ANDn an interview with the Lusa agency regarding the emergence — and popularity — of major language models, such as ChatGPT, and their possible impact on the field of Arts, the president of APPIA underlined that the question of conscience, which can be intersect with the discussion about the concept and intent of the Art, “it is a discussion that is to come when (there are) models that live in the real world”.
“Imagine a ChatGPT that is placed in a physical robot and that, as long as the robot exists, it learns successively. But I think it is a future discussion. We will certainly have this discussion, but I think that right now is not the right time”, said the also professor at the Faculty of Engineering of the University of Porto.
For Luís Paulo Reis, the definition of AI “parts of the general definition of intelligence”, that is: “Intelligence is the ability to solve new problems through a skillful use of knowledge”.
Luís Paulo Reis stated that, at the moment, AI is capable of “generating art, at various levels, writing books, creating images, making videos, creating sounds of the most varied types, from music to other types, in a way that for one On the other hand, it can be said that it is creative, but that in reality they use all the images, all the texts, all the sounds that were given to them as ‘input'”.
For this reason, the president of APPIA questioned: “Is this creative or not? Humans do the same. The way I can paint a picture has to do with all my perceptions throughout my existence. We can think that this type of Artificial Intelligence has perceptions different from ours, they don’t smell, they don’t feel, but that doesn’t mean that what this AI can create isn’t totally original”.
For his part, Lusófona University professor and researcher Phil Lopes explained to Lusa that these large language models use “a ‘corpus’ of data that can be thousands and thousands of images by different artists and, from that, constrains the search space and manages to generate images based on the thousands of images that have already (been seen) by other artists”. So it is not possible “to say that at this moment an AI can create its own style”.
“(The model) uses a lot of what other people have already done and then does that ‘mish-mash’, it’s a kind of parrot that spit out things it has already found on the Net and manages to assimilate these things and generates a text. However, the algorithm does not understand what is being generated, it is not aware, it is simply probabilistic in the sense of: ‘I’ve seen this many times, this sequence makes sense, I’m going to generate it this way'”, said the researcher.
The fundamental question for Phil Lopes is related to legislation, in the context of copyright and the misuse of material in the database of language models, and the social impact of the application of technology: “What are the security measures that exist people are not affected by this technology in ways they are not expecting”, remembering the cases of ‘deep fakes’ in pornography as an example or misinformation.
Also Read: Portuguese entrepreneur creates viral company from ChatGPT
Always be the first to know.
Seventh consecutive year Consumer Choice for Online Press.
Download our free App.
