Stay Informed with Free Updates
Sign up to the Artificial Intelligence myFT Digest — delivered directly to your inbox.
Oh, Deutsche . . .
The Future of Financial Analysis
By “you” they mean “us”.
Deep Research Widget
OpenAI’s launch on February 2 of a so-called “deep research” widget (currently available only to subscribers on the $200 per month plan) provides the report’s hook.
The Power of Iterative Reasoning
In an attempt to mimick logical reasoning, “deep research” responds to prompts in iterative steps. It’s part of a coming generation of AI bots with long-term memory, internet access, and the power to keep improving autonomously, says analyst Adrian Cox, who calls it a step towards artificial superintelligence.
The Future of Human Involvement
“To paraphrase an old expression: give a model a prompt and you feed it for a day; teach it to prompt and you enable it to come to life,” he tells clients.
The Impact on Financial Analysis
And in a world of self-improving, autodidactic logic engines, the role of the Deutsche Bank analyst human will be reduced to prompt engineering and proof-reading:
Much of the rest of the cognitive process will be offloaded. This will favor those with the least experience, whose performance will be instantly lifted to a new minimum level, and those with the most, who will know where to focus, identify the value (and errors), and make the most of the output. Those in the middle will struggle to add value.
Example:
In our own tests [ . . .] deep research took eight minutes to produce a 9,000-word report, citing 22 sources (with links), assessing the impact of new US steel and aluminium tariffs. It’s not perfect, but would improve with iteration and is a step-change above a standard ChatGPT response in analysis, relevance, and clarity.
Limitations and Future Developments
Its choice of largely credible sources and focus on attribution seems to reduce hallucinations. Even so, it is not entirely able to shake off the inherent weaknesses of generative AI, such as a lack of context. In our tariff example, we needed a second attempt to get it to address the most recent round of tariffs (but it then did so successfully).
Conclusion
These are obviously still early days for deep research tools. Both they and the underlying LLMs that power them will continue to improve. And only time will tell how far they will eventually move beyond rehashing other people’s ideas to generating entirely new concepts altogether.
FAQs
Q: What is Deep Research?
A: Deep Research is a new widget launched by OpenAI, which responds to prompts in iterative steps, mimicking logical reasoning.
Q: How does it work?
A: It’s part of a coming generation of AI bots with long-term memory, internet access, and the power to keep improving autonomously.
Q: What are the limitations of Deep Research?
A: Its choice of sources and focus on attribution reduces hallucinations, but it is not entirely able to shake off the inherent weaknesses of generative AI, such as a lack of context.
