OpenAI, which launched at the end of 2022 with great success ChatGPT, capable of generating all kinds of text on command, unveiled GPT-4, a new version of the generative artificial intelligence technology that operates the famous chatbot.
Launched GPT-4 to improve ChatGPT
OpenAI qualifies GPT-4 as “final milestone in its efforts to develop deep learning” and claims that he has demonstrated human-level performance on various professional and academic criteria. For example, he passes the exam to become a lawyer with a score as good as the top 10%. The previous version, GPT 3.5, was at the bottom 10% level.
The company has rebuilt its deep learning stack over the past two years and worked with Microsoft Azure to design a supercomputer for workloads. GPT 3.5 was the first test of the new system.
Although the new system is more accurate than its predecessor, it still makes factual errors. For example, in one of the examples provided by OpenAI, Elvis Presley is considered the son of an actor. GPT-4 also has no knowledge of events that will occur after September 2021 and cannot learn from its experience. However, OpenAI has improved many security properties.
The company reduced the system’s tendency to respond to requests for banned content by 82% compared to GPT-3.5. Additionally, the model responds 29% more often to sensitive requests, such as medical advice and self-harm, in accordance with company policies.
“We hope that GPT-4 will become a valuable tool to improve people’s lives by powering many applications”says OpenAI. “There is still a lot of work to do and we look forward to improving this model through the collective efforts of the community to build on, explore and contribute to this model”.
GPT-4 is available today for paid ChatGPT users. This will come a bit later for free users. Also, Microsoft reveals that its Bing chatbot was already using GPT-4 in secret.