9621 Agnes Crossing, Lake Suzanneview, New Mexico Island 84604-9295.
OpenAI has already held its first event for developers. He did this last Monday, and of course ChatGPT was at the center of all eyes. And as they
OpenAI has already held its first event for developers. He did this last Monday, and of course ChatGPT was at the center of all eyes. And as they explained, they did not disappoint, at least on paper Interesting developments in artificial intelligence chatbot coming soon.
After releasing GPT-4 a few months ago, OpenAI has now announced an improved version they call GPT-4 Turbo. Among other things, new turbo adds more up-to-date information It’s dated April this year, but there’s more news.
One of the biggest complaints raised about GPT-4 was the length of context it could cover (32K) and the fact that its information was still remaining at the end of 2021. Here are the two The main weak points that GPT-4 Turbo attacks.
On the one hand, as we said at the beginning, GPT-4 information at least until April of this year. It goes without saying how useful it is to contextualize the events that occurred in the year and a half that separated GPT-3.5 and GPT-4 from this new ‘Turbo’. Additionally, OpenAI promises to make every effort to ensure it “does not become outdated again.”
Regarding the length of the context say: GPT-4 Turbo Supports 128K context length, equivalent to 300 pages of a standard book. This refers to the number of words we can use in a prompt so that we can understand and interpret everything when presenting your answer, without losing the language model.
There is also good news on the subject price reductions for tokens As we can already see on the OpenAI website, it states. Currently, input tokens remain at $0.01 per 1,000 tokens, while output tokens remain at $0.03 per 1,000.
there is also Improvements for advanced users interacting with JSON elements In the GPT-4 Turbo API. This makes specific reference to what are known as function calls, allowing actions such as “open the car window and turn off the air conditioning” to be requested without requiring different input and output tokens.
Another announced innovation that we cannot ignore is related to DALL-E 3, an artificial intelligence model that can produce images and can currently be used in ChatGPT variants such as Bing Chat. From OpenAI they confirm: ChatGPT Plus and Enterprise users will be able to integrate it directly into their applications.via API.
By the end of this November, OpenAI also plans to: Opening a store where personalized chatbots will be distributed. They call them ‘GPT’, and the aim is to “empower researchers to build a large, personalized model designed specifically for them” using tools OpenAI has.
Not many details about this store have been disclosed, prices are much less, but They say it won’t be cheap. They even state that “many companies cannot do this at the beginning”, and this is actually understandable if we take into account the high business and operational costs this will bring.
More information | OpenAI
Cover image | Álvaro García M. via Bing Image Creator
Xataka on Android | Google Bard vs ChatGPT, the definitive duel on your mobile
Source: Xatak Android
John Wilkes is a seasoned journalist and author at Div Bracket. He specializes in covering trending news across a wide range of topics, from politics to entertainment and everything in between.
