HIGHLIGHTS
- OpenAI released a number of updates and new products.
- GPT-4 Turbo has a deadline of April 2023 and offers 128k context windows.
- Users of ChatGPT can design unique GPTs for certain use cases.
OpenAI revealed a number of new features and improvements for developers and ChatGPT users at DevDay, an exhibition of state-of-the-art advances in artificial intelligence. Among these was the capability to build personalized GPTs. Let’s examine the most recent GPT version first, though.
GPT-4 Turbo
The launch of GPT-4 Turbo, an enhanced version of an already powerful AI model that promises unheard-of efficiency and a deeper comprehension of current global events, was one of the most notable announcements. With its enhanced capabilities, which include an amazing 128k context window and awareness of global events through April 2023, GPT-4 Turbo can handle more than 300 pages of text in a single prompt.
Compared to its predecessor, GPT-4, this upgrade is twice as inexpensive for output tokens and three times cheaper for input tokens due to its cost decrease. All developers who pay can access the new model, which is presently in preview. Updates to OpenAI’s function calling feature enable more precise and intricate multi-function commands to be sent in a single message.
Another noteworthy improvement is the addition of JSON mode for developers, which guarantees proper JSON outputs and is especially helpful for jobs that require creating specialized forms.
Assistants API
In order to make the process of developing AI-powered applications easier, OpenAI has released the Assistants API, which includes features like enhanced function calling, code interpretation, and retrieval.
This new API is a step toward more sophisticated and adaptable AI solutions that can better manage challenging tasks.
Changes For The GPT-3.5 Turbo
Furthermore, OpenAI improved speed on a range of workloads and upgraded its GPT-3.5 Turbo model with a default 16K context window.
The revisions show a 38% improvement in format after completing activities like YAML, XML, and JSON generation.
Multimodality
GPT-4 Turbo has advanced to the point that it can already process images, a capability that companies such as BeMyEyes are already using to help people with visual problems. In addition, DALL·E 3 may now be integrated by developers into their image-generating applications, and OpenAI offers a strong moderation system to guard against abuse.
OpenAI hasn’t ignored interactions that are similar to those of humans; it has released a text-to-speech API that produces speech of human quality and offers a variety of voice presets and model variations for a range of applications.
Price Drops For Developers
In order to facilitate application scalability, price reductions for a number of services have also been announced, coupled with higher rate limitations.
The most notable differences are that GPT-4 Turbo output tokens are 2x less at $0.03 and input tokens are 3x less than GPT-4 at $0.01.
Copyright Shield
With the release of Copyright Shield, OpenAI’s dedication to copyright protection is further reinforced. Developers of APIs and enterprise users of ChatGPT will be protected by this.
Speech Recognition
Furthermore, Whisper large-v3 and the Consistency Decoder, which OpenAI is developing to enhance voice recognition, are expected to yield better performance and image quality.
GPT-4 Fine Tuning
Additionally, the corporation is providing experimental access to GPT-4 fine-tuning, with a focus on enhancing quality and safety. The unique Models program, which offers customized GPT-4 training with exclusive access to unique models, has been created for enterprises that require bespoke solutions.
Custom GPTs
Additionally, OpenAI has introduced GPTs, a tool that lets users tailor ChatGPT for particular uses and promotes community involvement in AI research. The planned GPT Store, which will display user-created GPTs, complements this feature.
GPT-4 All Tools
The GPT-4 Turbo-powered ChatGPT, which covers data until April 2023 and does away with the need to switch between models, has been made easier to use by the corporation.
Developing AI’s Future
The most recent developments from OpenAI, which were unveiled at DevDay, mark a significant step in AI technology by providing developers and organizations with more powerful, effective, and easily available tools.
