Re:OpenAI API
by wildermuthn on Friday June 12, 2020 @09:18AM
(#23498910)
Odd analogy to use with Microsoft Windows, since GPT-3’s source is available, along with a series of papers that enables anyone with the money and knowledge to implement it themselves.
Re:OpenAI API
by YeGoblynQueenne on Friday June 12, 2020 @09:35AM
(#23499065)
The reasons why MS windows and GPT-3 cannot easily be modified by anyone are different, but the result is the same: you're stuck with what you're sold.
To clarify: MS windows is closed source, but you can't very well train a large GPT model unless you're someone with the resources of OpenAI. So you're stuck with whatever they choose to train and make available to you.
The API allows you to fine-tune existing models on your own dataset [1]
[1] Cf second paragraph of https://openai.com/blog/openai-api/
Re:OpenAI API
by YeGoblynQueenne on Friday June 12, 2020 @02:48PM
(#23502377)
"fine tuning", i.e. transfer learning is still limited by the training of the original model.
For instance, if the original model is trained on English text exclusively and you want to fine-tune it on Greek text you are S.O.L.
Imagine MS saying they "democratised" operating systems because, hey, you can buy their binaries, so everyone can use their operating system. Compare that kind of "democratisation" with open source oSs.
No, the truth is that as more and more resources are necessary to wring the last few drops of performance out of the current generation of deep neural net models it is only large, well-funded companies that have the resources to innovate - and everyone else is forced to follow in their wake. Any expectations that progress would lead to "democratisation" of deep neural networks research has gone out the window.