© Unsplash/Thom Milkovic
At the end of 2023, an astonishing thing was brought to light by users of GPT-4, the OpenAI model, which is offered to paying ChatGPT users. It seems that artificial intelligence has become lazy. And if OpenAI has not given precise explanations for this phenomenon, some have put forward the hypothesis that by learning from humans, GPT-4 also learned to be less productive during the holiday season. 'year.
In any case, the problem was (more or less) recognized by OpenAI, which promised a solution in December.“[…] we've heard all your comments about GPT4 getting lazier! We haven't updated the model since November 11, and that's certainly not intentional. The behavior of the model can be unpredictable and we are seeking to correct it”, we could read in a publication on X, on December 8.
And a few days ago, OpenAI announced an update to its technology, which should fix this laziness problem. More precisely, this announces an update to GPT-4 Turbo, which is an improved version of GPT-4. “This pattern performs tasks such as code generation more thoroughly than the previous preview pattern and aims to reduce instances of “laziness.” where the model does not accomplish a task”, explains OpenAI. It also specifies that 70% of requests from app developers who exploit GPT-4 via OpenAI APIs are already processed by GPT-4 Turbo. Indeed, in addition to being more efficient, the latter costs less to operate.
A price drop
OpenAI has also just announced a price reduction for developers using the less advanced model: GPT-3.5 Turbo . As a reminder, this is the model that OpenAI offers to users of the free version of ChatGPT. And this is also used by many third-party apps.
Starting this week, integrating this technology into apps will cost less for developers, which will help developers to expand their activities. “Entry prices for the new model are reduced by 50% to $0.0005/1,000 tokens and exit prices are reduced by 25% to 0.0015 $/1 000 tokens”, specifies the developer of ChatGPT. Besides that, GPT-3.5 Turbo becomes more efficient, since OpenAI has improved its precision and corrected certain bugs.
Otherwise, it should be remembered that OpenAI is already working on its next model, GPT-5. It is not yet known when this will be available, but it is possible that OpenAI will make the announcement towards the end of the year. In the meantime, this has already given a small idea of the improvements that we could benefit from thanks to this new model.
- At the end of the year, users complained that GPT-4 had become lazy
- Some thought that by learning from humans, AI learned to work less during the end of year period
- In any case, OpenAI has just announced an update which should make its artificial intelligence less lazy
- He also announced a drop in prices for developers who use GPT-3.5 Turbo on their applications
- And at the same time, the ChatGPT developer is already working on the GPT-5 model
< blockquote>
📍 To not miss any news from Presse-citron, follow us on Google News and WhatsApp.
[ ]