(US & Canada) Sarah Hoffman, Director of Research, AI at AlphaSense, speaks with Martin Couturier, Director of AI Innovation at Saige Consulting, in a video interview about her role, how generative AI (GenAI) is different from traditional AI, GenAI challenges, and how companies can benefit from current and future AI tools.

AlphaSense helps enterprises remove uncertainty from decision-making. Its universe of public and private content includes equity research, company filings, event transcripts, expert calls, news, trade journals, and clients’ research content.

Shedding light on her role, Hoffman says it revolves around helping customers understand the role of GenAI. Having led AI efforts for over a decade, her prior roles included leading AI research for the Fidelity Investments Innovation Department and leading FactSet’s ML team.

Highlighting the distinctive features of GenAI compared to earlier AI models, Hoffman points out the creativity aspect. She notes that while earlier AI models were also enablers, GenAI is interesting because it generates, in addition to increasing efficiency.

As a regular user of GenAI, Hoffman says that there is also the element of strategy. For instance, she uses the technology to ask questions and then ask follow-up questions, making it a discussion. Hoffman recalls an instance when ChatGPT helped her with a report on healthcare by suggesting she include the aspect of mental health and how GenAI is changing the mental health landscape.

Speaking of challenges, she says that one of the biggest challenges with Gen AI is that it hallucinates, and more so, it makes things up. Additionally, it does end up giving different answers to the same query.

However, Hoffman finds it interesting that with each new version of the technology, there are fewer hallucinations. Commenting on the different ways to use this technology, she says that businesses can reduce hallucinations.

Delving deeper, Hoffman says that AlphaSense, like many other companies, uses the Retrieval-Augmented Generation (RAG) approach with its GenAI systems. Through this, the company ensures the system uses the organization’s trusted content to answer questions.

With this approach, the answers can be trusted more, says Hoffman.

Another thing that AlphaSense does is ensure that the technology links back to the source document for every answer it provides. This significantly reduces hallucinations, as the source can be traced and users can verify the original material and understand the answer’s context.

This transparency is crucial, especially when the technology is being used for high-stakes applications like making multimillion-dollar investment decisions, says Hoffman. That said, she believes that in some cases, hallucinations can be acceptable.

In her role as an AI researcher, Hoffman appreciates when tools like ChatGPT present ideas that are imaginative or speculative. She maintains that even a made-up response can inspire creative thinking and new perspectives.

However, it boils down to the use case, says Hoffman, as predicting potential developments in technology is different in context from making critical investment decisions. Moving forward, she states that for companies to benefit from current and future AI tools, it is critical to not only allow employees to use the technology but also train them on using it right.

According to Hoffman, the training should include real-life examples that highlight both pitfalls and benefits of AI. For instance, there are cases where lawyers have used AI tools without realizing that the technology can fabricate citations. Showing such examples can demonstrate how not to use the technology, emphasizing the importance of not relying on it blindly.

Conversely, training should also showcase positive use cases, illustrating how the tools can save time and boost efficiency, says Hoffman. A grave and common challenge, she says, is that users do not understand the limitations of the technology.

For instance, Hoffman recalls someone who altered their investment strategy based on ChatGPT’s suggestions without knowing that the model then was using outdated data. Effective training can bridge this gap, she says.

Thereafter, Hoffman encourages employees to use this technology for personal projects as well. Individual usage of AI can help them decide how it can be applied effectively in professional settings and where it would not fit.

In conclusion, she says, engaging with the technology is critical to reimagine roles and notes that this technology will transform the roles of knowledge workers.

CDO Magazine appreciates Sarah Hoffman for sharing her insights with our global community.