Have you ever tried chatting with a robot? This is what happens on many automated chats and the result is far from convincing. On the other hand, with ChatGPT, the experience is breathtaking. This chatbotchatbot developed by OpenAI has been talked about a lot for a week. This chatbot uses its artificial intelligenceartificial intelligence to appear as credible as a human interlocutor. And it is amazing! All you have to do is ask a question for the chatbot to answer it clearly and precisely instantly. We can thus “play” with the AI and ask it to write the texts on the most incongruous questions. With its faculties, the AI ChatGPT can imagine the scenario of a school writing and write it. The chatbot is also capable of creating a credible corporate communication. To use the tool, it is necessary to create a free account with OpenAI and log in. The rest of the operation is childish. In a field located at the bottom of the screen, all you have to do is enter your request. The chatbot will then write the lines as it goes, as if it were writing them in real time.
To achieve this result, OpenAI made its learning algorithm ingest astronomical amounts of data, whether it is literature, television, public online forums, press articles or film scripts. . On the other hand, the AI does not look for information outside its database. Chat is based on OpenAI’s GPT-3.5 language model. Its goal is to gain a deeper understanding of language. And it must be said that he imitates it very well and it is rather disturbing.
The limits of AI hide behind a nice speech
On the other hand, after a few tests on specific subjects, we can realize the limits of this AI. As soon as it is a question of being discerning and nuanced in front of a subject, ChatGPT states answers which are held at the level of the language, but which lose in direction and which are often false. This even happens on fairly simple general knowledge subjects. The eloquence of the AI then poses a problem because its assertions remain convincing. This phenomenon is a reminder of what happened to Meta’s Blenderbot chatbot when it made conspiratorial and anti-Semitic remarks.
A youthful mistake that could well continue, because if the AI is only in its first experimental version, these difficulties are difficult to correct by OpenAI’s own admission. But, the chatbot is not specifically designed for this. Moreover, when asked to describe itself, the AI explains that it is mainly intended for companies to improve their relationships with customers. via their Chatbot.
ChatGPT is not OpenAI’s first spectacular creation. The firm, which is run and financed jointly by American businessman Sam Altman and Elon MuskElon Muskis also the origin of DALL-E, a generatorgenerator images by keywords. Here again, it is a question of entering a sentence to create in a few seconds a work of art whose rendering is often impressive. In any case, we must never forget that these are machines.