A reflection on the evolution of thought and the current discussion on AI concerns.

“technology sufficiently advanced is indistinguishable from magic.” Arthur C. Clarke

ChatGPT is the center of so much controversy today and brings excitement and concern. Some people see a world of possibilities, while others see a world of dangers. I see both worlds happening simultaneously, but more importantly, I think this dilemma has a deeper meaning that reflects a potential issue in the way science and knowledge have been approached for a long time. Curiously, people are now concerned about the ethics, transparency, interpretability, morals, conscience and values of these new machines and the people using them. However, it seems paradoxical that these characteristics correspond to the metaphysics that was rejected when we adopted the “enlightenment” agenda a few centuries ago, in which Descartes was one of the essential proponents and some of his ideas behind the current mindset of today’s positivist scientific inquiry.

Continue reading “A reflection on the evolution of thought and the current discussion on AI concerns.”

Reproducibility in Keras

I bet I am not the only one that tried to figure this out! Of course I google’d around but really could not find a solution to make a Keras model expel the same results every time I execute the notebook cell with the compile and fit of the model. I read and applied the recipe in the official Keras documentation but still my code was returning different results. The solution came when I read the details in the Tensorflow documentation for tf.random.set_seed. It turns out that in order to make a Keras model compilation and fit methods reproducible we need to wrap our model in a function to leverage what the TensorFlow documentation says about functions:

Continue reading “Reproducibility in Keras”

Foundations: is low R-squared a problem?

R-squared nowadays is a strange metric that is not even mentioned in the ML community anymore because it is not used to evaluate modern ML models. However, understanding the intuition behind it and its proper interpretation is helpful to understand how noise in the data can impact model parameters. This post is from a white paper I wrote several years ago but I think it is still a good piece to reflect on how noise in the data can impact model parameters.

Continue reading “Foundations: is low R-squared a problem?”

ETSI GANA

ETSI provides a great framework for guiding the strategy of wireless operators towards the vision of autonomous networks through the GANA framework. The major contribution of this framework is the inclusion of the “Decision Engine” (DE) concept that enables the integration of AI components in different planes of the future wireless networks architectures. Check the whitepaper for far more details here https://www.etsi.org/images/files/etsiwhitepapers/etsi_wp16_gana_ed1_20161011.pdf