Machine understanding artificial intelligence

At the Future Work Summit held on January 12, 2022, listen to the opinions of CIOs, CTOs, and other C-level and senior managers on data and AI strategies. to know more information

This may be the first time that Eastern European Bank is cooperating with the Hungarian government to deploy artificial intelligence supercomputers to create a large-scale Hungarian language model.

OTP Bank was established in Hungary and operates a bank in the region. The bank made an agreement under which the government provided approximately half of the funding for the supercomputer developed under the contract with SambaNova Systems. Péter Csányi, deputy chief executive officer and head of the digital department of OTP, stated that the government will be able to use the system for public and academic research.

The contract with SambaNova Systems utilizes the generative pre-training converter (GPT) announced by SambaNova in October to generate large language models and mark them as data flow as a service. SambaNova product vice president Marshall Choy said: "Building such a system to run the GPT model has not been done by any bank in the past."

Csányi said the bank believes that it is in its own interest to create artificial intelligence applications for the Hungarian language in an increasingly digital economy. "Our ability to adapt to a changing world is one of the core capabilities we need to introduce internally."

In addition, the OTP bank and the government agree that if you do not take the initiative, Hungarian is unlikely to obtain its own AI language model in the short term, because it is a difficult language to learn and few people can speak it. He said that the experience of working on the project will also equip the bank with expertise that can be used to generate language models for the languages ​​of other countries in the region.

SambaNova said that this artificial intelligence supercomputer will become the fastest computer in Europe and is scheduled to be put into use in 2022.

"We have been observing the trend for several years, and things will definitely move towards larger and larger models, which require larger and larger resources similar to AI supercomputers to build. This is why we made them into products," Choy said. SambaNova's products allow companies to create their own GPT models as an alternative to shared resources such as GPT-3 that rely on OpenAI.

"We have also greatly reduced the burden on Péter and his team to hire hundreds of data science and machine learning professionals. In fact, he will be able to do this with a dozen people," Choy said.

The GPT method for creating large-scale language models uses AI deep learning technology to enable software to discover patterns in the language without explicitly training them. This is the "generating" part that allows the software to write its own rules based on the analysis of large amounts of textual content from the language. Although large language models speed up development and reduce the labor required to generate models, they also have shortcomings, such as a tendency to incorporate biases and errors derived from the text they use.

Regardless of their flaws, a recent state of artificial intelligence report found that large-scale language models are so important that many countries are "nationalizing" their research to avoid being left behind.

Choy said that companies are also focused on keeping up. "We believe that artificial intelligence will have a restructuring effect on all industries, similar to the Internet 20 years ago."

Csányi said that when his IT team members first proposed this idea to him, he was skeptical, believing that the budget and talent needed were beyond the bank's capabilities. "What SambaNova did was to provide it at a reasonable cost," he said. Although cultivating IT talents to use the AI ​​model will be a challenge, he believes that the bigger challenge will be to prevent his employees from devoting all their time to it and ignoring the daily operational needs of the bank.

Csányi said that OTP Bank hopes to use natural language processing applications for customer service, fraud prevention, loan origination, and network security, and for purposes that may not become obvious before the technology goes into production. "We have no shortage of use cases."

AWS Activate provides startups with free tools, training, etc. to help you build and scale quickly-in addition, you can earn up to $100,000 in Activate points.

Hear the opinions of CIOs, CTOs, and other C-level executives on data and AI strategies.

© 2021 VentureBeat. all rights reserved.

We may collect cookies and other personal information from your interactions with our website. For more information about the categories of personal information we collect and the purposes for which we use them, please review our collection notice.