[komposit]

This skepticism absolutely baffles me. Have you been using gpt-4? To unlock gpt for real you have to be careful to prompt it correctly and find a way to improve the feedback loop for improving code. It is only a matter of time until tools arrive that integrate this into your development environment and give it access to test/console output such that it can suggest code and iterate over the result. It's not perfect yet, but I'm seriously feeling the nature of our work will change fundamentally over two years already.

[illiarian]

So... nothing changes. It will be the tool for which you will need to manually construct prompts and clean up output (including imagined non-existent APIs).

The availability of a button inside an IDE doesn't make this a fundamental change in how we work

[cortesoft]

I don’t know, I feel like it really does change how we can interact with a computer.

It feels like we are headed to a world where we can interact with a computer much more like they do in Star Trek; you ask the computer to do something using plain English, and then keep giving it refinements until you get what you want. Along the way, it is going to keep getting better and better and doing the common things asked, and will only need refinements for doing new things. Humans will get better at giving those refinements as the AI gets better at responding to them.

It is already incredibly good for being such a new technology, and will continue to rapidly improve.

[visarga]

The step up in accuracy from one shot solutions to iterative ones is large.

[deeviant]

If the button can do, let's say, half or more of the work for you when you press it, you're lying to yourself if you think it won't change anything.

[dbtc]

Nothing changes the same way that there is no difference between writing software in assembly and writing it in python.

[crop_rotation]

It is so far ahead of even what the best IDEs do. For one, I have not seen GPT4 ever use non existent APIs. You don't need to carefully construct prompts. It tolerates typos to a good extent. You can just type a rough description and the output won't need cleaning manually. You might need to reiterate it to focus on some thing (like remove all heap allocations and focus on performance).

[49531]

I've seen it use non existent APIs a lot. Working on a project that uses a dialect of python it told me it knew (Starlark) was like pulling teeth. It would tell me to use a python feature Starlark didn't have, I'd ask it to rewrite it without using that specific feature and it would with another feature Starlark didn't have access to, so I'd ask it to write the solution using neither and it would just give me the first solution again.

[illiarian]

> For one, I have not seen GPT4 ever use non existent APIs.

Have you asked it to use any API that appeared after September 2021 (that's the cut off date for its data)?

Have you asked it to write code in less popular languages (e.g. Elixir)?

Have you asked it to write code for less popular or unavailable APIs (smart TV integrations)?

[leishman]

Yeah it was basically useless for an Elixir project I was working on. That will probably change at some point I’m sure.

[crop_rotation]

I have used it to write Nim and Zig code (both not too popular languages).

I also asked it to write using non existent but plausible sounding APIs, and it flat out says "As of my knowledge cutoff in September 2021, I have no knowledge ...."

Ae you talking about GPT4 or the default ChatGPT?

[illiarian]

I've seen similar claims about GPT 3.5 and Copilot, so I won't hold my breath.

To quote GPT-4 paper:

"GPT-4 generally lacks knowledge of events that have occurred after the vast majority of its pre-training data cuts off in September 202110, and does not learn from its experience. It can sometimes make simple reasoning errors which do not seem to comport with competence across so many domains, or be overly gullible in accepting obviously false statements from a user. It can fail at hard problems the same way humans do, such as introducing security vulnerabilities into code it produces.

GPT-4 can also be confidently wrong in its predictions, not taking care to double-check work when it’s likely to make a mistake".

> I also asked it to write using non existent but plausible sounding APIs, and it flat out says "As of my knowledge cutoff

Ask it to write a deep integration with Samsung TV or Google Cast. My bet is that it will imagine non-existent APIs (as those APIs are partly unpopular and partly closed under NDAs)

[raincole]

How do you know GPT4's cut off date...? I mean it says that, but it can totally be it "learned" its (supposed) cut off date from the GPT3.5 output all over the internet, right?

[illiarian]

> How do you know GPT4's cut off date...?

"GPT-4 generally lacks knowledge of events that have occurred after the vast majority of its pre-training data cuts off in September 202110, and does not learn from its experience."

GPT-4 paper, page 10: https://arxiv.org/pdf/2303.08774.pdf

[crop_rotation]

The model repeats it all the time "As of my knowledge cutoff date"

[raincole]

Yes, and this fact doesn't tell me anything, as I know LLM is completely capable to say things that aren't true.

[angarg12]

Is not skepticism, it's curbed optimism.

I don't feel that my job is at risk of disappearing. Instead I think we'll be using LLMs as tools to do our job better.

[elif]

I think it's safe to assume anyone trying to criticize chatGPT who has access to gpt4 would specify that their attempts are using even the latest and greatest. The disclosure is in the interest of their core argument.

Therefore the inverse can be safely inferred by nondisclosure.

[orangesite]

"You're holding it wrong"

[vineyardmike]

There’s a difference between the iPhone “you’re holding it wrong” argument and not using a tool correctly. If you try to hammer a screw, it may enter the wood but that doesn’t mean it’s the correct way to use it.

[ilaksh]

I am working on this. Broke so have to do odd GPT jobs from Upwork to make ends meet so paused on development. But the front end stuff works. At least as far as skipping copy paste.