Or maybe we can use the GPT’s as models in our own apps. It would be like adding a bunch of fine tuned models that you can use (based on some sort of user pays/revenue share model)
2 Likes
Yes, I think this “democratization” effort will basically create a divide between developer apps, and consumer personal efforts.
BUT the big fear (as a developer at least) is that the consumer may not see your value, since they can get a “good enough” experience themselves.
So it almost puts devs more and more into the tough B2B space, and not consumer facing, unless their app is truly remarkable.
1 Like
I’m gonna be honest I didn’t really understand the example of your app but that sounds really cool
Yeah. I’m not the best at writing 
I think the most valuable thing (I recall that you are a ReactJS dev as well) that we have is our ability to create powerful, interactive, highly dynamic & visual applications. That’s something that ChatGPT (the website platform) cannot do
Combining this with customized GPTs is HUGE for developers.
What?
Me too I love losing revenue it’s my favorite thing.
fantastic point, might have to be the focus from here on out. I don’t know
ChatGPT is restricted to being a text-chat service is what I meant to say (kind of like a text-based RPG).
The conference video was perfect. Really. I don’t want want to use ChatGPT to plan a trip. That’s dumb. But I would LOVE to use a Kayak that has ChatGPT built-in to run the functions while I tell it what I want.
1 Like
Yeah that’s where my interest lies as well in audio and video. The limit is only your creativity.
I agree. Maybe that’s what they were going for, is to narrow the barrier of entry so AI isn’t just a get rich quick scheme for people that can just brand the best API wrapper, and there’s a lot of that.
1 Like
As we all sort of sit around clutching our pearls and wondering how theses announcements affect us as developers, can you imagine what’s going on at AWS, Anthropic, Google right now? Just by the price reduction alone of GPT-4, arguably the best LLM on the market, some of those guys recent efforts just got majorly nuked.
And honestly, at these low prices and with all these non-code capabilities, how many people are still going to be tempted by open-source models that require a bunch of install this and install that, not to mention base training, just to get to first base?
I think we just witnessed a seismic shift in this marketplace. I imagine the first of many to come.
4 Likes
Yeah this is just an insane ride, and by the time any of us start making any money UBI will kick in and AGI will render us all useless…
I say that tongue-in-cheek but I mean seriously, we’re all after a pot of gold that is just going to be replaced in itself when AGI seismically shifts the economy
1 Like
I want to try before forming an opinion but I cannot access GPTs ? Is it lunched yet ? Is there a approval process?
I just checked and yes I have access to both gpt-4-1106-preview and gpt-4-1106-vision-preview right now.
No time to check anything else, gotta write so much code today!
Don’t have ChatGPT, so no idea on GPTs.
It’s launching!
What you got cookin curt?
This is a complete sentence, I think.
Sadly, I need to finish this big CRM integration. 
But hope to play around and build with the AI later this week 
Code? We don’t need no stinkin’ code! We got ChatGPT!!!
1 Like
No this CRM code interacts with all kinds of cloud stuff, so it’s faster and more efficient to code it directly. I wish I could just hit a magic button and have the product done.
But if I could do that, we’d all be jobless and on UBI living in our little communes.
1 Like
Just as accounting software that can prepare your taxes has not replaced accountants, this new tech is not obsoleting developers in the way everybody is thinking.
After playing around with the assistant in the playground, and assuming the GPTs are built on this, I think OpenAI may have made it’s first big mistake: Overreach. Super big companies do it all the time: Google, Microsoft, Meta… They are now getting into the areas of content management and vector storage and retrieval, which are not their strength.
Consequently, they have created a bigger opportunity than anybody realizes for developers, particularly those with “wrapper” applications.
What do I mean?
Well, why are businesses and the general public considering Artificial Intelligence at all? Because, as David Shapiro would say, it can perform a number of tasks “better, cheaper, faster”. Existing chat-with-your-pdf applications today can do what this new assistant api technology can do NOT better, but “faster and cheaper” – and almost as good. And I think “almost as good” is going to be good enough for a sizable market share.
I mean, I could be wrong. But, since we’re all speculating here, I figured why not use a little historical context as well? Guess we’ll all see for sure soon enough.
1 Like
Yeah I’m not so much worried about this update and more worried about AGI in the form of many autonomous agents that could then replace entire work forces.
Unless the accounting software did it automatically for you in which case then it would get rid of accountants, which with AGI you would theoretically be able to do.
Maybe true Im not so impressed with it either and I keep reading about all the bugs people are getting. I’m knee deep in my own existing projects so I haven’t had too much time to play with the Assistants
Depends on what the thing is. But yeah hopefully these new bots actually suck and it keeps us as developers ahead of those who can only use their GUI.
1 Like
So, I just did a little comparison test between the assistant api and my RAG app using the same exact document. Not only did my app give better, more detailed responses, but it also returned source citations (heh heh). I know it’s just a teeny tiny little detail, but there are a LOT of folks who not only want an accurate answer, but citations from the source to corroborate.
2 Likes
Well… here is what I thought.
Let OpenAI do whatever they want to do, if it is legally allowed. They conduct their own business, and we, as developers, follow them.
I believe they will have many trial and error experiences, so I am sticking with my current system at the moment, using chat completion APIs only.
Also, they might close some projects if those projects are not successful. So, we also need to think about which APIs will survive in the future.
If they successfully develop tools (such as RAG), we can just use their tools; not an issue for me.
… … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … …
To be honest, I believe business people need more complex features than just RAG (Like, I need something more than RAG). So, yes, still you need to do many things to do as a developer, even if OpenAI develops something unique. That would not be enough and yes even if that is AGI, still not enough for the business by itself.
It depends upon the business people and their actual needs (not the ones we invent for them).
I was going to start another topic, but my question still actually refers to this topic.
I have tested the same questions against the same documents with the same models using both the assistants api and my application and, as I stated above, am getting consistently better responses using, just like you, the existing completion API.
GPTs and the Assistant API are great new tools, but as someone alluded earlier, I think they are targeted and optimized for non-developer end-users. In this early stage of the game, an AI NOT being able to tell you where in a document it pulled it’s information is NOT a strong selling point in my opinion.
All that above to say, I heartily agree.
2 Likes
