The Bloomberg Terminal has long been inspired by its notable tenacity devotion, bordering on obsession. Among traders, the ability to carve a path through the software’s dizzying scrolls of numbers and text to isolate distant information is the mark of a seasoned professional.
But as more data is fed into the terminal – not just about earnings and asset prices, but also about weather forecasts, shipping logs, factory locations, consumer spending patterns, private loans, and so on – valuable information is lost. “It’s becoming increasingly unsustainable,” says Shawn Edwards, Bloomberg’s chief technology officer. “You miss things or it takes too much time.”
To try to remedy this problem, Bloomberg does this chatbot-style interface testing in the case of the ASKB terminal (pronounced “ask-bee”), built on a basket of different language models. The overall idea is to aid financial professionals condense labor-intensive tasks and enable them to test abstract investment thesis against data using natural language prompts.
At the time of publication, the beta version of ASKB is available to approximately one-third of the software’s 375,000 users; Bloomberg did not provide a full release date.
In early April, WIRED spoke with Edwards at Bloomberg’s palatial London headquarters. We discussed the impetus to modernize the Terminal, the possibility of traditionalist opposition to the change, and Bloomberg’s attempts to eliminate hallucinations.
The following conversation has been edited for length and clarity.
WIRED: Shawn, tell me about the reasoning behind the terminal renovation.
Shawn Edwards: Over the years, Bloomberg has continued to expand the extensive data set at its disposal. Often, finding the right data in a sea of information determines success or failure. This becomes increasingly unsustainable: you miss things or it takes too long.
The fundamental problem we solve with generative AI is helping users find key insights and synthesize a worldview around a specific idea.
The concept is that untapped alpha is lurking somewhere in the data and ASKB will aid bring it out?
Yes. The user can ask a high-level question – a thesis that is in their head – rather than asking about specific data points. “How will the war in Iran and the change in oil prices affect my wallet?” This is a very, very crucial question with so many dimensions. Can we synthesize this answer in a few minutes?
In a scenario where anyone can wade through the maze of data, what will separate average traders from the best?
These tools are not magic. They don’t average [employee] suddenly great. The difference will be your ideas.
In the hands of experts, it allows them to do better analysis and deeper research – review 10 great ideas when they might only have time for one. If you are an average analyst, these will be 10 average ideas.
Bloomberg presents ASKB as a form of agent-based artificial intelligence. On the surface, it looks more like a chatbot interface than something that necessarily automates tasks. What is agentic about ASKB?
There are earnings that come out every quarter. My job as an analyst is to prepare for what may come up on the earnings call. For every company I prepare for, I compare their prices with other companies, search many documents, analyze their fundamentals, and so on. I don’t sleep during the earning season.
With ASKB I can create workflow templates. I can write a long query and say, “Hey, here’s all the data I’m going to need.” Please give me a summary of the bull and bear cases, what the Street says and what the guidelines are. Now I want to plan [the workflows] or I trigger them when I see this or that state of the world.