your123 - stock.adobe.com
SambaNova unveils GPT Banking for financial industry
The product is aimed at giving financial institutions deep learning language capabilities. However, deep learning is hard to navigate in the finance industry.
SambaNova Systems, an AI hardware and software vendor, on Wednesday unveiled GPT Banking, a deep learning system product for the financial services industry.
The new product comes after the vendor in October added a GPT (generative pre-trained transformer) language model to its Dataflow-as-a-Service product.
GPT Banking
GPT Banking enables banks to create large language models for applications such as: translation; language generation for processing, transcribing and prioritizing claims; reducing human error and manual repetitive work; and understanding market, investor and stakeholder sentiment.
SambaNova said GPT Banking will help close the gap in deep learning deployment in banking and help financial services companies build and deploy AI models faster. The vendor claims that with the new product, banks will be able to deploy advanced language models within days or weeks.
Deep learning challenges
While this implementation of GPT in finance is promising and new, challenges with deep learning may be hard for enterprises to address successfully, said Gartner analyst Moutusi Sau.
"Market wise, commercial growth wise, this is definitely very big … because of this language capability," Moutusi said, adding that the technology will be beneficial in non-English-speaking countries because of its translation feature.
"Deep learning done in such a massive scale has been very hard to achieve for most of the financial institutions. And that's why [the financial sector] generally stays away," she continued.
Moutusi SauAnalyst, Gartner
The difficulty is mainly in achieving explainable AI and explainability of deep neural networks.
Since deep learning and neural networks try to model how the human brain comes to a specific decision, processes involved in creating the models can be hard to explain to those using the models.
Neural networks also deal with unstructured data, which is harder to explain than structured data, Moutusi added. There are not many examples of neural networks being used in the banking and investment sectors, she said.
Therefore, while SambaNova's GPT Banking seems promising, the results will determine whether it will succeed.
"It's not a black box," said Marshall Choy, senior vice president at SambaNova, in response to whether GPT Banking will provide explainable AI.
SambaNova works with banks and others in the financial industry to make sure the vendor's models and services can be explained to regulators and internal auditors, Choy said.
Also, SambaNova separates itself and its service from customers' data, and therefore never sees customers' data, he continued.
"That's one of the safeguards against things like privacy and other things that are of concern to regulators," Choy said.
GPT and pre-built models
While GPT is a relatively new technology, using pre-built models is not new, said Mike Gualtieri, an analyst at Forrester. "Natural language processing has taken a leap with GPT models. I expect many companies to productize it for specific use cases and industries just as SambaNova has done for banking."
SambaNova GPT Banking is available now as a subscription service. The vendor did not disclose pricing.