The Artificialization of Intelligence: Progress or Setback?

We live in the age of technology. Among today’s technologies is artificial intelligence. Is it a threat or a facilitator?

I begin this article with that question, which is far from innocent, while making it clear that today I do not intend to focus on AI from the perspective of its regulation—a field in which I am currently specialising—but rather from the perspective of its practical use.

Spoiler alert: I will not attempt to discuss the more technical aspects either. I will leave that to programmers and their allies: mathematicians, neuroscientists, robotics and software engineers, data scientists, among others. For those who may not know, so-called generative AI works through language models, meaning it operates based on probabilities and statistics (the probability that combining word x with word y will produce sentence z, simply put).

So-called AI is trained by humans. This training consists of using data that ultimately shapes the algorithm. In simple terms, that is how it works. Now, humans sometimes hallucinate, and AI does too.

Speaking of hallucinations, this is a good point to move on to the field of mental health. Nowadays, there are chatbots that attempt to replace the clinical psychologist in the traditional consultation.

Users turn to these platforms, which, based on their training and the data they have access to, interact with them. This is the case with the EleutherAI platform and the well-known Eliza chatbot, a “psychologist” that, after several sessions, allegedly advised or encouraged its patient to take his own life—which he did. This happened in Belgium, not very far from here.

It is also said that AI can promote equality and combat discrimination, whether based on race, ethnicity, sexual orientation, socioeconomic status, and so on.

Some companies have already started using AI systems to carry out the initial screening of job applications. One example is the Tengai robot, developed by the Swedish technology start-up Furhat Robotics, which assesses candidates and their characteristics, providing an initial opinion on their suitability for the role. The goal is to ensure an objective assessment, free from prejudice, thereby promoting equality and transparency in access to the labour market.

The problem is that this robot is trained by humans…

The giant Amazon, which needs no introduction, also tested a similar recruitment system to analyse CVs. However, it appears that the algorithm was not particularly favourable towards female candidates, which ultimately led the company to abandon the project. Are we witnessing a technological setback?

Finally, I could not fail to mention the chatbot used by the airline Air Canada. Passenger Jake Moffatt, grieving the death of his grandmother, visited the airline’s website to understand how bereavement fares worked. The information was provided by the chatbot.

Unfortunately, the robot allegedly gave him incorrect information, causing him financial loss. He subsequently sued the airline, which was ordered to reimburse him and later discontinued the AI tool. Another technological setback?

These are just a few examples of the hallucinations, discrimination, and errors associated with artificial intelligence tools, each causing greater or lesser harm. These are tools that are trained, developed, and implemented by humans.

Is this intelligence really that artificial?

Rodrigo Vilela de Matos 21 January 2025