Ask HN: Why did Google threaten its own business by publishing the GPT paper?
Maybe Google researchers have too much freedom in what they can publish?
14 comments
I am not convinced that ChatGPT is all that much of a threat to Google. Exactly how difficult would it be for Google to reproduce ChatGPT? They have access to all the training data and crawl it regularly. They have the infrastructure in place to make the model and integrate it into search. What am I missing here?
Unless OpenAI has some IP they cannot use it would seem that they have done Google a favor by demonstrating the potential of the technology. If my memory is correct one cannot patent an algorithm.
Unless OpenAI has some IP they cannot use it would seem that they have done Google a favor by demonstrating the potential of the technology. If my memory is correct one cannot patent an algorithm.
ChatGPT's threat to Google isn't that Google can't match it--it's that it changes the approach to search results from displaying lists of results where it's "natural" to insert sponsored listings to a more conversational approach where ad opportunities will be rarer (or at least entirely different).
If users prefer the newer, interactive approach to exploring search results, Google's long-standing business model will require a significant update.
If users prefer the newer, interactive approach to exploring search results, Google's long-standing business model will require a significant update.
In my mind you might as well place sponsored content into a conversation and hide it maybe even better than in search results as a list.
Product placement.
Google already has something like it.
It also has imagine
It also has imagine
OpenAI made GPT, not Google. Did you mean "Attention is all you need", the paper that introduced transformers?
Yes.
Google did not "threaten their own business" by publishing basic research in machine learning. Products threaten businesses, not research.
But research does attract funding, and top tier employees.
Okay, and that funding and top tier talent was attracted to Google, right? I'm not sure what you meant by that.
Soviet scientist Pyotr Ufimtsev's research publications lead to the development of US stealth aircrafts like SR-71 BlackBird etc. Pyotr's research was ignored in the USSR.
https://en.wikipedia.org/wiki/Pyotr_Ufimtsev https://www.goodreads.com/book/show/101438.Skunk_Works
https://en.wikipedia.org/wiki/Pyotr_Ufimtsev https://www.goodreads.com/book/show/101438.Skunk_Works
ML Research is currently at a stage where a lot of low hanging fruit is available for picking. And if you're not quick (as in: publish), someone else will get to eat instead.
Google would probably attract a lot less talent if they didn't have such high-profile publications. People want their names on papers.
Google would probably attract a lot less talent if they didn't have such high-profile publications. People want their names on papers.
I don’t think ChatGPT is of any threat to Google. I am pretty certain that Google and maybe other tech companies such as Meta probably have their own text generation models that are as good as (if not better) ChatGPT. Why they haven’t released them to the public is anyone’s guess.
For the same reasons large companies release open-source software.