In exploring the varying beliefs about generative AI, we can categorize opinions into three distinct forms reminiscent of the three versions of the Efficient Market Hypothesis.
1. The Weak Believers: They view GenAI as a mere statistical parrot, reassembling existing ideas without genuine creativity. This group either closely scrutinizes AI development, focusing on its mechanics, or observes from a distance, emphasizing AI's errors and perceived limitations.
2. The Semi-Strong Believers: This majority group is attracted by the practical value of GenAI, using it for tasks like drafting emails, preparing sales pitches, or summarizing podcasts. They like using GenAI for querying and chatting. They mostly take GenAI for its observable, everyday applications and have little time to theorize on its mechanics or other actual or potential use cases in exotic fields. In some ways, they perceive transformer models as another type of new, helpful set of algorithms, like Google's page rank decades ago. Effectively, they acknowledge its imperfections but appreciate its utility without delving too much into whether it surpasses humans in creativity or other fields.
3. The Strong Believers: A small group, like this author, believes transformer model method to create models with a higher level of complexity than the best of us overall. They argue that generative AI can create novel solutions, addressing complex scientific challenges beyond human capabilities. This belief extends to its potential in solving intricate mathematical problems and advancing scientific discovery, possibly leading to a transformative technological leap, if not today, then in a short while, given its continuously improving capabilities. The solutions will not be perfect, but they give us starts and details that were not there pre-LLMs. On the GenInnov website, https://geninnov.ai/articles/, we list dozens of examples of how foundation models are proving their utility in turbocharging innovations across fields almost every week.
Unlike in the case of EMH, the three belief sets will not keep debating past each other without resolutions for long. The developments in the next few quarters will conclusively settle whether GenAI and subsequent technologies are brains that can tread where humans could not. The resolution will significantly influence not just financial markets, particularly the fate of today's tech cyclicals, but also the overall direction of technological progress.
Until then, we all must decide which school we belong to for many commercial, investment, and other community/family decisions. GenInnov website is perhaps an excellent place to start for a summary of external articles showing different facets of LLMs.