Google is back competing head to head with OpenAI. This raises the question of how a large language model like this can be plugged into real-time internet data. In the past, models like GPT were not trained with up-to-the-minute data, but there is no fundamental obstacle to training a transformer model with real-time data. OpenAI initially released their model as a research project, but Google has the advantage of having invented transformers and having a strong AI team.
While OpenAI has made significant progress in training large language models, they have focused primarily on releasing them as products, even in the early stages of development. On the other hand, Google has been more cautious in releasing their models publicly, prioritizing stability and avoiding unreliable responses.
The competition between Google and OpenAI in the field of large language models is inevitable. However, it is not clear who will maintain the lead in the long run. The open-source community has the potential to create even more advanced models, and smaller tech companies are also investing in open-source models. The advantage of having a model with more parameters and being a big tech company is not necessarily sustainable.
OpenAI’s lead in the conversational AI market does not guarantee long-term success. Google and Microsoft are also strong competitors in this field, and they have the resources and expertise to catch up. The future of language models is uncertain, and the next breakthrough may come from a fundamentally altered or improved architecture.
In terms of business models, both OpenAI and Google face challenges. OpenAI’s revenue model relies on API calls and subscriptions, which can be replicated by other companies. Google’s search-based advertising model does not easily translate to conversational AI. The economic landscape of large language models is wide open, and it remains to be seen who will emerge as the dominant player in this new era.
