Maybe because the products suck and there are very few business models for software? I'm gonna focus on the software product segment, and my evaluation might be flawed, but I think that software is an inherent race to the bottom for three important reasons:
1) Dependence on proprietary software is at best a short-term support boon, and in the majority of cases a crippling liability when the vendor changes support terms or goes under (personal experience talking).
2) Open source software is a slow but inevitable creep towards feature-matching and quality that has borne ample fruit already.
3) As ephemeral information with near-zero material cost other than labor, software finds most low-hanging fruit quickly, while higher-hanging fruit attract the formation of non-profit foundations to continue development; there just aren't many more interesting software products to offer.
Working business models in the tech industry seem to revolve around the provisioning of hardware because buying software just isn't an appealing proposition to customers in most cases.
Sure, LLMs are interesting software, but they're a perfect demonstration: Open source LLMs already exist and gather huge momentum when released; the field is still so new that novel and efficient solutions are not the exclusive realm of proprietary research; and because of these factors, business models can't revolve around interesting model capabilities so much as provisioning the hardware they run on.
1) Dependence on proprietary software is at best a short-term support boon, and in the majority of cases a crippling liability when the vendor changes support terms or goes under (personal experience talking).
2) Open source software is a slow but inevitable creep towards feature-matching and quality that has borne ample fruit already.
3) As ephemeral information with near-zero material cost other than labor, software finds most low-hanging fruit quickly, while higher-hanging fruit attract the formation of non-profit foundations to continue development; there just aren't many more interesting software products to offer.
Working business models in the tech industry seem to revolve around the provisioning of hardware because buying software just isn't an appealing proposition to customers in most cases.
Sure, LLMs are interesting software, but they're a perfect demonstration: Open source LLMs already exist and gather huge momentum when released; the field is still so new that novel and efficient solutions are not the exclusive realm of proprietary research; and because of these factors, business models can't revolve around interesting model capabilities so much as provisioning the hardware they run on.