- GPT-4 update will offer improved language model
- As well as fewer logical errors and fakes
Startup OpenAI has unveiled a new version of ChatGPT artificial intelligence technology. It’s called GPT-4 and will offer better search and processing results.
What can this chatbot do?
At the presentation, the management acknowledged that the technology had some problems – for example, the bot often invents imaginary facts or refers to non-existent sources. The company promises that now such collisions should occur less frequently. But they won’t disappear completely.
The new version can also make simple logical errors and spread fakes. In addition, GPT-4 only knows the facts that happened before September 2021 – and does not learn from experience, OpenAI emphasized.
GPT-4 analyzes images well and describes them in words. But so far, OpenAI has not made this feature available to customers.
How to access
The new version of GPT-4 will be made available to paying OpenAI customers who sign up on the waiting list. However, some companies are already using this technology. For example, the language learning app Duolingo implemented GPT-4 for teaching dialogue. But to do this, companies pay for a different, more expensive subscription plan.
Microsoft has also confirmed the use of GPT-4. Their Bing search engine has been testing a new model for the past few weeks.
See also: “Love is gone? Elon Musk will now work on AI, not cryptocurrencies
