To disarm the big tech bomb, including “AI”, the public needs to be on high alert — as AI firms start to visibly wobble, we need clear demands and firm commitments from our leaders that if it all comes crashing down, the public won’t be the one left holding the bag.
I don’t know that there needs to be anyone bailed out here. Companies with viable business models making things and services people will use and actually need will probably still be around.
The dot-com hysteria eventually washed out a lot of companies - the rather clueless people claimed “it was all a fad” also oversteered as much as or more than the people pumping up bullshit like pets.com and idiotic VCs that were throwing money at people because they were in their twenties and because they could spell HTML.
What remained were obviously viable business models and of course the internet was not and is not a fad.
I suspect the same for LLMs. People saying it will all go belly-up are either huffing serious copium and/or absolutely clueless or have an axe to grind. It’s just not true. This stuff is definitely not going to go away. What form it may take, what the business models are and so on…that remains to be seen.
I could see all these hyperscalers having way overextended themselves finding that companies and individuals are standing up their own systems to run models. I can also see how more and more efficiencies are gained such that people don’t need quite the hardware - or the subscriptions - that these companies are hoping they do. Of course, neither the models, nor the hardware, nor the ways that they are being used by, say, things like opencode or Claude code have been standing still. For instance, there is this open source project, that runs qwen 3.8 even faster - https://www.mtplx.com/
Even for paid models, you look at the costs of tokens over time, they have dropped, like a rock, since 2023. Does Anthropic, etc., think they are going to be able to raise prices on people to cover their costs?