OpenAI has just announced the highly anticipated general availability of GPT-4, its latest cutting-edge text-generating model. The powerful model is now accessible through OpenAI’s API, marking a significant milestone in the advancement of natural language processing technology.
Early Access for Existing Developers, Wider Availability Soon
Starting today, existing OpenAI API developers with a proven track record of successful payments will have the privilege of accessing GPT-4. OpenAI plans to gradually expand access to new developers by the end of this month, with availability limits set to increase in the future, depending on compute availability.
Unprecedented Demand for GPT-4
OpenAI has received an overwhelming number of requests for access to the GPT-4 API since its announcement in March. The company expressed its excitement over the growing range of innovative products that leverage GPT-4 and envisions a future where chat-based models can cater to any use case.
Enhanced Capabilities: Text, Code, and Image Inputs
GPT-4 surpasses its predecessor, GPT-3.5, by introducing several notable improvements. Unlike its predecessor, GPT-4 can generate text, including code, and accept both image and text inputs. OpenAI boasts that GPT-4 performs at a “human level” across various professional and academic benchmarks. The model was trained using publicly available data, including web pages, as well as licensed data obtained by OpenAI.
Image Understanding in Testing Phase
While GPT-4’s image-understanding capability is not yet available to all OpenAI customers, the company has initiated testing with a single partner, Be My Eyes. However, no specific timeline has been provided for expanding this capability to the wider customer base.
More from News
- UK Joins WTO’s Digital Trade Agreement For Global Ecommerce
- Will The Latest AI Action Plan Improve AI Use In The UK?
- How AI Is Assisting Cybercriminals
- Parents Say Social Media Is The Better Option For Kids’ Finance Advice
- Freedom Holding Corp. Advances Fintech Ecosystem With Strategic Tech Developments
- Studies Reveal Mental Health Effects Of ‘Doomscrolling’ On Socials
- Meta’s New AI Can Generate Images With Text And Your Posts
- Apple Maps Finally Launches on Web
Challenges and Limitations
Despite its impressive capabilities, GPT-4, like other generative AI models, is not flawless. It is prone to generating inaccurate information and reasoning errors, occasionally with unwarranted confidence. Additionally, GPT-4 does not possess the ability to learn from experience, leading to failures when faced with complex challenges like introducing security vulnerabilities into generated code.
Developer Fine-Tuning and Future Plans
OpenAI has announced plans to enable developers to fine-tune both GPT-4 and GPT-3.5 Turbo, another recent text-generating model, with their own data. This functionality, which has been available for several of OpenAI’s text-generating models, is expected to be introduced later this year.
Growing Competition in Generative AI
The unveiling of GPT-4 in March triggered intensified competition in the field of generative AI. Recently, Anthropic expanded the context window of its flagship text-generating AI model, Claude, from 9,000 tokens to a remarkable 100,000 tokens. Context window refers to the amount of text considered by the model before generating additional text, with smaller windows resulting in a higher chance of veering off-topic.
OpenAI’s DALL-E 2 and Whisper APIs Now Available
In addition to GPT-4, OpenAI has made its DALL-E 2 image-generating model and Whisper speech-to-text model APIs generally available. This move expands the scope of OpenAI’s API offerings and provides developers with access to a range of cutting-edge AI capabilities.
Deprecation of Old Models to Optimise Compute Capacity
OpenAI plans to deprecate older models accessible through its API to optimise its compute capacity. Starting from January 4, 2024, models such as GPT-3 and its derivatives will be replaced with new “base GPT-3” models, which are expected to deliver improved compute efficiency. Developers using the older models will be required to manually upgrade their integrations by January 4. Furthermore, those who wish to continue utilising fine-tuned old models beyond this date must fine-tune replacements based on the new base GPT-3 models.
Supporting Developers through the Transition
OpenAI is committed to assisting users in making a smooth transition to the new models. In the coming weeks, the company plans to reach out to developers who have recently utilised the older models and will provide more information once the new completion models are ready for early testing.
OpenAI’s release of GPT-4 and its future plans signify a major leap forward in the field of text generation and set the stage for a new era of AI-driven applications and innovation.