I’m back from GreenBiz24, which was both amazing and overwhelming. There were estimates of up to 4000 attendees, more vendor exhibits than you could shake a stick at and more sessions/panels than anyone could possibly attend. While valuable and worthwhile, the conference was also unfortunately a microcosm of one major problem with ESG in general: it was overwhelming, too complicated and impossible to grasp in its entirety. Perhaps the most ubiquitous topic was artificial intelligence (AI), although the voluntary carbon market (VCM) was a very close second.
While VCM is complex and multifaceted, its fundamentals are generally known/understood. The same can’t be said about AI – at least not in a crowd of ESG, climate and sustainability professionals. In one session, I stood up and asked the panel “what IS AI, really?” That wasn’t facetious or rhetorical – I truly had no idea. It turns out many attendees also lack basic knowledge on AI. Luckily, I had a long conversation with an AI expert who educated me on basics, including these three fascinating points:
- The vast majority of energy used by AI is during the training period. The energy needs for producing query results are minimal in comparison. At the moment, however, training needs seem to be essentially continuous as explained in the third point below.
- AI doesn’t think, it really just identifies patterns or trends in the data on which it was trained, and responds to queries based on those patterns/trends – although frequently in a complex or nuanced way. Results are therefore highly sensitive to the quality and validity of that original data and how queries are structured. He mentioned that incidents of AI “hallucinating” were attributable to those facts rather than a faulty AI algorithm.
- But this last point left me dumbfounded. After the learning phase is done, the data that was processed and parameters for learning are deleted. In other words, after the AI has become learned, there is no way to review, evaluate or validate data on which its learning was based, nor the original learning instructions. It loses its memory. This has huge implications for auditing/assurance. Further, AI systems therefore can’t be retrained with new or additional data – the learning phase must begin anew when increasing the data universe or changing original learning parameters.
Perhaps AI is less SkyNet and The Terminator, and more akin to W. W. Jacobs’ The Monkey’s Paw. There is so much for ESG and sustainability professionals to learn about AI. We will be announcing a few things in the near future to help. Stay tuned.
If you aren’t already subscribed to our complimentary ESG blog, sign up here: https://practicalesg.com/subscribe/ for daily updates delivered right to you.
