How does ChatGPT differentiate itself from other AI language models in terms of providing unique answers to user queries?

[prateekkumar]

Hello everyone,
I am using ChatGPT but whenever I ask any question related to web development that time ChatGPT provides same answer. If you have any idea then share me.

 

[ocnbrze]

why not ask chatgpt this question.........lol

 

[techfunhj]

Make sure that your prompts are descriptive enough otherwise the AI won't know what you want.

 

[Harro5]

why not ask chatgpt this question.........lol

Don't slam the gentleman

 

[Hadron]

why not ask chatgpt this question.........lol

One good reason not to is that these LLMs make garbage up and regurgitate fallacies from their training samples. I've come across several examples in the last week, including Bard getting high school physics spectacularly wrong and one of my students trying to use ChatGPT to research something and discovering that much of what it tells you is rubbish (which is OK if you are smart enough to check and not just assume that because it can produce good English that the content is reliable - though if you have to check everything it says it's of limited use).

 

[EdwardMatthew]

Ensure that your prompts are sufficiently descriptive otherwise, The AI may not comprehend your intentions accurately.

 

[Hadron]

Technically the "AI" doesn't comprehend anything. It's just something that generates text based on probabilistic associations from its inputs. It has no concept of true or false information, no understanding that either the input you give it or the words it produces as output actually mean anything, and actually no concept of "meaning" at all.

It's true that the more detail you put in your question the more associations there are for generating a response. But don't forget that that also means more scope for "making up" answers that have no basis in reality.

 

[Davdi]

I've been playing with some of the AI image generators (Dall-E and others at tne night cafe website. Sometines a basic description gets really good results. It also uses a random seed, and changing that can produce dramatically different results, but still following the original prompt text,

 

[ocnbrze]

occassionally?
the AI just ruined a lawyer who was to lazy to do the work themselves.

 

[Hadron]

Being a lawyer doesn't mean you aren't an idiot. My favourite was a law firm that went under some years ago because one of the partners thought that a proposal to help a "Nigerian prince" move some money around sounded like a good deal to them, but decided to use the firm's bank account rather than their own...

I see that the post you were replying to has disappeared (I suspect along with the member). But yeah, "occasionally" doesn't cover it: that thing constantly makes stuff up. Indeed that phrase was one of the things about that post that made me think "I wonder where this was copied and pasted from?" or "did they get ChatGPT to provide this text?" (claims about it "understanding" and hence giving more accurate answers also had me thinking "would you like to buy a bridge from me?").

 

[PearlMiller]

ChatGPT stands out by its extensive training on diverse data, making it adaptable to various topics. Its ability to understand context and generate human-like responses sets it apart. OpenAI constantly updates and refines the model, ensuring it delivers accurate and unique answers to user queries, making interactions more engaging and informative.