ChatGPT has played a crucial role in reshaping multiple industries by bringing AI to the masses. OpenAI, the company behind ChatGPT held its first developer event earlier this week. The company has launched multiple improved models and new APIs to hold a lead over the competition in the AI race. The upside is that news and media publishers now have access to better AI tools to improve their efficiency and quality of work. OpenAI has also added the ability for users to create their own GPTs as well, something that I would encourage news publishers to build for their internal teams to improve workflows. That said, let’s dive deep into what OpenAI showcased at the event.
OpenAI unveiled an improved version of its already popular GPT-4 model. This new model comes in two versions, one used purely for text analysis and the other can analyze text as well as images. OpenAI has also increased the context window for GPT-4 and it can now process much longer prompts (equivalent to 128K tokens). To increase adoption, the prices for input as well as output token has been reduced. Input tokens are three times cheaper on GPT-4 Turbo compared to GPT4 while output tokens are 2 times cheaper.
OpenAI’s DevDay also witnessed the launch of Assistants API. These can be used by developers to create their own assistants. For example, these can be used to retrieve knowledge from an external source, call a particular function to perform an action, and more. I believe this can be used by news publishers to bring automation in the newsrooms by automating certain tasks using the Assistants API.
Users of ChatGPT Plus had access to Dall-3 image creation and the Text to Speech capabilities with the latter being release a few months ago. These features can now be accessed via API which eliminates the need to use the ChatGPT Plus interface. News publishers looking to explore AI-based image creation for their articles, or those who wish to add a text-to-speech module on their articles can do it using this API.
OpenAI has announced that users will be able to create their own GPTs for different use cases. The creation of GPTs does not require coding knowledge and can be done purely by prompts. OpenAI also revealed that it will allow enterprise customers to make an internal GPT which is built on the company’s knowledge resources. This eliminates the risk of leaks and helps companies leverage AI effectively. For ChatGPT Plus and Enterprise users the company has revealed sample GPTs from Canva and Zapier AI Actions to try out. News publications can create such internal GPTs to train the AI on their writing style and way of presenting information to help journalists speed up their publishing.
There's an app for that. It’s a common phrase popularized with the growth of third party app stores for mobile phones. OpenAI will launch a GPT store where users can publish their created GPTs for others to use. The store will have GPTs from verified builders to start with. OpenAI is also exploring options to pay GPT creators based on how many people are using theri created GPT.
OpenAI has announced that it will protect its customers and pay the legal costs involved, if they face a legal claim for copyright infringement. This plan dubbed Copyright Shield is “generally available” for features of ChatGPT Enterprise and OpenAI’s developer platform.