Challenges and Realities of Using Generative AI and Large Language Models

Generative artificial intelligence (AI) tools have been touted as a way to save time and boost productivity, but according to Peter Cappelli, a management professor at the University of Pennsylvania Wharton School, the backend work needed to build and sustain large language models (LLMs) may require more human labor than the effort saved up front. In a recent MIT event, Cappelli pointed out that while AI may create more work for people on a cumulative basis, it is important to consider the gritty details and realities on the ground when implementing these technologies.

Cappelli highlighted that projections from the tech side are often wrong, as seen with the delayed rollout of driverless trucks and cars that were predicted in 2018. He noted that while AI has the potential to transform various industries, the rollout is often hindered by factors such as regulations, insurance issues, and software complexities.

Furthermore, Cappelli raised concerns about the cost and validation of generative AI output. He mentioned that as more people utilize LLMs, the computer space and electricity demands will increase, leading to higher costs. Additionally, he emphasized the need for human validation of AI output, especially for complex reporting or critical undertakings.

One of the key challenges highlighted by Cappelli is the potential inundation of information and contradictory responses generated by AI. He cautioned that organizations may struggle to adjudicate the differences in output and ensure the reliability and accuracy of AI-generated reports.

Despite these challenges, Cappelli sees potential for generative AI in data analysis and decision support processes. He suggested that AI could be particularly useful in sifting through large data stores and delivering analyses to support decision-making. By leveraging AI for data analysis and database management, organizations could potentially improve their efficiency and effectiveness in handling large datasets.

In conclusion, while generative AI presents exciting possibilities for innovation, it is essential to consider the practical implications and challenges associated with its implementation. By addressing issues such as cost, validation, information overload, and human preferences in decision-making, organizations can harness the power of AI to enhance their processes and drive better outcomes.