Behind the mechanism

In recent years, artificial intelligence (AI) has become a relevant topic on Brazilian social media. According to research led by Getúlio Vargas in Brazil, entitled “Imaginaries about Artificial Intelligence in the Brazilian digital debate” (2024), there were 365,000 posts about AI on Facebook, X and Instagram between January 1 and May 31, 2024. Of these posts, 41% were classified as optimistic, 7% as pessimistic and 42% as neutral. I believe these percentages will change as understanding of this topic improves over time.

On certain networks, I noticed very optimistic and very pessimistic opinions, especially regarding generative AI (a subset of AI that uses models to create text, images, videos and other forms of data). For any extreme case, a very useful recommendation: "Watch out the hype!".

Examples of discussions that, in my opinion, could be considered very optimistic and very pessimistic: “How to create a profitable company using only ChatGPT” and “Will AI replace consultants?”

AI will never replace technical professionals or create a business plan for a profitable company. In most cases, if not all, the important information for the decision is available in private datasets (e.g. market analyses, articles, surveys, etc.). And what we need to keep in mind is that generative AI uses publicly available data to train its models.

A simple example to illustrate my point of view: Consider asking GPT: What are the benefits achieved by companies using AI to support the supply chain? Base your answer on this article: “Cannas, V. G., Ciano, M. P., Saltalamacchia, M., & Secchi, R. (2023). Artificial intelligence in supply chain and operations management: A multiple case study research. International Journal of Production Research, 62(9), 3333–3360”

The answer will be based only on the Abstract of the article. And it should be! The remaining information is private and access to it will require a subscription.

Based on this scenario, I wrote this short article aiming to improve the understanding of some aspects of the AI ​​framework. I will cover some important concepts: the training process, how machines “understand” words in a given context or how they “see” images. I will start with an overview of AI and then delve into one of the main mechanisms of generative AI: the Transformer, used in models such as GPT (Generative Pre-Trained Transformer).

Good reading!

Leandro Santos

Behind the mechanism.pdf

Download PDF • 1.21MB