3 ms·
OpenAI’s moat is stronger than you think
- verdverm 3y agoI disagree with the author's points, they seem to be missing that the open source is catching up and moving faster, that eventually smaller models will be on par with these super sized models. They are basing their arguments on the situation today, things are moving at an incredible pace. > I don’t believe that there will be free alternatives that actually are comparable in quality There is so much effort going into this, it's hard to believe the companies will be able to keep up. > On the B2B side, having a robust API that meets enterprise SLAs is not trivial. OpenAI has not done great at production, throwing errors regularly and having security issues. Further, they don't have a strong backwards support story. Having built a prompt, then having them update their model, our prompt broke. This is a big issue missing from the discussion. These LLMs are inconsistent at one version, and the versions are changing under us without the ability to lock for more than 3 months. OpenAI got an early lead, but that does not mean it will dominate. The one thing they have going now is that Microsoft owns more than anyone else and they will be shilling it hard. I suspect many businesses will rely on their cloud provider products and APIs, rather than OpenAI, though the brand argument has some weight. We are waiting excitedly to see what GCP offers with PaLMv2 and the VertexAI product. We want to move over ASAP, OpenAI is not meeting our needs.
- two_in_one 3y agoI think we are getting into a second stage of LLMs development. The applications. That's where OpenAI has only half of expertise needed. The other half is problem specific. Except for the trivial once.