What are the limitations of using AI and machine learning in Investment Banking?

There is much overhype about data and what you need. Tremendous value can be gained by doing things as simple as downloading economic data from (Federal Reserve Economic Database) FRED and many other free sources. Using that data to analyze how businesses react to varying economic conditions can create more value than most other AI projects combined. We do this with our retail banking clients by looking at close to 100 local and national inputs and they are commonly shocked to find out what really drives their businesses to change. AI is like anything else if you pick the right project getting data can be simple and provide shocking value.

#If your data isn’t ready for Generative AI, your business isn’t ready for Generative AI! Banks should first develop a comprehensive data strategy based on the business use cases they plan to explore Generative AI solutions for — the data strategy should include structured, semi-structured, and unstructured data spanning banks internal data, industry third-party data, public records, news feeds, market & economics data, streets & customer’s earning calls data, SEC filings, market research, etc. Additionally, banks should plan for redesigning and preparing their data architecture and intelligent data platform for LLM data (e.g., chunked text and vectors) management, speed of access and computing.

It is good to see my long-time friends Kumar and Saroj chiming in here. Saroj is not wrong in how to set everything up for sustainability. It's just that so many are struggling to get started with AI they need to walk before they can run. Once you have some solid legs then you can put in the proper infrastructure to support a sustainable process.

Agree. The simple steps are always the best in EDA. 95% of the time I find I can augment the model wit Fed or data.gov as my data source. This time for open data. There is so much useful data in these sources. You just need a repeatable approach to use the, in conjunction with your organization own data.

Uma das limita??es mais significativas: Privacidade e Seguran?a Prote??o de Dados: O uso de IA e ML requer quantidades significativas de dados, e levanta preocupa??es sobre a privacidade e a seguran?a dessas informa??es.

Regula??o e ética Quest?es Regulatórias: A regula??o em torno do uso de IA e ML ainda está em desenvolvimento. Os bancos de investimento precisam garantir que est?o em conformidade com as leis e regulamentos. Viés e Discrimina??o: Algoritmos de IA podem reproduzir e amplificar preconceitos existentes nos dados, resultando em decis?es discriminatórias. é importante mitigar esses viéses e garantir que a IA seja usada de forma ética.

Considérez l'IA comme un collaborateur, non pas comme un rempla?ant. Elle peut traiter et analyser des quantités de données inimaginables à une vitesse et avec une précision que l'esprit humain ne peut égaler. Cependant, elle ne possède pas votre capacité à interpréter ces données dans un contexte complexe, à comprendre les subtilités des relations humaines ou à faire preuve de jugement éclairé face à des situations inédites.

I believe this is a bit misleading... While this can be true if you're trying to build a custom model from scratch, in reality there are many cheaper ways of getting high leverage from AI applications. Using pre-trained models, public data, and even Large Language Model applications can all be utilized for a fraction of the cost of building custom models and can still provide incredibly high value. Another thing to weigh is how much money you're actually saving by augmenting your team with better data and analysis. If the model costs a firm 100K/yr but generates 1M in value, then it's a no brainer.

Lorsque vous déployez des technologies d'IA, il est impératif de maintenir une communication claire sur la manière dont ces outils fonctionnent et prennent des décisions. Les clients sont-ils en mesure de comprendre et de faire confiance aux recommandations fournies par l'IA ? Une absence de transparence peut engendrer des doutes et de la méfiance, nuisant ainsi à la relation que vous cherchez à batir ou à maintenir.

AI is fundamentally a multiplier If you have a bunch of valuable data, AI can help you create 100x value by sitting on top of that data. And this is the key point here: data. I could deploy the exact same model and application to 100 firms, but the results will not be the same. The firms with better data and engineered data will win. That's the competitive advantage. I don't believe AI will erode any advantages, the firms that are able to build on top of better data and thesis will continue to win and now will just be winning harder.

Em resumo, embora a IA e o ML ofere?am muitos benefícios para o Investment Banking, é crucial estar ciente das limita??es e desafios associados. Os profissionais devem usar essas tecnologias de forma consciente, complementando e julgando os insights gerados pela IA para tomar decis?es informadas e éticas.