AI Companies Should Sink or Swim on Merit without Government Intervention
There is lots of talk of the next Administration subsidizing the buildout of AI infrastructure. I am not in favor of the Federal Government funding the cost of training LLMs.
I am not in favor of the Federal Government funding the cost of training LLMs with direct subsidies nor with SEZs (special economic zones), which I am familiar with from my days investment banking days when I visited Technology companies located within SEZs in Beijing, China, nor by any other means. Generally I favor tax breaks, but I would prefer that the Federal Government not discriminate as to which companies or industries it awards tax breaks to.
From my perspective, all U.S. citizens and U.S. headquartered companies should not be required to pay Income Taxes. The U.S. Income Tax (personal and corporate), ought to be abolished and replaced with a Consumption Tax (i.e. Sales Tax).
If the Gen AI model builders are looking at a cost of $10 billion to train a frontier LLM in 2026, $100 billion for the following LLM generation 2 years later and $1 Trillion for the subsequent LLM generation, let them figure out how to raise the capital, or, how to train their LLMs less expensively. Perhaps Smaller Language Models (SLM) are the correct course to pursue rather than monolithic LLMs that are developed in a vacuum.
