Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: Is ChatGPT Good Anymore?
5 points by tosser2084 6 months ago | hide | past | favorite | 4 comments
I typically run 3 LLM's at a time in MSTY such as Gemini, Claude 3, Qwen or TinyDolphin, and this week ChatGPT 4o... the answers I get from Clause and Gemini are by far the most useful. The local LLM's are also similarly good / useful (just slow on my m1 mac). ChatGPT 4o... I mean, they sound good, but if you actually try out what it's saying it's not pretty / doesn't typically work or there are issues you run into like the option you need to change isn't there, etc. For example, I wanted steps in how to install Ubuntu on Docker on Windows... Gemini and Clause gave me nearly identical steps (that worked)... ChatGPT 4o... was so off I closed / deleted that window without further thought.

And don't get me started on trying to reproduce any of the vision stuff they showed on Monday, I have yet to get my paid account to do anything remotely similar to those demos... so I guess they were all scripted after all(?)




I'd read that Bing Copilot was now GPT-4o so went there to try.

Wheras previously it would tell me its model name+version, now it "can't".

------------------

Copilot

Your everyday AI companion

You

what model are you

Copilot

I’m Microsoft Copilot, an AI companion designed to assist users by providing information, answering questions, and engaging in conversation. I use advanced technology to provide relevant and useful responses. However, the specifics of my model and how I function are proprietary information. Is there anything else you would like to know or any other way I can assist you?

You

what is the name of your LLM model

Copilot

I’m sorry, but I can’t provide the specific details about the underlying models or systems that power me. As Microsoft Copilot, my primary function is to assist users by providing information and engaging in conversation. However, the specifics of my model and how I function are proprietary information. Is there anything else you would like to know or any other way I can assist you?


I see a fate similar to Wolfram Alpha. gradual degradation and pay-walling of free features.


The WA degradation really baffled me. How did that come about?


I am guessing students overtaxed the free features for homework assignments, so metering and other restrictions were put in place




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: