(W.E. Talk) Fausto Giunchiglia: How to Embrace the Ongoing Development of AI?(2)

CNS: What challenges will AI bring to our future life?

Fausto Giunchiglia: A first challenge deriving from the widespread adoption of GenAI systems is an important loss of white-collar jobs. Hopefully, this effect will be transitory as new jobs will be created around the use of GenAI systems in old and new problem domains. 

A second more subtle challenge relates to the safety issues which arise because of the low quality of the answers provided by GenAI systems. This problem arises because of the very nature of GenAI systems and of how they are trained. GenAI systems have no access to reality, as we perceive it, with our eyes, ears, smell, tact and taste. They have no idea of what reality is. They learn mainly from online text, as provided by third parties, most often humans, sometimes other AI systems.

Where does the material used to train AI systems come from? How to guarantee that the privacy of individuals is guaranteed? Issues like these raise a foundational medium-term question about the potential and not necessarily positive impact of GenAI on the future of humanity. It is easy to conceive a future where people will ask GenAI systems about everything, and most often about things they do not know about. Most often, people will have no means of checking whether what the GenAI says is true, or properly grounded in the current local context, as perceived by them. 

CNS: The launch of DeepSeek-R1 this January by Chinese AI company DeepSeek has attracted much global attention. For you, what's the special point of this model? 

Fausto Giunchiglia: The key important enhancement brought by DeepSeek is higher quality results with substantially lower energy consumption. This is definitely the way to go. But, in my opinion, this is only a first small step, a first set of results which can be amplified of orders of magnitude, leading to AI systems which are more affordable, more tuned to reality, and more useful to people.