Just wanted to highlight this as such a great, concise way to look at the Buy vs Build with pretty much any cloud service, thanks!
Once marketing gets in charge of product, it's doomed. And I can't think of a product startup that it hasn't happened to. Particularly with this type of growth, at some point, the suits start to out number the techies 10:1.
This is why openeness and healthy competition is primordial.
If you set money on fire -- eventually there's a time when you need to stop doing that.
Yes, OpenAI might be (we don't know how much) burning through their $5B capital/Azure credits now, but I think the `turbo` models are starting to addressing this as well. And $20/month from a large user base can also add up pretty quick.
> This is very early in the maturity cycle for this tech.
Think about what value you get out of the services and what migration might look like. If you are making simple completion or chat calls with a clever prompt, then migration will probably be trivial when the time comes. Those features are the commodity that everyone will be offering and you'll be able to shop around for the ideal solution as alternatives become competitive.
Alternately, if you're handing OpenAI a ton of data for them to opaquely digest for fine tuning with no egress tools, or having them accumulate lots of other critical data with no egress, you're obviously getting yourself locked in.
The more features you use, the more idiosyncratic those features are, the more non-transferable those features are, and the more deeply you integrate those features, the more risk you're taking. So you want to consider whether the reward is worth that risk.
Different projects will legitimately have different answers for that.
Time to market is more important, you build users you get an edge, you can swap models later on (as long as you own the data).
There's very little there there in most of these folks.
(For the few where there is, though, I agree with you.)
I use it instead of Bing Chat now for cases where I really need a search engine and Google is useless. Mainly because it's faster, but I also like not having to open another browser.
This is very early in the maturity cycle for this tech. The options that will be available for private inference and fine tuning, for cloud-gpu/timeshare inference and fine tuning, and for competing hosted solutions are going to vastly different as months go by. What looks like squeezing value out of OpenAI today might look a lot like technical debt and frustrating lock-in a year from now.
That's what they're hoping you chase after, and if your product is defined by this technology, maybe that's what you have to do. But if you're just thinking about feature opportunities for a more robust product, judiciousness could pay off better than rushing. For now.