OpenAI has just released GPT-4, its latest AI model. GPT-4 is a large multimodal model which is able to accept image and text inputs and generate text outputs. This new AI model also exhibits human-level performance on various professional and academic benchmarks.
GPT-4’s capabilities boast a clear improvement over the previous model (GPT-3.5) with regards to creativity as well as reliability. One major improvement being the ability to accept both text and images as input, making it capable of generating text outputs based on inputs consisting of both text and images.
This new AI model has also been tested on simulated exams designed for humans – outperforming existing large language models. Apart from just performing well in the English language, GPT-4 also has the ability to understand and respond in Latvian, Welsh, and Swahili.
API users will have the privilege of customizing their users’ experience within bounds, allowing for significant personalization.
GPT-4 is no way near perfect and still contains similar limitations as earlier GPT models. For example, this AI model does not know about events after September 2021, which can result in reasoning errors as well as accepting false statements as true.
As we already know by now, GPT-4’s capabilities are pretty significant, however this now poses new risks, code containing bugs, inaccurate information and even generating harmful advice.
According to OpenAI, GPT-4 is 82% less likely to give inappropriate content than the previous version – following policies better with regards to sensitive topics.
Microsoft has confirmed that the new Bing search experience is currently powered by GPT-4. GPT-4 can also be accessed with a ChatGPT Plus subscription. OpenAI may also look to be factoring another subscription tier to allow for more GPT-4 usage in the near future.
To access the GPT-4 API, you must sign up for the waitlist.