Raw LLM Responses

At this point I do wonder whether our form of capitalism is eating its own tail. When manufacturing jobs left there was a surge into service jobs. But if service jobs go, what next? (Not that everyone was able to make the services transition, but many did and found new things to do that have benefited society at large). But I’m not the first person to point out that a company with no workers at all can only be sustained in a society where other companies have workers. The “economy” is us (humans) interacting with each other. If no companies have workers, then it follows that no one has jobs either, so how do the former attract and retain income?? In reality I think what will happen is that competition and supply/demand will drive the price of labour down to a point where it is cost effective to employ people rather than outsource to (very expensive — don’t forget!) general purpose AI services. Publicly available, non-specialist AI in this regard is essentially competition in the labour market. What remains to be seen is what the real cost of AI is, per human watt-hour equivalent, vs the amount produced. Up to this point investors have provided a lot of cash, which has artificially driven the cost of AI services down to the consumer. Whilst the cost of AI hardware fluctuates — and may well fall in future — right now a single GPU can be upwards of US $50k. If you take into account the lifespan (about 3 yrs) and the running costs, then — excluding model training entirely — a single GPU can cost $170k over three years. So that is more than a junior person earning $50k/yr would cost you — and at the end of that 3 years in the AI case you have naught (the GPU is degraded) whilst with the junior employee you have instead an intermediate employee (they grew). If you add in profit margins, model training costs, and the cost of employing more senior people who are still needed to use the GPU, then the question becomes quite clear — how many humans can a GPU replace? Can it run fast enough, uplift the “productivity” of the senior employees high enough to make it cost competitive to hiring juniors? Some scientific studies say yes and others say no — and I think what that tells us is “it depends”. It depends on the tasks, the nature of the work, how effective the senior can be using AI vs a junior, how sustainable the senior’s skills are in using the AI in the long run. Short term upside dynamics (replace juniors with AI) may turn into medium or long term downsides (no seniors in the pipeline). Plus there is the simple fact of the energy efficiency of the biological brain being much much better, and displaying different qualitative and quantitative characteristics, to AI — in some cases AI will be better and in others human brains will be. So on economic grounds I am not 100% convinced just yet that AI will actually replace all services jobs, even if it theoretically could do. But I’d also draw attention back to what the “economy” is — us. It is humans interacting to benefit each other. Those humans don’t go away if their current jobs do. Even if the world ends up with people laid off all of a sudden, then at least a few of them will find something new to do and very quickly people will start working with each other again. And when they do maybe they’ll boycott AI? Or maybe we’ll just find some new medium ground. Lol — who knows!