It’s very corporate, isn’t it? “Just keep scaling what we have.”

That being said, a lot of innovation is happening, but goes unused. It’s incredible how my promising papers come out, and get completely passed over by Big Tech AI, like nothing matters unless it’s developed in house.

The Chinese firms are picking up some research in bigger models, at least, but are kinda falling into local maxima too.

They're investing this much because they honestly seem to think they're on the cusp of super intelligent AGI. They're not, but they really seem to think they are, and that seems to justify these insane investments.

But all they're really doing is the same thing as before but even bigger. It's not going to work. It's only going to make things even more expensive.

I use Copilot and Claude at work, and while it's really impressive at what it can do, it's also really stupid and requires a lot of hand holding. It's not on the brink of AGI super intelligence. Not even close. Maybe we'll get there some day, but not before all these companies are bankrupt.

I knew it was a bubble since Computex January 2024 when Derb8uer showed an "AI PC case". He asked "What's AI about this PC case?" and they replied that you could put an AI PC inside it.

The problem is there is little continuous cash flow for on prem personal services. Look at Samsung's home automation, its nearly all online features and when the internet is out you are SOL.

To have your own Github Copilot in a device the size and power usage of a Raspberry Pi would be amazing. But then they won't get subscriptions.

what really is lacking is now more hardware dedicated to single user sessions to improve quality of output with the current models

That is the exact opposite of my opinion. They're throwing tons of computing at the current models. It has produced little improvement. The vast majority of investment is in compute hardware, rather than R&D. They need more R&D to improve the underlying models. More hardware isn't going to get the significant gains we need

more development on offline services

There is absolutely massive development on open weight models that can be used offline/privately. Minimax M2, most recent one, has comparable benchmark scores to the private US megatech models at 1/12th the cost, and at higher token throughput. Qwen, GLM, deepseek have comparable models to M2, and have smaller models more easily used on very modest hardware.

Closed megatech datacenter AI strategy is partnership with US government/military for oppressive control of humanity. Spending 12x more per token while empowering big tech/US empire to steal from and oppress you is not worth a small fraction in benchmark/quality improvement.