OpenAI has received a new cash injection. And the company immediately announces new actions. It includes a new user interface. Canvas aims to make working with ChatGPT more intuitive – but it applies exclusively to text and code, not images, as the name suggests. There are also new tools for developers, a real-time API, fine-tuning vision, and an API for speedy caching.
Advertisement
A canvas is a separate window that you can open in ChatGPT – and it opens automatically when ChatGPT thinks it might be helpful. The function is used to improve collaboration with ChatGPT. AI should be a better help here too. In the future you can work on text and code. The only option so far is to ask something and ask again if you are not satisfied with the answer. The canvas allows you to mark areas and change them. ChatGPT can also provide feedback on individual pieces, considering the entire project upon request.
There are some shortcuts available, like suggesting changes and language corrections, making something longer or shorter, adjusting the difficulty level for reading comprehension, and inserting emojis. The latter certainly makes creating many AI-generated LinkedIn posts even easier. There are other shortcuts to coding too.
OpenAI’s canvas
(Image: OpenAI)
As always, this is an early beta version that you can test now and it is based on GPT-4o. Subsequent training took place to optimize the model for collaboration work. According to OpenAI’s blog post Without additional data created by humans. Data from OpenAI’s o1-preview was incorporated using distillation. People with ChatGPT Plus and Teams can access Canvas, with Edu and enterprise customers joining in the next few days. Canvas will be available for free once it comes out of the beta phase.
New API features from DevDays
At DevDays, a developer event in San Francisco, OpenAI also introduced new features for the API. A real-time API makes it possible to offer multimodal applications with low latency. It works with preset voices of Advanced Voice Mode.
Model distillation is a way of transferring knowledge from larger models to smaller models. It is now available for GPT-4o Mini. According to OpenAI, prompt caching provides developers with a 50 percent discount and faster processing times by reusing previous input tokens.
There is also an option to do finetuning for images for GPT-4o. For example, this may improve image understanding.
Just a few days ago, OpenAI received $6.6 billion from investors in the form of convertible bonds. This represents OpenAI’s new company valuation of $157 billion. However, investors are demanding changes in the structure of the company. As previously reported, the non-profit status will be eliminated and OpenAI will become a for-profit company. This is the only way investors can get more money back. However, to do this, OpenAI must first make more money. Insiders currently expect a loss of US$5 billion at the end of the year.
(EMW)
Source code editor simplifies Visual Studio Code 1.94 file searching
