Google’s Gemini is competing with more than just ChatGPT
Google’s timing in announcing their new AI model, Gemini, may be strategic marketing against the backdrop of OpenAI’s recent drama with Sam Altman.
Not only can they leverage the results showing Gemini Ultra outperforms GPT-4, Open AI’s state-of-the-art language model, but they can ride the new wave of distrust in a company whose unusual structure has shown the world it can lead to uncertainty.
However, this may do Google more harm in the long term. Recent events at OpenAI have reminded enterprises of a risk that predates ChatGPT and recent AI. Building an enterprise system on the back of proprietary technologies, owned by technology giants, puts businesses at their mercy to control pricing, service level, and access.
In the spirit in which OpenAI was first founded, but has since dropped, companies may instead choose to consider open-source AI models. Traditionally, the most performant models were open source. Google’s pioneering language model BERT, released in 2018, allowed consumers to adapt the model freely for their use case and host it on any technology stack. This also instilled more confidence in methods of responsible AI, as the data used to train the model, as well as the model structure, was in the public domain.
The costs of training a new generation of language models changed this norm. Altman has previously stated that the cost of training GPT-4 was over $100 million. Pushing forward the frontier of AI required enormous investor backing, leading OpenAI to restructure as a for-profit company. And these proprietary models now come with a price tag for consumers, investors are not willing to part with a property that was so expensive to produce.
Companies now need to weigh the benefits of using the most performant AI models against the inherent risks that come along with them. There are plenty of open-source alternatives, including Meta’s LlaMa 2 model, that are still strong AI models, and cloud platforms are continuing to create tooling that makes deploying these models simpler and cheaper.
In the end, Google’s battle for AI supremacy may not be against OpenAI. Instead, they may need to create a narrative that reignites the dwindling public confidence in big tech’s AI supremacy.
