawei is helping financial institutions on that front by] building a highly efficient detection engine into our storage solutions. We also offer network-storage security integration, wherein the network quickly notifies storage once ransomware is detected so that storage can take immediate actions to protect the data,” claims Lee.
He also highlights that the new built-in bait file solution in the Huawei OceanStor Dorado All-Flash Storage will enable more proactive cybersecurity. Designed to attract viruses and malware, bait files are used to help detect and protect against viruses and malware. “Once a ransomware threat is detected, the bait file solution will respond to it. This does not mean we’re deleting that ransomware because we’re not a cybersecurity company. Instead, the solution will create an immutable snapshot to contain the damage of the files.”
“In short, Huawei is making data infrastructure as seamless as possible through zero touch, giving financial institutions the confidence to train their AI model reliably via zero wait, and helping them protect their data and AI models with zero trust,” concludes Lee.
