Last Thursday, July 6, OpenAI announced the universal availability of GPT-4, its latest model of generative artificial intelligence, through its API.
All existing OpenAI API developers with a history of successful payments can access GPT-4. The company intends to expand the availability of machine intelligence configuration for developers of other categories by the end of July. Also, the firm’s plans provide for increasing the ability to use this model of generative artificial intelligence, depending on the load on computing power.
The company’s blog published information that since March of this year, millions of developers have requested access to GPT-4 API. The firm also reported an intensive increase in the number of innovative products using this model of generative artificial intelligence. Representatives of the company said that in the future they imagine AI configurations based on a chatbot and supporting any use cases.
The GPT-4 functionality provides for the possibility of creating text materials. This chatbot differs from its predecessor, which was GPT-3.5, in its ability to accept image input. GPT-4 was trained using publicly available information arrays and data licensed by OpenAI.
The image recognition function is currently not offered to all users of digital products by the developer of the most popular chatbot based on artificial intelligence. OpenAI is testing this feature with one partner, Be My Eyes. It is not yet known when image recognition will become publicly available.
GPT-4 impresses with efficiency and capabilities in terms of intellectual activity, but it is still not an ideal digital development. Sometimes chatbot invents facts in response to user requests and defines them as existing in objective reality. Also, GPT-4 may be mistaken in its reasoning on a certain topic, but confidently prove the reliability of the generated maxims. Another disadvantage of the chatbot is that it does not learn from its mistakes and does not draw the right conclusions from negative experiences when solving complex tasks.
OpenAI claims that developers will be able to fine-tune the GPT-4 and GPT-3.5 Turbo based on the data they have. The possibility of this activity will appear before the end of 2023.
GPT-4 was presented in March 2023. Since then, competition in the field of generative artificial intelligence has increased significantly. For example, Anthropic has expanded the context window for Claude, an AI model that creates text materials and is still in the preview stage.
OpenAI announced that from January 4, 2024, some of its old artificial intelligence configurations will not be available for use. This applies, in particular, to GPT-3 and its derivatives, which will be replaced by new GPT-3 base models. Experts say that it is highly likely that new versions of this product will have more efficient computing capabilities. Those who intend to continue using the old AI models, after January 4, will have to set up a replacement on top of the new basic configurations.
