I think the problem is that people working in the field have a much different definition of what "fully-functional" means than users who believe that they're communicating with an intelligent, all-knowing or otherwise infallible being.

For my use case I use it for input processing, it’s just amazing that way. Hallucination is a huge problem, but you can avoid it.

I believe it. For text->text workflows it's a game changer. For anything touching a physical systems, even the highest-level path planning for robotics, there's a long way to go.

It’s all about being very deliberate with your use. People are getting ahead of themselves right now for sure. It’s still a monumental leap in UX!

Fully functional in what way? As far as I can tell, ChatGPT is a box that I put sentences into, and I get grammatically correct sentences that contains topics or words loosely statistically correlated to what were in my sentences, that may or may not be correct and often are not. The box has little to no memory. I honestly don't see what's so useful about this box.

The results are approximately as good as asking an intern, but the response is approximately instantaneous, where an intern takes a week to do anything. And similar to the intern, you can get better results with a bit of guidance and iteration.

In short, humans kinda suck. LLMs also kinda suck, but faster than humans.

> humans kinda suck

Agreed, but that's part of my critique. These systems are written by humans and are trained with barely curated data generated by humans. There is the concept of emergence, but I'm not sure how emergence suddenly fixes a terrible foundation full of biases and errors.

> and I get grammatically correct sentences that contains topics or words loosely statistically correlated to what were in my sentences, that may or may not be correct and often are not.

It's correct enough to pass the bar exam, medical exams, it scores 90-93 percentile on the SAT. This is way more complex and efficient than what you make it to be imo.

> It's correct enough to pass the bar exam, medical exams, it scores 90-93 percentile on the SAT.

So is Google Search, and we had that for a long time now. Is a slightly different and more verbose UI really a game changer?

(Let's ignore the fact that Google Search is broken from all the SEO spam and monetization. Especially when we have no evidence that ChatGPT is any more resitant to this than Google was.)

Google Search scores 93% on the SAT? On questions that have no solution online?

When hooked up to plugins or tools it does feel like a fully functional interface. With access to a browser, it can do basically whatever I want it to.

I hooked GPT-4 up to a shell and asked it to use the GPT-3.5-turbo completions API (not in the dataset), and it successfully did it through trial and error with curl and error messages. This example is of course not something you would actually do regularly, but rather shows that you don't need a lot of context for it to do useful things right now. With a complete explanation of the OpenAI endpoints, it would most likely make the request perfectly on the first try.

Yes, exactly this is what I meant. You can really leverage the API in other systems. CharGPT on its own is cool, but my mouse comment was about NL being a bolt on UX tool. Which is crazy!

You can teach it to act as a support personal.

'only answer the following questions and use these API endpoints doing the task'.

Now you have a multi lingual human Isabel Interface.

Have you seen the Wolfram alpha examples?

You can tell it what you need to do in terms of data processing, and ask it to write Python code that does that. E.g. ever had to write a convoluted ImageMagick command to do something complicated? GPT will write it for you.

Think of it as a natural language interface to anything that has an API.