Hacker News Reader

Comment by jliptzin | Hacker News Reader

I don’t understand arguments that say google will be left in the dust when it comes to AI. There’s no network effects like Facebook, it just takes a lot of money to train the models, which google obviously has. Do we really believe google, with all its resources and engineering talent, has no ability to develop something on par with OpenAI?

View on HN

They also already have best in class AI. Google's PaLM is a much larger model than GPT3 and even larger than most estimates of GPT4. They also have Imagen which blows the doors off of Dall-e 2. In no way is google behind OpenAI and they have access to WAAAY more training data and resources in general. Google just hasn't shipped a consumer or publicly available product with any of it yet.

> Do we really believe google, with all its resources and engineering talent, has no ability to develop something on par with OpenAI?

Yes, they’re a large, bureaucratic organization, that’s incapable of developing new products or avenues of growth.

For the sake of argument though, let’s say Google created a product on par with ChatGPT, they would just kill it after a few months like every other Google service that they’ve made.

In fact Google has a bunch of SOTA LLM models, some of them are (conservatively speaking) several months ahead of OpenAI's. It's really interesting to see those people who believe (or want to see) Google being an underdog in the area of machine learning, but it's actually not. It's something similar to say that Tesla is years ahead of Waymo or Cruise.

Google acquired DeepMind almost a decade ago. I'm not sure if anything significant has been produced yet? Maybe it has, and it's all behind the scenes.

Reply on news.ycombinator.com