How will AI change the research and investment worlds? Our view is that large language models (LLMs) will soon surpass human analysts in assimilating and summarizing information. Hence this video explores three areas where human analysts can continue to earn their keep, and possibly even help decision-makers beat the ‘consensus engines’.
We have spent much of 2024 writing about the rise of AI, and how it will change the energy industry: unlocking new step-changes in industrial efficiency, next-gen DAC or autonomous vehicles; while re-exciting gas generation, compounding grid bottlenecks, wolfing up grids’ spare capacity, boosting fiber-optics, industrial cooling, transformers and harmonic filters.
But how will these AI models change the research and investment worlds? This video sees decreasing value in research that assimilates and summarizes information already floating in the public domain. Machines can increasingly do that. But where will the machines struggle?
By definition, Large Language Models are ‘consensus engines’. These AIs are trained by throwing billions of tokens at an algorithm, which must learn to guess the most likely tokens to follow on from previous tokens. Or in other words, they will average out all of the wonky views on the internet, thereby arriving at a stale consensus!!
AIs are also not human. Humans may retain an edge in understanding what is on the minds of decision-makers, undertaking novel analysis to address these debates, pitching the conclusions in ways that will engage human readers, and relating conclusions back to their actual human experiences. Being a good analyst is ultimately about empathy.
Man versus machine? We are considering a new series of videos, where we will identify a topic that is particularly on the minds of our clients, then pit Rob against ChatGPT. Rob will present his best answer to the question, based on the TSE research library then ask ChatGPT to critique his answer. Then Rob will critique ChatGPT’s answer.
Since AIs are consensus machines, this exercise may help to draw out counter-consensus ideas. So please do write in if you would like to suggest any topics for Rob to debate with ChatGPT in the first instalment of ‘Man versus Machine’.