OpenAI CEO Sam Altman gave a keynote on November 6, 2023 that highlighted the current state of ChatGPT and showcased a series of product updates, including ChatGPT-4 Turbo. Altman explained that increased context length, more control, better world knowledge, new modalities, customization, and higher rate limits will be part of the new offerings. In addition to these enhancements, he noted that the pricing will come down, making Turbo 2.75x cheaper to work on, and that speed will begin to increase. Will these enhancements be a catalyst for the development of GPTs?
Figure 1: OpenAI’s new announcements
Current stats: OpenAI
- 2 million developers building on the API
- 92% of Fortune 500 companies are building on OpenAI products
- 100 million weekly active users on ChatGPT
What’s new: ChatGPT Turbo and 6 major new enhancements
- Context length will now support 128,000 tokens of context
- More control over responses and output
- Better world knowledge by bringing knowledge into the model — now up to April 2023
- New modalities will go into the API today — vision and voice capabilities launched
- Customization — fine-tuning program and custom models in the API
- Higher rate limits will double the tokens per minute for all ChatGPT4 customers
Microsoft and OpenAI partnership
There was a cameo appearance at OpenAI DevDay by CEO of Microsoft (MSFT), Satya Nadella, during which he discussed the Microsoft partnership with OpenAI. According to Nadella, Microsoft is building in partnership with OpenAI and that this is causing the shape of Azure to shift with this alignment. Nadella also highlighted that Microsoft aimed to build the best system so that OpenAI could build the best models. Will this effort start to drive upside in revenues for Azure, Infrastructure-as-a-Service (IaaS) and Platform-as-a-Service (PaaS) businesses within Intelligent Cloud?
Figure 2: Microsoft CEO Satya Nadella and OpenAI CEO Sam Altman at OpenAI DevDay
GPTs: Gateway to the agent
GPTs are a way to create a tailored version of ChatGPT and are precursors to agents. These will integrate instructions and its expanded knowledge and then ultimately deliver tailored actions. Users will be able to build a GPT with natural language instead of having to code. These can be public, private, or specific to an organization. OpenAI plans to offer revenue sharing for GPTs and give developers $500 of OpenAI credits to incentivize the creation of new applications. In addition, these agent concepts will all come to the API.