ChatGPT responses to actual code

[BarryG]

I just saw this Reddit post:

https://www.reddit.com/r/software/comme ... _code_and/

So I decided to throw a couple of PureBasic snippets to ChatGPT to see how it would explain them. I must say, its answers impressed me!

The last snippet about LoadImage has an error about what #PB_Any does, but it still summaries the code correctly as to its intended purpose.

[HeX0R]

I'm not sure, but I have the feeling, the more threads we have with "ChatGPT" in the subject line, the more of those damn bots are willing to be part of our community...

[Rinzwind]

Still not impressed. ChatGPT scrapes the internet for questions and human given answers and API's and just looks for matching keywords in your questions and then glues together text from its sources without mentioning where it came from. A bit like the chess programs, brute force database to lookup possible results. This is really obvious if you answer it to the point questions for a specific knowledge area. You see it pastes similar answers from a similar area, but totally wrong/not applicable. Just a matching algorithm and mixing answers from others. The bad part is, that it sounds legit at first sight if you indeed do not know the answer. Still fun to pay with, but dangerous for this age of internet simpleton users who cannot do their own research, comparisons and validations. Which movie did portray a blank future with humans not knowing anything anymore?

Let's put it this way: another case of misuse of the term AI.

[Mijikai]

What a waste of time.

[Little John]

Still not impressed. ChatGPT scrapes the internet for questions and human given answers and API's and just looks for matching keywords in your questions and then glues together text from its sources without mentioning where it came from.

And even worse: ChatGPT sometimes “cites” sources that do not exist!
See e.g. https://futurezone.at/science/chatgpt-w ... /402255681

[Bitblazer]

ChatGPT acts dangerously similar to an idiot imho. There is no real understanding of the code, but a lot of babbling and googling power to argue about it

It might be useful in 5 years and awesome in 10, but right now it just seems to be an interesting new toy.

I don't really see any benefit for a human expert who knows how to use google. It would be bad as a bot in a beginner chat, because it does give dangerously wrong answers sometimes and for an advanced chat, it would likely just be a nuisance. So where is the practical use right now for everybody else beside the developers of ChatGPT itself?

ps: ok, it could be a fun new "chat game" like "this is the question and here is chatGPTs answer. Point out where and why ChatGPT failed and actually made a huge mess that contradicts the original questions purpose. Combined with Bonus Points for correct quick, specific and short answers

[BarryG]

If you post my above questions to Google, you don't get (semi-correct) explanations like ChatGPT gave, or any explanations at all. That's what I mean by it's impressive. Even the simple "why does a=1/b fail" has no hits in Google to explain why it fails:



So for ChatGPT to actually explain why it fails, is impressive (to me at least). Maybe I'm just easily amused.

[jacdelad]

ChatGPT is scam like all AIs in general. They claim to be intelligent but are nothing more than well programmed and masked google searches.

[Mijikai]

Im sure all the data is appreciated

[KosterNET]

I agree with @BarryG. Even when considering that ChatGPT is nothing more than a lingual model, it still is very good able to translate questions that are not entered into the system on a structured way into a well-generated response.

Yes, it has it's perks. Yes, it depends on the data that was 'learned' into the system, but an incredible amount of data was input for the system. Therefor you get an anwer about an incredible amount of different subjects.

And for the 'AI'-name; in the contrary to systems where programmers program any amount of possibilities, these systems are programmed using an algoritm that uses 'learning-data' to build their dataset.

The system is not retrieving extra information from the Internet as I understood, what means that the system must contain an incredible amount of data. That also impresses me.

I also agree that 'intelligence' is debatable. But that also applies to a lot of people

[BarryG]

a lot of babbling and googling power

ChatGPT scrapes the internet for questions and human given answers

nothing more than well programmed and masked google searches

Not according to the FAQ -> https://help.openai.com/en/articles/682 ... ow-about-x

Quote: "[ChatGPT] cannot access the internet, search engines, databases, or any other sources of information outside of its own model."

[Bitblazer]

Not according to the FAQ -> https://help.openai.com/en/articles/682 ... ow-about-x

Quote: "[ChatGPT] cannot access the internet, search engines, databases, or any other sources of information outside of its own model."

They already work on "live searches" (aka googling - yes basically it is done by interfacing with a bunch of search engine API's, but that is just an implementation detail and might make it less responsive ..., but thats just a minor technicality to me, like the natural language interface that a tech mag already did - for the show effect).

[BarryG]

They already work on "live searches" (aka googling - yes basically it is done by interfacing with a bunch of search engine API's

You say it's googling; the developers say it's not - who do I believe? Why do I take your word over the developer's? Have you worked on the project to be able to make such claims?

[Bitblazer]

You say it's googling; it says it's not - who do I believe? Why do I take your word over the developer's?

I just use "googling" as a synonyme for using search engines data/api aka life searches. The rest is from the developers themselves. Call it "bing'ing" if you prefer

[BarryG]

But they've said they don't use search engines (Googling, Binging, etc). So why do you assert that they do? Do you have inside knowledge of how it's coded?