GPT-3, the algorithm that writes like a human arrives in Italy

Gpt-3 is a writing algorithm that manages to produce texts of the same quality as a human being. How do you tell? How can this be established? He called it the Turing test, no less.

gpt-3 it is the latest algorithm developed by OpenAI which creates texts of the same level as those of a human being. I say this through gritted teeth, because this big virtual machine could undermine my own job position! But let's get back to the news. This incredible result is made possible by a new language model, completely re-evaluated to allow clear, fluid, error-free writing on any topic.

Let's make a quick comparison with gpt-2, the older version of the same algorithm. Gpt-2 knew approximately 1,5 billion parameters. Well, it seems an absurd figure, but Gpt-3, which has just arrived in Italy, has already stored 175 billion.

What does it mean? It means that these inputs, or specific information, can be used and inserted into the context through a dynamic learning path which does not need human supervision. The value of a similar algorithm? Millions of dollars, of course. This splendid little jewel has just obtained the green light from Microsoft and can be integrated on the platform of the startup Indigo.Ai.

As reported by Gianluca Maruzzella, CEO and co-founder of Indigo.Ai, the potential of this technology is immense. “We are among the first companies in Europe to experiment with the most developed technological language model in the world today. The artificial intelligence created by OpenAI integrates with the features of the no-code platform created by Indigo.ai – cutting the time to create a virtual assistant in just a few minutes.

What are the concrete implications of Gpt-3?

In the current state of affairs still none, but it is expected that the country could greatly benefit from it, since it is such an advanced system. It is even thought that an international use could lead to an economic scheme based on Gpt-3. Also from the interview: "I imagine that in the future, perhaps in the very near future, a micro universe of tools will be born that will base their business on this language model, for the simple reason that it allows not having to spend time training the artificial intelligence . Indeed, it will be able to self-learn the tools it needs”. A new step in the construction of the neural network, basically.

The capacity of Gpt-3 is not given only by the amount of information, but also by its extraordinary ability to relate them to derive reliable data and facts. Not a small thing, given the amount of data stored. Wikipedia in Italy currently occupies about 4 GB. Gpt-3 takes 164. So imagine a scenario in which this ability is put at the service of chatbots, effectively allowing them to build an algorithm intelligent enough to provide a useful and unsubstantiated answer. The customer REALLY finds the solution to his problem and is no longer irritated by the idea of ​​talking to a somewhat limited robot.

GPT-3 is part of the language models popularized with the BERT, Or the Bidirectional Encoder Representations from Transformers of Google, and is no longer limited to predicting the possible thought of the user, going to find autocomplete solutions. In short, take Gmail which helps you write emails based on your previous ones.

GPT-3 is in able to construct meaningful sentences and have discussions without you having the horrible suspicion that you're talking to an AI. What is the ethical limit in using a GPT-3? There is a broad spectrum behind this amazing technology. His ability to learn so well and so quickly is almost a double-edged sword. Its ability to relate information, right or wrong, makes it an excellent propaganda tool.

In short, this tool must be used within certain ethical and non-speculative parameters, in order to serve humanity and not to be exploited as an element of manipulation. Gpt-3 will have to be used in the defense of the freedom of others and in support of man. Social norms and public policies, as well as educational ones, will have to be established to build a language model capable of helping, and not destroying, the work that has been done so far to create a sustainable internet model. And there is much to be said about this. In short, we will see what happens to Gpt-3, in the hope that it will be useful to us, and not backfire.