The first is large standalone data centers processing cloud based systems where end users use terminals to communicate with the AI data processing center
This works now because high performance AI is proprietary to a few major players and computational overheads are high
As AI evolves, there will be a potential migration path for commodization of AI analysis tools facilitated by AI assisted coding tools with a corresponding increase in computing horsepower for stand alone systems to allow stand alone computers to run powerful AI algorithms on non cloud based systems where end users control their AI systems and computational overheads assets
This second path is death to AI monopoly that Altman wants build and control
Altman is also ideologically motivated to develop AI systems in a direction that facilitates social control along the lines of the high tech, neo Marxist Chinese Social Credit system and his cloud based system enables facilitates that collectivist goal
Computing horsepower has outpaced the needs of most users but AI systems are going to put a lot more pressure on hardware designers . With the current pace of technology, AI systems and very high computational power standards alone. Systems will quickly emerge
This computer which with its 128GB of fast unified memory and a Ryzen AMD Strix Halo Al Max+ 395 was selling for $2000 last November. It is very capable of running very capable AI LLMs. It is currently available on Amazon for $3650. But my guess is that the price will be dropping significantly in the not-too-distant future.
https://www.amazon.com/GMKtec-Computers-LPDDR5X-8000MHz-EVO-X2/dp/B0F53XL9DP?crid=2V728GRP43QC
Here is a video that describes recent breakthroughs using this hardware:
https://youtu.be/d4EWA6yd5cE?list=LL
summary: https://share.gemini.google/MCQtfE8CEkra