During his presentation, he argued that the current debate on generative artificial intelligence is marked by extreme viewpoints that make it difficult to accurately assess its scope and implications.
Mangarelli explained that generative artificial intelligence cannot be analyzed from a single perspective, as it simultaneously impacts various areas. Among these, he highlighted technological, economic, labor, organizational, educational, social, and geopolitical dimensions, which makes assessing it more complex.
This cross-cutting nature reinforces the idea that it is a general-purpose technology, with diverse applications depending on the context and objectives of each organization.
Between hype and underestimation
One of the central themes of the conference was the difficulty of striking a balance between unbridled enthusiasm and skepticism.
Mangarelli noted that the current ecosystem is characterized by conflicting messages: ranging from views that predict immediate disruptive impacts to reports that downplay their effects or even point to failures in their implementation.
This tension between “reality and noise,” as he put it, creates an environment in which making informed decisions becomes particularly challenging.
How they work and why they matter
To grasp the scope of generative AI, the dean emphasized the importance of understanding its fundamentals. He explained that these systems are based on language models trained on large volumes of data, capable of generating content based on instructions, and that they are part of an evolution spanning more than two decades in the development of artificial intelligence systems.
In this context, he emphasized that the emergence of more advanced models is giving rise to systems that not only respond but can also plan, verify, and execute complex tasks.
From chatbots to autonomous agents
One of the most significant changes highlighted at the conference is the shift from conversational interfaces to more autonomous systems.
Mangarelli explained that current trends point toward agents capable of understanding objectives, breaking down problems, executing multiple steps, and adapting based on the results.
This shift marks a new paradigm: moving from tools that answer questions to systems that operate to achieve defined objectives.
Productivity with mixed results
Regarding the impact on productivity, Mangarelli presented evidence showing mixed results. On the one hand, some studies point to significant improvements, including a doubling of productivity in certain contexts. On the other hand, other research shows declines or limited effects, particularly among more experienced employees.
This reinforces the idea that the benefits of generative AI are not automatic, but rather depend on the context, the skills involved, and how it is integrated into work processes.
Beyond the code
Mangarelli also addressed the impact on software development, noting that building systems goes far beyond writing code. Aspects such as architecture, security, integration with existing systems, user experience, and governance remain fundamental, even in a context of increasing automation.
AI does not replace technical expertise; rather, it requires a broader and deeper understanding of systems.
A New Form of Literacy
Another key concept was the need to develop new capabilities, both at the individual and organizational levels. Mangarelli argued that the rapid pace of innovation in artificial intelligence is creating a gap between available technological capabilities and the actual capacity for adoption.
Closing that gap involves not only adopting tools, but also developing judgment, critical thinking, and an understanding of how these systems work.
Understand to Decide
In closing, the dean emphasized that the real challenge lies not in adopting or rejecting artificial intelligence, but in understanding it.
We can only fully grasp the true impact of generative AI systems once we have a solid understanding of how they work, their capabilities, and their limitations.
In a rapidly changing environment, striking that balance between underestimating and overestimating becomes key to making informed decisions and harnessing the potential of technology.