OpenAI is optimistic about the future affordability of frontier AI models, including its own GPT family. Olivier Godement, the head of API Product at OpenAI, discussed this topic at VB Transform 2024. He mentioned that the cost of a token/word on the GPT-4 model has already been reduced by 85-90% since its introduction 15 months ago. Godement believes that this trend will continue, making AI models even more accessible in the future.
When OpenAI launched the GPT-4o model, it received a massive response from users. The model was not only half the price of its predecessor, GPT-4 Turbo, but it was also twice as fast. The demand for GPT-4o was driven by OpenAI’s efforts to improve both its capabilities and affordability. Users were impressed by the multimodal capabilities of GPT-4o, which made it feel like a truly human-like interaction with voice. Additionally, the company made significant efforts to make the models cheaper, enabling new use cases that were previously considered expensive.
OpenAI is committed to optimizing costs at both the hardware and inference levels to further reduce the cost of running frontier AI models. Godement draws a parallel with the smartphone and television industry, where costs have significantly declined over time. OpenAI’s goal is not to maximize revenue margins but to enable people to build and experiment with more use cases. Any cost optimization achieved by OpenAI will be passed on to customers, suggesting that there is still room for improvement in both intelligence and cost.
Godement also highlighted the success of ChatGPT Enterprise, which has gained popularity across various sectors such as consulting, finance, marketing, and sales. The platform has been adopted by teams to enhance knowledge work, with Moderna employees using it to create a GPT model for calculating vaccine dosages during clinical trials. Godement envisions a future where every employee in enterprises will have a super assistant to boost productivity and job satisfaction.
Regarding competition and rival models, Godement expressed a positive outlook. He sees advancements and innovations in the field as a reason to be happy. At the same time, OpenAI focuses on building strong customer relationships and trust to prevent customers from switching to competitors. While slight changes in metrics from vendors may not lead to significant customer swapping, maintaining strong relationships remains a priority.
Lastly, Godement discussed OpenAI’s recent partnership with Apple. The partnership serves two purposes: it helps OpenAI leverage Apple’s hardware capabilities, and it allows Apple to enhance its Siri virtual assistant using OpenAI’s technology. This collaboration aligns with OpenAI’s mission to make AI more accessible and useful for everyone.
In conclusion, OpenAI is dedicated to making AI models more affordable in the future. The company has already made significant progress in reducing the cost of running AI models, and it expects this trend to continue. OpenAI’s focus on affordability and innovation has led to the success of models like GPT-4o and ChatGPT Enterprise. By building strong customer relationships and collaborating with industry leaders like Apple, OpenAI aims to create a future where AI assistants are commonplace in enterprise settings, boosting productivity and satisfaction in the workplace.