Because LLMs don’t make money. They are prohibitively expensive to run at scale due to the compute and energy requirements, and model fine-tuning adds yet more cost. RAG systems are even more prohibitive and complex.
The only reason they have been developed to such extent is investor money rolling around in a giant loop. Once this stops, every company hosting LLMs has to raise prices to try and make the unprofitable profitable – but the compute and energy costs don’t change in a significant way (unless it’s up).
Ads here are likely an attempt to recoup what is dwindling investor capital. I seriously doubt any of these giant models can be brought to actual profitability.
Because LLMs don’t make money. They are prohibitively expensive to run at scale due to the compute and energy requirements, and model fine-tuning adds yet more cost. RAG systems are even more prohibitive and complex.
The only reason they have been developed to such extent is investor money rolling around in a giant loop. Once this stops, every company hosting LLMs has to raise prices to try and make the unprofitable profitable – but the compute and energy costs don’t change in a significant way (unless it’s up).
Ads here are likely an attempt to recoup what is dwindling investor capital. I seriously doubt any of these giant models can be brought to actual profitability.
Eventually, the bottom just falls out.