Google “We Have No Moat, and Neither Does OpenAI”

Re:Google “We Have No Moat, and Neither Does OpenAI” by wing-_-nuts on Friday May 05, 2023 @10:46AM (#35829898)

There are two things that make a good LLM. The amount of data available for training, and the amount of compute available. Google's bard sucks in comparison to Open AI, and even compared to Bing. It's pretty clear that GPT4 has some secret sauce that's giving them a competitive edge.

I also don't think that Open Source LLMs are that big of a threat, for exactly this reason. They will always be behind on the amount of data and compute available to the 'big players'. Sure, AI will increasingly be incorporated into various software products, but those products will be calling out to big tech apis with the best model. There will be some demand for private LLMs trained on company data, but they will only be useful in narrow specialties.

[ Parent | on HN ]

Re:Google “We Have No Moat, and Neither Does OpenAI” by codethief on Friday May 05, 2023 @04:01PM (#35834427)

Did you read the article? It refutes almost every claim you're making and, I must say, rather convincingly so.

[ Parent | on HN ]

Re:Google “We Have No Moat, and Neither Does OpenAI” by wing-_-nuts on Friday May 05, 2023 @04:18PM (#35834631)

I'll admit, I skimmed it. I went back and re-read it, and the timeline of events was especially shocking. I still think the big models hold an edge, simply because it will most likely be better at handling edge cases, but wow, my days of underestimating the oss llms are certainly coming to a middle

[ Parent | on HN ]