The costs associated with building an AI company, particularly for developing models like ChatGPT, are substantial due to the expensive training process and the necessary computing resources for inference.
The release of ChatGPT in 2022 ignited significant investment into the AI industry, driven by the belief that training cutting-edge models would become increasingly costly, which indeed proved true.
Training costs have soared, with models like GPT-4 costing an estimated $78 million, while inference has become more affordable, allowing companies to deploy AI products at a lower price point.
While building a frontier AI model might remain beyond the reach of smaller firms, training models with competitive capabilities has become relatively economic, influencing product deployment strategies.
Collection
[
|
...
]