Artificial intelligence (AI), a skeptical perspective
These days, everyone is talking about Artificial Intelligence. While it has been a topic of work and debate for years, it is increasingly being used, and in recent months, it has been impossible to go unnoticed.
Chat GPT appeared less than a year ago and changed the way humans interact with technology forever. While artificial intelligence had already existed before, OpenAI’s project made it the center of attention not only for experts in the field. With Chat GPT, artificial intelligence began to be accessible to everyone.
This led many to talk about its virtues, how it could facilitate certain tasks, and the numerous solutions it could provide. As we observed in this debate, ethical dilemmas also emerged. In this same critical context, Ariel Torres, a linguist, professor, and specialist in the field, gave a lecture for the Chair of Information Technology at the National University of Rosario (UNR) titled “Impact of AI on Professional Work”.
Intelligence and Artificial Intelligence (AI)
The first question posed by the speaker was as follows: “What is intelligence, and what is artificial intelligence?” In this regard, he responded that there are many types of intelligence, mentioning, for example, emotional intelligence or cleverness itself. Not all of them necessarily involve a series of steps to solve a problem, as one might think in a general definition. “The fact that AI follows a certain logic does not mean it is intelligent in solving all problems,” he added.
He then highlighted a distinction within artificial intelligence itself: General AI (AGI) and Generative AI (GAI). The former is the goal that everyone wants to achieve. According to the Journalist Estefanía García, AGI can think, reason, perceive, infer, everything that humans can do, and could perform any intellectual task, like a human, but with fewer or no errors. GAI includes any AI that, as its name implies, generates new material from previously input data.
Taking all of this into account and as a preliminary conclusion based on these definitions, Ariel Torres stated: “What we define as artificial intelligence is quite limited.” Although there have been giant strides in the development of these technologies, it still seems that there is a need for further development and, above all, learning.
Limits and Issues
The specialist also discussed the problems generated by Chat GPT. He mentioned, for example, the potential modification of certain professions, possibly leading to unemployment. Additionally, he emphasized two other problems that happen just for Chat GPT to exist: the human cost and the environmental cost.
Ariel Torres cited a BBC article that talks about the ‘hidden workforce,’ which refers to people in poor countries subcontracted for low wages to ‘train’ AI systems. Their task “involves labeling millions of data and images to teach AI how to behave.” According to experts, this affects mental health not only due to the tedious nature of the task but also “because of the toxic material some labelers are exposed to.“
On the environmental front, the columnist for Argentina’s La Nación newspaper referred to the impact caused by this technology. In this regard, we can start by mentioning the high energy consumption, leading to carbon emissions. The previous chatbot model ‘generated emissions equivalent to over 550 tons of carbon,’ as reported in El Comercio, which also analyzed the technology’s water consumption: “For every conversation, AI “consumes” a water bottle.” “OpenAI needed 700,000 liters of water to train the previous ChatGPT language model,” they revealed.
In addition to all this, we can add concerns about algorithmic biases and the so-called “stochastic parrots,” which are essentially “an operational model capable of little more than the semi-random repetition of options in databases selected by the operator’s biases.” For this reason, Ariel Torres remarked about AI that “they work well as long as they are not in a completely unprecedented situation.“
As another point to mention, the columnist noted that these technologies cannot be empathetic because they simply have no interest in being so. They don’t care. Despite attempts, he also finds AI far from art or humor, which are very human characteristics, even geniuses. Ariel Torres maintains, nonetheless, that AI is very good at some activities. However, they cannot do everything, so “we need to be skeptical.” Therefore, he concluded that “there is room to grow and work on.” Let’s focus on that!