On Tuesday, Morgan Stanley Wealth Management announced they are just “one of a handful” of organizations receiving these Early access to OpenAI’s new ChatGPT 4 offering.
The news coincided with that of OpenAI Announcing that the next iteration will be released its hugely popular ChatGPT-3 artificial intelligence model.
ChatGPT-4 is still not widely adopted, and Morgan Stanley is “currently the only strategic client in wealth management” that can currently use it, according to the company.
“Built specifically for and by Morgan Stanley with appropriate controls, financial advisors and their teams will leverage the in-house ability to ask questions and view large volumes of content and data, with answers delivered in an easy-to-digest format sourced exclusively from MSWM content and with links to the source documents,” the company said in a statement.
Morgan Stanley also has other ongoing AI projects, including ones focused on advisor workflow and communicating with clients and prospects.
William Trout, director of wealth management at Javelin Strategy & Research, said Morgan Stanley “has a track record of leveraging digital tools and data to empower the financial advisor.”
“The integration of the OpenAI GPT-4 chatbot follows the implementation of Next Best Action capabilities into the Advisor workflow, built upon years of effort by Jeff McMillan (Chief Analytics and Data Officer) and Morgan Stanley to bring them to the table Alignment of massive amounts of customer, market and reference data,” Trout said in an email wealth management. com. “The launch of this internal chatbot promises to further break down information silos in the wirehouse. Connecting the data points will both increase consultants’ efficiency and enable them to identify new opportunities for the client. The effect will be to dramatically improve Morgan Stanley’s financial advisor.”
This is just the latest in a long line of announcements from wealthtech companies claiming to integrate ChatGPT into their applications. Companies like FMG, Orion, and Broadridge have already started integrating ChatGPT, while others are taking a more cautious approach.
“We’re not currently using Open AI formally, but are in the process of exploring its possibilities in our SEO efforts and in our blog and newsletter content,” said Rishi Bharathan, the CEO of WiserAdvisor, a platform for prospective advisors. “Because we primarily produce consumer-facing content that our users engage with as they make financial decisions, we must be extremely diligent before publishing content.”
Meanwhile, unsurprisingly, AI was a hot topic at the Technology Tools for Today (T3) conference in Tampa, Florida.
Raj Madan, head of technology for wealth management platform AdvisorEngine, pointed out during a panel discussion that machine learning is a subset of AI, but it needs a lot of data to be useful.
“The whole idea of machine learning is more programmatic,” Madan said. “It uses algorithms to look at sequences and patterns and use those sequences and patterns to make some kind of predictions.”
Madan said his first ML project 13 years ago failed.
“And it taught me a great lesson,” Madan said. “Machine learning doesn’t work without a wealth of data.”
Machine learning applications were limited because they really analyze patterns, observe those patterns, and then make predictions, Madan said — good for things like credit card fraud monitoring, but of limited use in situations where data was scarce or generated infrequently.
The breakthrough, Madan said, was the arrival of generative AI.
“Now we’ve gone from small predictions to actually creating text or images,” Madan said. “ChatGPT is a chatbot. You give him a request. It gives you content back.”
ChatGPT uses the discipline of natural language processing, which draws from “a very large corpus of material,” Madan said. ChatGPT-3 uses around 45 TB of data, including books, articles and websites, and creates an extensive language model.
“(This) ultimately allows ChatGPT to … go out and create a complex sentence based on the patterns it understands from all the material it’s read,” Madan said. “Everything in machine learning is math. It’s all statistics. ChatGPT uses probabilities to figure out how to complete a sentence.”
ChatGPT uses a library called Generative Pre-Trained Transformers under the covers. Transformers are often referred to as “neural networks”. Madan said since 2018 when Google came up with the idea For a wider audience, Microsoft, Google, and OpenAI “have been in an AI race to essentially see who could produce the most sophisticated solution for this transformation algorithm.”
Madan said that ChatGPT-3, which became widespread in November 2022, is an example of “the singularity”. This idea was first suggested by Dr. Ray Kurzweil, and is defined as the moment when “technology will be so advanced that growth will be exponential,” Madan said.
Madan said that another milestone will come in a few months when ChatGPT-4, which is many times more powerful than ChatGPT-3, becomes available to the general public.
So how can consultants use ChatGPT-3 in their own practices? Madan gave examples of customer communication and coding, especially with Excel, at the beginning.
But that doesn’t mean it will replace the role of advisor.
“ChatGPT and all the chatbots out there essentially mimic your question,” Madan said. “If your question isn’t right, you won’t have the right answer. It just knows there’s a pattern out there. That’s all it really understands. We will not replace anyone with this technology. This technology will be an assistant to the advisor. It will help them get a head start or a leg up.”
However, there are many concerns for consultants looking to use the technology in their businesses, Madan said. This includes so-called “hallucinations”, where chatbots make mistakes and give information that not only cannot be true, but are absurd statements that defy logic.
“It’s all about probabilities and statistics,” Madan said. “They will not let ChatGPT run wild. You will have some supervision. Especially if you work in the corporate sector.”
Madan said that using unrestricted AI for customer communications could generate content that easily violates SEC marketing rules and FINRA 2210 rules.
And then there’s the problem of bias that can be added to AI without users or programmers even realizing it.
“ChatGPT takes a range of data and takes a range of texts, books, articles, newspapers and essentially processes them and creates this model,” Madan said. “The problem is, what if the content produced is biased? If it’s biased, what will happen to the model? Are there people in this room who are not biased? No. …And developers can actually program bias inadvertently, not intentionally.”
Copyright and plagiarism are also concerns for those who use AI to create content.
“The fair use doctrine says you can use copyright for certain things. You can use it for research. Does ChatGPT conduct research?” said Madan. “These are some of the legal issues that arise here.”
Another concern is that ChatGPT-3 stopped “training” in 2021, meaning it has stopped adding new content to its network.
“So it’s been two years since it stopped training,” Madan said. “What if you have content that was out of copyright before 2021 and is now in copyright? That’s going to be a problem.”
Madan said many of those fears would dissipate over time, however.
“It’s all about the maturity of this area,” he said.
Despite this, Madan said, “ChatGPT requires oversight and cannot be trusted blindly. You can’t overdo it.”