Sarah Hoffman, Director of Research, AI at AlphaSense, speaks with Martin Couturier,  Director of AI Innovation at Saige Consulting, in a video interview about integrating GenAI with other technologies, her take on artificial general intelligence, the role of transparency in AI, prioritizing human users, and measuring the strategic impact of GenAI tools.

Hoffman states that it fascinates her when people discuss GenAI; the focus is often on next-gen models — how these will be better, more accurate, and have fewer hallucinations. On the contrary, for her, the most significant breakthroughs have come not from improving LLMs themselves but from integrating those with other technologies.

As an ardent follower of this space since the launch of GPT-3 in 2020, Hoffman notes that the biggest impact was created by combining language models with search engine technology. This integration addressed two major challenges — the lack of real-time information and the problem of hallucinations. Real-time information became accessible, and hallucinations could be cross-checked against sources, significantly mitigating those issues, she adds.

Speaking of artificial general intelligence (AGI), Hoffman says that it has been a hot topic lately but believes that it is far from being achievable. AGI refers to creating AI that is as intelligent as humans; however, even with the latest models, there are moments when it says things that are entirely incorrect.

Having said that, Hoffman says that what already exists looks a lot like human intelligence, and is improving steadily and making fewer mistakes. As it continues to evolve, it will become even more indistinguishable from human intelligence in practical applications, she remarks. The focus should be on how to best utilize the AI that is available today rather than worrying about AGI’s timeline.

According to Hoffman, having transparency around AI is critical to making sure people know when they are interacting with AI and helping them understand how it works, at least to a certain degree. Beyond transparency, having diverse groups involved in creating these technologies is crucial.

Another key point is not losing sight of the human user during the development process, says Hoffman. She shares that AlphaSense prioritized talking to users to understand how it works for them when releasing its generative search tool.

Consequently, some users reported saving anywhere from 11 to 50 hours a month with the tool. The focus is on creating something truly helpful — something that enables users to do their jobs better, whether they are in investment roles, corporate strategy, or consulting.

Moving forward, Hoffman states that while one can track the statistics on the time saved by using GenAI, it is harder to measure the strategic value these tools add. For instance, beyond saving time, they might uncover insights users would never have found otherwise—simply because they couldn’t have accessed them using traditional methods.

Hoffman states that assessing strategic impact might involve setting up scenarios where users try their traditional search methods first, and then compare those results to what they uncover using the generative tool.

Another angle might involve looking at questions users wouldn’t have thought to ask before, like identifying overlooked risks in an investment strategy or spotting hidden opportunities, she says. Measuring this would require having the right questions and the right kinds of users who are open to exploring.

In conclusion, Hoffman states that there are significant insights that could be uncovered now by asking follow-up questions and things that were not possible with traditional search.

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