>>105990404
When locally ran models can good enough to compete with models ran on centralized servers. There's no point for most business to take on the energy cost of running LLM's, especially when you shift those burdens onto the consumer. The near future of """AI""" is your model running a local model, but need to pay a subscription to access features that'll require API integration or internet capabilities. I could see Gemini on phones needing a google one account to access much of the "advanced features", it's somewhat talking place already imo, pic related. Right now it's pro, but as the tech bro wallets gets tighter, Gemini on phones will get more stingy on features and access after normies get accustomed to using it.