OpenAI's New Reasoning Model & Hong Kong Explores New AI Regulation
Here's what's happening in AI x Banking this week.
Hey everyone! It’s already the middle of September—how time flies. We’ve been adding more team members to Team Greenlite and getting ready to go to some conferences over the next few weeks.
The Greenlite team will be in Las Vegas for the ACAMS conference next week. If you’re around, would love to meet you all there! Feel free to send us an email to schedule a meeting.
We’ll be at Money 2020 too—stay tuned for that.
Tons of news to get through this week—let’s dive in!
AI Roundup: OpenAI Releases O1 Model
OpenAI’s newest model helps with “reasoning”
OpenAI, maker of the popular ChatGPT, released a new upgraded model for the general public they call “O1.” The O1 model and the O1-mini model are designed to enhance reasoning and problem-solving capabilities in complex tasks, particularly in science, coding, and math.
These models are trained to think more deeply before responding, refining their strategies and learning from mistakes. In tests, they performed comparably to PhD students. The system shares its thinking process and how they’re breaking down each problem into structured steps in the chat function, which helps users understand how the LLM is breaking down the problem.
It’s a significant advancement in AI models—before, LLMs developed answers based on statistical patterns, not true understanding of what they’re being asked to answer. On top of that, previous LLMs produced answers in a single step and didn’t go back to refine or correct their thought process.
People are already making a huge impact with the new O1 model. People have been using it to create code and apps, as well as answering complex academic questions.
OpenAI’s O1 model is only available to ChatGPT Plus users (who pay $20 a month) and Teams, while Enterprise users will be getting it soon. Developers can start testing out new products with the API, but it’s quite limited. Weekly limits for ChatGPT users are only 30 messages for o1 and 50 messages for o1-preview.
Hong Kong Explores New AI Regulations Around Finance
Hong Kong, a hub for international finance, is moving quickly to set up AI guidelines for uses within financial services, according to Bloomberg. The report says that Hong Kong’s government is working with industry leaders to put out a framework that encourages the use of AI in banking.
The government is expected to announce the framework at the end of October, during the city’s Fintech Week. The policy guidelines are expected to cover everything from individual banking to securities and wealth management.
Hong Kong sees new technology as a way to attract more financial services business to the hub. Bloomberg writes that foreign investors are becoming more apprehensive of operating in Beijing given China’s strict rules, and Hong Kong operates with a separate government.
This wouldn’t be the first new tech that Hong Kong embraces—it’s also been a leader in crypto and fintech frameworks as well. But what’s notable is that it’s one of the first guidelines around AI in finance in the world, and it seems that it’ll have a positive take. Since governments are actively exploring how to regulate AI, if at all, we expect the Hong Kong AI policy to be a model that other countries look to guide them in future regulations.
Community Bank Employees Are Becoming AI Fans
A great new report from American Banker this week, outlining how community banks are helping employees benefit from new AI tools.
While employees might be hesitant at first—worrying about AI taking their jobs or making it more complicated—it seems like after seeing real use cases or getting sufficient training, AI adoption among employees jumps, according to a panel at the Independent Community Bankers of America forum last week.
Mercantile Bank turned on live chat functionality for small business customers, and while employees were initially worried about what questions they would get, it’s turned into a very helpful resource for customers.
Bankwell Bank in New Canaan, Connecticut, piloted a new AI platform to standardize policies for small-business lending. The platform uses large language models (LLMs) to analyze and improve policies by identifying gaps and conflicts. After testing the technology, the bank found that it was 80% to 90% accurate when asked about SBA (Small Business Administration) policies. Chief Innovation Officer Ryan Hildebrand told American Banker that "Some of the other employees at the bank may be nervous. But they are going back and double-checking and realize it's the right answer. It's starting to build into the culture that we can trust this stuff."