There's a lot more to GPT-4 than ChatGPT. Here's what it means for advisors

OpenAI announced in March that their next generation technology for analyzing language data, ChatGPT, would be available for purchase by paid subscribers of ChatGPT, who are charged $20 a month for the service. ChatGPT itself is an impressive service, but it pales in comparison to GPT-4, the next generation technology.

‍

A simulated bar exam performed by GPT-4 offers a very high level of success. ChatGPT remains in the bottom 10% of test takers, whereas GPT-4 has posted outstanding scores on exams like the GRE, SAT, and LSAT, which is outstanding by any standard. 

‍

This is a powerful system that has the capability to understand and recognize text as well as create images. BPT-4 was able to create a live website using a handwritten sketch in an OpenAI demonstration.

‍

It has been suggested that GPT-4 has shown some “sparks” of AGI (Artificial General Intelligence), a level of intelligence close to human capabilities, in a recent paper published by Microsoft researchers. It is reported in their paper that the GPT-4 is capable of solving new and complicated problems that span mathematics, coding, vision, medicine, law, psychology, and more, and this is without the need for any special prompting. Further, GPT-4 has been consistently shown to be superior to previous models like ChatGPT, when it comes to all of these tasks, and has performed this task at human-level performance on every one of them."

‍

Despite the fact that this technology appears to be poised to affect many industries – including financial services – in the near future, it appears to be transforming these industries entirely. With GT-4, advisors will be able to trigger analytical tools or access long-forgotten but still relevant research within seconds, and it is expected to completely transform the advisory business as well. The CEO and co-founder of Toggle AI, which develops generative AI technology for investment, Jan Szilagyi claims that GPT-4 is going to revolutionize the advisory business. As more and more financial firms start to adopt this kind of technology, it does not seem as if a question of "if, but when," will be raised.

‍

There are many new applications in the financial industry that ChatGPT is useful for, for instance to brainstorm blog ideas or to help with the writing of emails for financial advisors.  

‍

This system offers even more than GPT-3. It will be able to output text that contains more than 25,000 words. It will also be able to add interactive charts and diagrams to the text. In the case of adequate data, a long-term financial plan, for example, may be constructed that correlates with a large percentage of desired outcomes. However, that is just within the limitations of this model, within the limitations of the field. The global chief information security officer of LexisNexis Risk Solutions, Flavio Villanustre, claims that past performance cannot guarantee future results. As a result of GT-4, financial forecasts that rival those of experts could be created.

‍

There are a few companies in the world that are strategic partners of OpenAI's GPT-4 system, and Morgan Stanley is one of them. Morgan Stanley Wealth Management's head of analytics, data, and innovation, Jeff McMillan, has stated that the bespoke financial advisor solution that we are developing with OpenAI will be one of the first of its kind in the financial services industry.

‍

An advisor from Morgan Stanley will be able to ask open-ended questions to this chatbot. A chatbot will ingest internal research, commentary, and analysis, which are mostly in PDF format. For example, you might be interested in understanding the earnings outlook for the S&P 500, alternative investments for a particular kind of investment, how certain investments perform when inflation spikes, or what the difference is between a retirement account and a traditional one.

‍

However, Morgan Stanley is not stopping there. The firm has scheduled a firm-wide hackathon to come up with new applications and ideas. They are also experimenting with how the system can be used to improve client communications.  

‍

There is still a problem with GPT-4, even though the improvements have been made, “which means GPT-4 still seems to be hallucinating,” says Muddu Sudhakar, the CEO and co-founder of Aisera, an AI startup that specializes in generative algorithms. This means that the model may sometimes produce false or misleading output. In the financial world, accuracy is of utmost importance, so it is essential that safeguards be put in place.”

‍

There is also a problem with the topic of data residency as well as data security. GPT-4 requests that information be delivered to OpenAI using an API (Application Programming Interface), which requires that data reside in OpenAI.  

‍

It is for this reason that financial services companies might have a hesitancy to use private client information in order to adjust their models. As an example, Morgan Stanley has already demonstrated this to be true.  

‍

A highly complex and regulated environment like ours makes it challenging to integrate technology, and McMillan points out that it is a very challenging process to do so with the appropriate controls in place. We are fortunate to have a proven track record in this area, but it is important to clarify that this is a brand new and exciting endeavor for us.”

‍

Fortunately, GPT-4 is not the only open source solution emerging to compete with it. Databrick, for example, has produced Dolly, a small, but capable open source solution, which is open to the public, and so is free to use. The key difference is that a customer has complete transparency with the model and can house it at his own data center.

‍

It is estimated that 19% of employees could see their tasks automated at least 50% of the time through GPT-4 according to a study from researchers at OpenAI and the University of Pennsylvania, which looked at the potential impact of GPT-4 on the workforce.

‍

Would GPT-4 eventually be able to replace the role of a financial advisor in the future? This seems very unlikely. Besides providing analytical solutions, financial advisors possess a number of skills that machines do not possess, including empathy. Furthermore, it seems unlikely that licensing organizations will allow GPT-4 to perform the role of a financial advisor in the future.  

‍

As a result, this technology will help to improve productivity as well as service levels, as well as help to make it a powerful assistant for companies. As a director of Confidential Computing at Fortanix, Richard Searle summarized the situation as follows: "With the use of GPT-4 technology in massive, centralized AI services, financial advisors are able to offer a better level of guidance to their clients by allowing the AI system to evaluate data volumes that may be too large for a human analyst to analyze."