OpenAI has announced the general availability of its latest language model, GPT-4, through a cloud-based API. The company, known for its heavily-funded artificial intelligence endeavors, has also revealed plans to phase out several earlier neural networks while making three other AI models’ APIs generally accessible.
GPT-4, introduced in March as the successor to GPT-3, offers improved user instruction interpretation, the ability to process image inputs, and enhanced reasoning capabilities. These upgrades enable the model to tackle complex tasks such as solving math problems more efficiently.
The newly launched GPT-4 API allows developers to integrate the model into their applications. Initially, the service is accessible to existing customers who have purchased access to OpenAI’s APIs. However, OpenAI plans to open access to new developers later this month, with intentions to expand the API’s usage limits in the future. As part of its long-term roadmap, OpenAI also aims to introduce a fine-tuning capability, allowing developers to enhance GPT-4’s accuracy by training it on custom datasets.
Accompanying the launch of GPT-4, OpenAI is also making three other models’ APIs available. GPT-3.5 Turbo, a predecessor to GPT-4, offers similar capabilities at a lower cost. Additionally, OpenAI is opening access to the APIs of DALL-E and Whisper, which are optimized for image generation and speech transcription respectively.
In conjunction with these developments, OpenAI has announced the discontinuation of several earlier language models offered through its APIs. These models will be taken offline on January 4, 2024, and developers using them will need to switch to newer OpenAI products. The changes come following an update to OpenAI’s API technology earlier this year.
The Completions API, initially introduced in 2020, is being phased out in favor of the newer Chat Completions API. OpenAI believes that the new interface, which offers structured prompts, provides better results and mitigates prompt injection attacks. These attacks involve hackers modifying AI models’ outputs using malicious instructions. The enhanced security measures aim to assist developers in ensuring the safety of their OpenAI-powered machine learning applications.
OpenAI’s latest announcements not only highlight the strides the company is making in language modeling but also its commitment to improving performance, expanding accessibility, and prioritizing security in its AI offerings. With GPT-4 and its accompanying APIs now available to developers, the possibilities for AI-powered applications and innovations continue to expand.
