Genuine question from a casual:
Seems like high quality models are becoming increasingly common. Yes 5.6 and Fable are still the best but seems like Anthropic & Open AI's value proposition has always been "these models are hard as fuck to create and expensive as fuck to deploy, so just pay us for inference"
If more companies are now creating high quality models, and we assume a Moore's law esque type of improvement in regards to the compute needed to run these models.
Is it not entirely possible that OpenAI and Anthropic completely lose their competitive advantage in the next few years?
I'm sure they will both continue to lead with the best models, but how long before the public says "5.6 level is good enough for my use case" and then is eventually able to run a model like that locally?
This is the first time since ChatGPT launched that I've really felt like it's a real possibly their dominance in the coming years could degrade rapidly.
@Maskoff023 Chat does OSI model actually matter? I've never been looking at logs and been like what layer is this shit in? I'm actually asking genuinely though