top of page

Exploring the Salesian Approach to Artificial Intelligence - Part 3 - By Joel Chacon Ph.D.

The Evolution of ChatGPT

An example of this a popular tool known as ChatGPT[1], which is an incredible tool that provides an interactive dialogue interface for users to communicate in real-time, it is widely used in the academy and in the industry, however initially ChatGPT required to be feed with data generated by humans, this is better known as Reinforcement Learning from Human Feedback (RLHF) which means that experts “teaching” to the AI what it suppose to do. Note that ChatGPT works with an insane amount of text that highly depends on the language which is constantly changing, for instance the language English used nowadays has changed from Shakespeare’s days, Isn’t it?.

The Limitations of AI: Uncharted Territory and the Importance of Human Judgment

Since IA is trained with old data, What will happen with unprecedented events which IA has no information about? The short answer is that it will depend on how the application was designed, however there are jobs that just can not be wrong, imagine a machine based on IA that can diagnostic if either a patient has cancer or not, but since the patient has information with no precedent, then the IA diagnostics him/her free of cancer, but instead the patient does has cancer and requires a treatment with quimios, Would you prefer to be attended by this amazing IA machine?.

AI's Role in Enhancing Workforce Performance and Productivity

Not everything is bad with AI, a promising application of IA in the industry is to boost the performance of workers, for instance a study revealed a 14% more productivity of customer support agents[2], a similar study for programmers revealed increases from 10% to 67% which is a huge improvement[3]. Therefore, it is quite probable that next jobs will require using IA in their workstations.

Embracing the Future: Empowering Students with AI in Education

Certainly, IA has a direct effect on schools. A lot of students are using ChatGPT to solve a lot of homeworks, however there is a huge problem; ChatGPT doesn’t provide reliable information, and a lot of students believe that it does, however this is not a reason to ban IA from schools.

Personally I believe that instead, students must be taught how to use current IA applications, which implies to teach them the capability to decide if the result of IA for a given task is good enough or it must be hand improved.

Rethinking Education and AI in the Classroom

It is surprising how a lot of teachers reject IA for schools with arguments like “I learned to look definitions from the book, my students must do it too”, the main concern of a lot of teachers is that students might be dependent to use IA for solving basic tasks like addition or division by hand, but What if AI can provide different methods or even faster methods[4] to do it?

I agree that students must know how to do basic tasks like addition and division by themselves, but Is it really that teachers want students to learn the basics or teachers just don’t want to learn new tools?, Do you remember how to do division by hand?, I do remember, How about getting the Greatest Common Divisor by hand?, not at all.

Striking the Balance: Harnessing the Power of AI while Preserving Human Agency

At the end of the day, I think that IA must be included in our daily lives as a tool to boost our performance, however as users we must know how to do each task from scratch, thus we don’t create a dependency with each tool derived from IA, because most probably that in a future we will need to do more personalized tasks that IA can not do by itself. This principle might be a good indicator for the inclusion of IA in both academy and industry. Even more, this might be used in our previous example of diagnosing cancer, using IA to run fast promising diagnoses, then using these results with the health professional to decide if a patient has cancer, in other words, using IA to boost the performance of the productivity.

[1] [2] [3] [4]

bottom of page