Dataframed

There are a few caveats to using generative AI tools, those caveats have led to a few tips that have quickly become second nature to those that use LLMs like ChatGPT. The main one being: have the domain knowledge to validate the output in order to avoid hallucinations. Hallucinations are one of the weak spots for LLMs due to the nature of the way they are built, as they are trained to correlate data in order to predict what might come next in an incomplete sequence. Does this mean that we’ll always have to be wary of the output of AI products, with the expectation that there is no intelligent decision-making going on under the hood? Far from it. Causal AI is bound by reason—rather than looking at correlation, these exciting systems are able to focus on the underlying causal mechanisms and relationships. As the AI field rapidly evolves, Causal AI is an area of research that is likely to have a huge impact on a huge number of industries and problems. Paul Hünermund is an Assistant Professor of Strategy and