Actually it IS google's problem. They don't publish through traditional academic venues unless it suits them (much like OpenAI/Anthropic, often snubbing places

Actually it IS google's problem. They don't publish through traditional academic venues unless it suits them (much like OpenAI/Anthropic, often snubbing places like NeurIPS due to not wanting to MIT open source their code/models which peer reviewers demand) and them demanding so many GPUs chokes supply for the rest of the field - a field which they rely on the free labor of to make complimentary technologies to their models.