Re:(Databricks cofounder) How Is ChatGPT’s Behavior Changing...
by nomel on Tuesday July 18, 2023 @11:29PM
(#36781802)
> Llama 2 is spectacular.
Sure, but it's significantly behind GPT-4, and somewhat behind GPT-3.5. Could they make money on it, with a paid service, if they tried?
If they actually had something competitive, that say beat GPT-4, or approached best-in-class in the future, would the still give it away for free?
My naive, but seemingly obvious, assumption is that they are doing this first for actual R&D, because they are behind, second for promoting the use of AI, both with the end goal of profit from a closed source model that is actually competitive.
What motive do you see for them releasing these models for free, especially with the massive increase in cost associated with catching up?
Sure, but it's significantly behind GPT-4, and somewhat behind GPT-3.5. Could they make money on it, with a paid service, if they tried?
If they actually had something competitive, that say beat GPT-4, or approached best-in-class in the future, would the still give it away for free?
My naive, but seemingly obvious, assumption is that they are doing this first for actual R&D, because they are behind, second for promoting the use of AI, both with the end goal of profit from a closed source model that is actually competitive.
What motive do you see for them releasing these models for free, especially with the massive increase in cost associated with catching up?