At present, the models are being replaced with better ones on very short cycles. While open models exist and are even useful, they're generally a few months behind the closed models in performance.
If you are (and most of us are) still encountering cases where the closed models are not good enough, the open models are (mostly) even worse, so they're not competition.
I anticipate the money will run out, and when it does the open models catch up (that appears to be cheaper than maintaining a leading position); at that point, I expect us to face the "commoditise your complement" situation.
Yeah, and also the capabilities are starting to plateau. So I feel like the market will be flooded with a 100 Anthropic like providers giving out models.
The true cost/token is something that I am still wondering about. The most expensive chinese model costs $3 per input to run. And we still have massive markup from the american closed source model companies.
I agree that money will run out. But what will policy do?
At present, the models are being replaced with better ones on very short cycles. While open models exist and are even useful, they're generally a few months behind the closed models in performance.
If you are (and most of us are) still encountering cases where the closed models are not good enough, the open models are (mostly) even worse, so they're not competition.
I anticipate the money will run out, and when it does the open models catch up (that appears to be cheaper than maintaining a leading position); at that point, I expect us to face the "commoditise your complement" situation.
Yeah, and also the capabilities are starting to plateau. So I feel like the market will be flooded with a 100 Anthropic like providers giving out models.
The true cost/token is something that I am still wondering about. The most expensive chinese model costs $3 per input to run. And we still have massive markup from the american closed source model companies.
I agree that money will run out. But what will policy do?