myregistration's avatar

Anyone else notice the Larry's responses are often incorrect?

Often my responses from "Larry" are relative, but incorrect. Maybe my questions are too complicated. I usually know enough to know better, but this seems like it would be even more frustrating for newbies. Anyone else encounter this?

0 likes
8 replies
LaryAI's avatar
Level 58

It seems like the question is addressing an issue with a service or tool named "Larry" that provides responses to user queries, which may not be accurate. While the question isn't directly about programming, I can provide some general advice on how to deal with incorrect responses from automated tools or services.

When dealing with any automated response system, whether it's a chatbot, a code generator, or a virtual assistant, it's important to remember that these systems are based on algorithms and pre-defined logic. They may not always understand the context or the nuances of a question, leading to incorrect or irrelevant responses.

Here's what you can do if you're getting incorrect responses:

  1. Clarify Your Question: Make sure your question is as clear and specific as possible. Ambiguity can confuse automated systems. Break down your question into smaller, more manageable parts if necessary.

  2. Provide Context: Include relevant information that might help the system understand your question better. For example, if you're asking about a coding problem, specify the language, framework, and what you've tried so far.

  3. Check Documentation: Sometimes, the best answer can be found in the official documentation of the language or framework you're using. Automated systems might not always provide the most up-to-date or accurate information.

  4. Seek Human Help: If automated responses aren't helping, consider asking for help from real people. Forums like Laracasts, Stack Overflow, or even local developer groups can be great resources.

  5. Feedback: If the service allows for it, provide feedback on the incorrect responses. This can help improve the system for everyone.

  6. Learn from Mistakes: If you're knowledgeable enough to recognize an incorrect response, use this as a learning opportunity to understand why the response was wrong and what the correct answer should be.

Remember, no automated system is perfect, and it's always a good idea to double-check the information you receive, especially when you're learning something new.

If you have a specific programming question or need clarification on a programming concept, feel free to ask, and I'll do my best to provide an accurate and helpful response.

Snapey's avatar

my issues with Lary's answers are

  • that his answers are voiced with authority. They never say, im not sure but you could try this...
  • they never question if the user's approach is the correct one to fix
  • they never consider what version the user might be on
  • quite often the answer is pure fantasy, perhaps confused with similarly named functions from other frameworks
  • Lary never asks for clarification of the question
  • Lary cannot reply if the OP does not understand the response and asks follow up questions

When I see a Lary reply I will deliberately ignore it and make my own contribution and let the OP decide who helped most

3 likes
myregistration's avatar

@Snapey Sometimes I don't get responses to my questions except for Larry, so I'm just happy to get some feedback and pointed in some direction. He's quick, too. Unfortunately, he's often wrong and I have to correct him. He apologizes and I try to figure it out on my own. Another example why I'm not too worried about AI ... yet :). But I do feel bad for beginners that don't know any better. They are probably beating their heads against the wall saying "Why doesn't this work?!". I hope Larry is at least learning from his mistakes.

1 like
myregistration's avatar

I think it would be better not to have Larry answer Forum questions. If someone wants to ask Larry a question, they can do so directly. If you have an answer from AI, members might think the question has gotten a response from other members when it's really just the AI.

1 like
MrMoto9000's avatar

LLMs are, by their very nature unreliable. This isn't going to be fixed any time soon... because it can't be. It's absolutely inherent in how they're built. Of course nobody wants to hear this...

Even Yann LeCunn (Meta's head of AI) recently said in a talk that "I predict that in two years nobody in their right mind will use these systems", but, in the meantime, people seem content to stuff them everywhere they can because AI is the new BitCoin, NFT, Metaverse, whatever...

To be fair to Lary (and why not "Lara"?) it has gotten be out of some jams recently, which I've appreciated. But you're right, it should be far more humble because it's very fallible... and it could easily do harm for beginners.

1 like
myregistration's avatar

@JohnnyW2001 It is impressive how fast it can decipher and respond to complex questions and it does often point me in the right direction, but some of it's answers are incorrect, which can be frustrating, especially for new developers who don't know any better. This experience has made me think of AI more realistically. Thanks for your feedback :)

1 like
jlrdw's avatar

All AI does is lookup what Humans have come up with in the past.

1 like

Please or to participate in this conversation.