Openai is already working on GPT-6, the next version of Chatgpt. According to the start-up, the IA model will arrive faster than you think, and it will be very different from its predecessor.
Beginning of the month, Openai lifted the veil on GPT-5, a new version of the IA model that animates Chatgpt. The deployment of GPT-5 has proven more laborious than expected. The start-up was forced to make several concessions to appease the anger of its subscribers, very attached to GPT-4O, the previous iteration of the AI. OpenAi in particular softened the way in which the AI responds to requests, and kept access to old models.
Read also: Openai plans to launch an essential security function on Chatgpt
GPT-6 will arrive earlier than expected
Shortly after the launch of GPT-5, Sam Altman, CEO of Openai, spoke of the arrival of the next version of Chatgpt, namely GPT-6. Faced with a lounge of journalists, he assured that GPT-6 was already under development and that he will not be expected too long, reports CNBC.
In fact, Altman says that Optai will not wait as long as between GPT-4 and GPT-5. The output of the two models was more than two years oldGPT-4 having been unveiled in March 2023, a few months after the start of the Chatgpt wave. During this period of time, the firm nevertheless occupied the ground by released a multitude of intermediate models. Clearly, it will not be necessary to wait until 2027 to discover GPT-6.
Better memory for chatgpt
The boss of Openai reveals that GPT-6 will be very different from his predecessor. The next version of Chatgpt will especially adapt more to its interlocutors. Openai wants internet users to be able to Easily customize the chatbot depending on their personal tastes. With this in mind, the AI giant has already inaugurated a variety of features, starting with different personalities who influence the way Chatgpt answers questions.
For Altman, the Chatgpt customization key is memory. To offer a really tailor -made experience, the conversational robot must keep in memory who you are, with your preferences, your habits and your somewhat bizarre mania, estimates the manager.
“People want AI to have memory. They expect features that require that we are able to really understand them ”says Sam Altman.
Openai has already considerably improved chatgpt memory over the versions. Chatgpt is able to reference all your past conversations in order to calibrate his answers. For example, Chatgpt can remember that a user is allergic to a fruit or prefers a certain response format, thus avoiding having to recall it with each new conversation. For our part, we asked the chatbot to favor the answers in the form of a chip list.
Finally, Altman has promised that future versions of chatgpt will be ideologically neutral. In accordance with the requirements of the United States government, OpenAi will leave each user calibrating the ideological positioning of the chatbot. Clearly, Chatgpt can be “super woke” or rather conservative.
🔴 To not miss any 01net news, follow us on Google News and Whatsapp.
Source :
CNBC