Have all the new features implemented into OpenAIs ecosystem "damaged" RAG based application development on the whole? I kind of get the feeling that OpenAI looked at this growing cottage industry, copied their open source features, and is now laughing all the way to the bank while leaving many dev teams in this space scrambling to pivot their ideas.
What's the point of building things in this space these days when OpenAI will just absorb those innovative ideas? Am I just looking at this wrong?
I'm asking cause my own solo project is having a bit of an identity crisis at the moment and I guess I'm just feeling self conscious about the whole thing lol - I'd love to hear other devs opinions on this. Thanks in advance!
I don't think so and if anything it's positive. We just got a bunch of new tools that will allow us to build even cooler projects. Change is always scary but after the hype of these releases settles, I think people will be more grounded again with their expectations. The market for RAG is growing, not shrinking.
Gauging from the sentiment of some of the clients I work with, they're super excited actually. Most of them have a bunch of new ideas on what they want built and are excited for the future. In reality, a lot of these things still aren't possible and these latest OpenAI releases don't change that.
I think we're still far away from a place where this kind of RAG deployment would be obsolete (talking AGI here). But I do think it's natural to be concerned with these types of releases, just cant lose focus of the big picture.
Obviously there are some things you want to avoid as a developer in this space like vendor lock in with LLMs, building products that can be squeezed out easily by large companies (think UIs for fine-tuning or apps like pdf.ai). But even these seemingly basic companies like pdf.ai are doing totally fine even with OpenAI releasing these features.
@Teemu hey genuinely thanks for that response. I've been teetering between feeling like all hope is lost vs being excited at the mainstream exposure of these new OpenAI tools, but your response has me going the more positive route for sure
There are hard problems to solve with RAG (the same problems we and everyone else is trying to solve).
It feels like OpenAI is going after a different audience and leaving developers behind a bit. Their current tools just obfuscate settings and leave little room for customization. People dislike black-box approaches, especially when things aren't working well.
Our goal is to empower developers to understand and build the best ways to connect LLMs to their data. There's no one-size-fits-all option for this
