An interview with investment bank J Safra Sarasin’s chief sustainability officer Daniel Wild about their experiences using chatGPT in ESG was just published by ESGClarity. Wild emphasizes the need for human involvement, assessment and ultimate decision making rather than deferring wholly to chatGPT – which is very good advice:
“We can summarise a very long report, hundreds of pages, and have a nice extract of that. We can maybe have access to outline of a new publication [that] we would like to produce. And we have also looked into some mechanisms to do some automated reading based on information that’s found out there. However, I would like to say that at this point in time at least we still need the human being, that’s for sure, because we have to review what ChatGPT produces. It’s maybe accelerating the work we have to do, so makes it easier, makes it faster, but I would not say it replaces the processes we have and not say that it creates a lot of innovation so far…
A second challenge I see in the kind of black box behaviour. If I referred to our clients, which often approach us and say, why is this company, in your view, sustainable or not, or why did you come to this conclusion? A bad answer would be, ‘I don’t know, but ChatGPT told me so’. That would not be an acceptable solution. So we still need to be very clear in what we are looking for, what is important for us, what are our convictions, and how do we also balance social aspects versus environmental aspects and so on, because it will influence the outcome…
… since we have a lot of unstructured data or qualitative data globally available, maybe even new data sources such as satellite imagery, news flow, we can use ChatPGT or other AI engines to combine these information points and bring them forward in a structured way for further interpretation by human beings.”
In my view, a great deal of caution is warranted in relation to understanding the data chatGPT uses in the first place. Yesterday, I asked chatGPT about how current its data is – here is the response:
“My knowledge is based on information available up to September 2021. I do not have access to real-time data or events that have occurred after that date. Therefore, I may not be aware of recent developments, news, or changes that have occurred since then. It’s always a good idea to verify information and consult more up-to-date sources for the latest information, especially when dealing with rapidly changing topics or events.”
I’m not sure many ESG folks know that chatGPT processes only data that is two years old – past the expiration date of much corporate climate and DEI information. I’ve written about my own set of experiments and resulting concerns with the technology (the final article is here, which contains links to other blogs in the series). Wild’s views and even chatGPT support the need for caution, significant human review, skepticism and ultimate decision making.
If you aren’t already, subscribe to our complimentary ESG blog here: https://practicalesg.com/subscribe/ for daily updates delivered right to you.
