Can a machine replace the position of an editor? In part, it does seem to be possible thanks to GPT-3 technology, which is nothing more than an artificial intelligence (AI) model whose main utility is the generation of texts from a seed text. “Although a priori it can be a threat to humans, GPT-3 is usually used more as an inspiration tool to generate a base and for the editor or content creator to improve it”, comments José Gabriel García Ortega, CEO of La Caja Company. Although many companies are reluctant to incorporate it, GPT-3 technology is one of the first systems to show the possibilities of general artificial intelligence and, therefore, it could become a great discovery of the 21st century. Thus, García Ortega has developed five keys on the GPT-3 and analyzes how it is revolutionizing the content generation ecosystem. In my essay, I want to highlight them.

Firstly, like all AI, GPT-3 has had a pre-launch training process in which a total of 500 billion words were extracted and processed from all public books ever written and available, the entire Wikipedia, and millions of web pages and scientific documents available on the Internet. Once passed, it was released in June 2020 by the artificial intelligence organization OpenAI.

Secondly, GPT-3 is truly a great revolution in the sector. In 2019, OpenAI released GPT-2, an earlier version that worked with 1.5 billion parameters. The revolution came from the hand of GPT-3 since it works with 175,000 million parameters, a very considerable increase in processing capacity, and for which it is increasingly demanded by companies. However, there are bigger tools like Google's GShard, which has 600 billion parameters. “We will see more and more companies betting on this way of generating texts since, in addition to saving time, it is a solution that provides knowledge and training on any subject”, says García Ortega.

Moreover, GPT-3 is still an API that applied to a product gives you a solution. In this specific case, from a text request, the API returns a finished text that must then be perfected or rectified. Thanks to the number of parameters it has, it is capable of convincingly writing by connecting letters, words, and phrases. “Its composition algorithms are trained to be able to complete a text easily, in the most coherent way possible, from an initial sentence, adapting it to various jargons, such as that of a lawyer or doctor”, adds Ortega García.

Its main utility is the generation of a written language through texts for articles, social networks, etc. The question is whether GPT-3 is capable of writing like a human without the reader being able to distinguish it. Although the majority opinion is yes, it could only be so for short texts such as social media copy. However, for longer texts, it is usually used as an inspiration that generates a base so that later a human person can touch it up.

The programming ability is the latest utility that is revolutionizing the sector. Programming does not stop being a generation of texts from a seed, facilitating the planning of the company at any level. The big difference with other AIs is that they needed training focused on a single task, while GPT-3 is a generic model capable of adapting and imitating various tasks.