PwC has become the first reseller of ChatGPT Enterprise through a partnership with OpenAI. This allows businesses to purchase the gen AI service through PwC instead of directly from OpenAI. PwC aims to upsell its services to clients interested in using ChatGPT to optimize their workloads. The firm already uses custom GPTs for various tasks and sees AI as a way to drive increased impact and productivity.
According to research from the University of Chicago, GPT-4, a large language model developed by OpenAI, has demonstrated the ability to conduct financial statement analysis with accuracy surpassing that of professional analysts. Using chain-of-thought prompts to guide its reasoning, GPT-4 achieved a 60% accuracy in predicting future earnings, outperforming human analysts. The researchers suggest that large language models may play a central role in decision-making despite challenges in numerical analysis.
However, A study from Purdue University reveals that 52 percent of programming answers generated by ChatGPT are incorrect. The research analyzed 517 programming questions on Stack Overflow and showed that 77% of the answers were verbose. Despite the inaccuracies, users still preferred ChatGPT answers 35% of the time due to their comprehensive and well-articulated language style.
We care about PwC’s goals – upsell services. That’s the opportunity they have identified. I see it too – and note that on the one hand, you have improved earnings reports, and on the other, incorrect answers to programming questions. And yesterday I reported on Google recommending glue on pizza. If I had the simple answer, I’d build the business there. It’s not that simple.