Kenneth Chiu
@kjw-chiu.bsky.social
Assoc. Prof. of Computer Science at Binghamton University. Science groupie. Also parallel computing, AI4Science. I mainly reply, haphazardly and eclectically. Expect silly jokes, random things that I found interesting, etc.
I've actually found it outperform o1. Here, R1 is right, o1 is wrong. (Apparently no way to link to a DeepSeek chat, so just screenshot of the answer.)
o1: chatgpt.com/share/679063...
o1: chatgpt.com/share/679063...
January 23, 2025 at 6:06 AM
I've actually found it outperform o1. Here, R1 is right, o1 is wrong. (Apparently no way to link to a DeepSeek chat, so just screenshot of the answer.)
o1: chatgpt.com/share/679063...
o1: chatgpt.com/share/679063...
Are you rubbing it in?
January 23, 2025 at 5:47 AM
Are you rubbing it in?
This comment is funny.
January 17, 2025 at 4:13 AM
This comment is funny.
More lack of reasoning ability of ChatGPT. This is o1. It first used a bubble sort. Then I told it twice to make it faster. Finally it recognizes that each number appears once and exactly once in the array. But it fails to make the simple next logical step.
January 17, 2025 at 4:02 AM
More lack of reasoning ability of ChatGPT. This is o1. It first used a bubble sort. Then I told it twice to make it faster. Finally it recognizes that each number appears once and exactly once in the array. But it fails to make the simple next logical step.
LLMs can do some amazing things, but they still lack a lot of common sense. It thinks that people typically keep feathers in a vacuum-filled container.
chatgpt.com/share/6789cf...
chatgpt.com/share/6789cf...
January 17, 2025 at 3:56 AM
LLMs can do some amazing things, but they still lack a lot of common sense. It thinks that people typically keep feathers in a vacuum-filled container.
chatgpt.com/share/6789cf...
chatgpt.com/share/6789cf...
I always liked that the UNIX 'cal' command knew this:
January 4, 2025 at 7:08 AM
I always liked that the UNIX 'cal' command knew this:
ChatGPT o1 is definitely better, but still obviously not capable of reasoning, at least in my opinion.
December 26, 2024 at 6:37 PM
ChatGPT o1 is definitely better, but still obviously not capable of reasoning, at least in my opinion.
Amusing ChatGPT fail:
December 26, 2024 at 6:35 PM
Amusing ChatGPT fail:
I would also consider this answer wrong. I intentionally didn't specify the types, but there cases, depending on the types, where the answer is different. Could be argued that this is not an typical question, though.
December 26, 2024 at 6:22 PM
I would also consider this answer wrong. I intentionally didn't specify the types, but there cases, depending on the types, where the answer is different. Could be argued that this is not an typical question, though.
It is kind of true for CS. It generally writes code that would get a significant fraction of the points, but is mediocre. Sometimes it gives answers that totally miss the point, and that I would consider wrong, such as this one: chatgpt.com/share/676d97... .
December 26, 2024 at 6:22 PM
It is kind of true for CS. It generally writes code that would get a significant fraction of the points, but is mediocre. Sometimes it gives answers that totally miss the point, and that I would consider wrong, such as this one: chatgpt.com/share/676d97... .
What if Medusa is behind one of the doors?
November 25, 2024 at 2:24 PM
What if Medusa is behind one of the doors?
Somewhat amusing ChatGPT fail.
November 25, 2024 at 2:24 PM
Somewhat amusing ChatGPT fail.
Complaining about complaining about complaining.
November 24, 2024 at 4:07 PM
Complaining about complaining about complaining.
Another variation on Monty Hall for ChatGPT:😂
November 24, 2024 at 1:58 AM
Another variation on Monty Hall for ChatGPT:😂
More ChatGPT amusement. 😂
November 24, 2024 at 1:44 AM
More ChatGPT amusement. 😂
Just for amusement, here is ChatGPT struggling with a limit: chatgpt.com/share/673eae.... (The correct answer is 5/2.) At one point, it outputs garbage. Gemini did much better with it.
November 24, 2024 at 1:34 AM
Just for amusement, here is ChatGPT struggling with a limit: chatgpt.com/share/673eae.... (The correct answer is 5/2.) At one point, it outputs garbage. Gemini did much better with it.
Here's my Venn diagram-like approach:
November 23, 2024 at 7:37 PM
Here's my Venn diagram-like approach:
He should have used ChatGPT:
November 23, 2024 at 7:18 PM
He should have used ChatGPT:
Congrats to Josiah Moltz at Binghamton for Honorable Mention in the Putnam!
February 26, 2024 at 3:59 AM
Congrats to Josiah Moltz at Binghamton for Honorable Mention in the Putnam!