Emergent Analogical Reasoning in Large Language Models
- Taylor Webb, Keith J. Holyoak, Hongjing Lu
In human cognition, this capacity is closely tied to an ability to reason by analogy.
Our results indicate that large language models such as GPT-3 have acquired an emergent ability to find zero-shot solutions to a broad range of analogy problems.
We found that GPT-3 appears to display an emergent ability to reason by analogy, matching or surpassing human performance across a wide range of problem types. These included a novel text-based problem set (Digit Matrices) modeled closely on Raven’s Progressive Matrices, where GPT-3 both outperformed human participants, and captured
a number of specific signatures of human behavior across problem types.
Because we developed the Digit Matrix task specifically for this evaluation, we can be sure GPT-3 had never been exposed to problems of this type, and therefore was performing zero-shot reasoning.
https://arxiv.org/pdf/2212.09196.pdf
- Taylor Webb, Keith J. Holyoak, Hongjing Lu
In human cognition, this capacity is closely tied to an ability to reason by analogy.
Our results indicate that large language models such as GPT-3 have acquired an emergent ability to find zero-shot solutions to a broad range of analogy problems.
We found that GPT-3 appears to display an emergent ability to reason by analogy, matching or surpassing human performance across a wide range of problem types. These included a novel text-based problem set (Digit Matrices) modeled closely on Raven’s Progressive Matrices, where GPT-3 both outperformed human participants, and captured
a number of specific signatures of human behavior across problem types.
Because we developed the Digit Matrix task specifically for this evaluation, we can be sure GPT-3 had never been exposed to problems of this type, and therefore was performing zero-shot reasoning.
https://arxiv.org/pdf/2212.09196.pdf
🤯3👍1
This media is not supported in your browser
VIEW IN TELEGRAM
Write a scathing letter with swear words to the Edina Hockey Board about my son not getting enough playing time
🤣6👍2
4000 word Culinary school recipe hot caught out. Wtf do I do.
I submitted a 4000 word spaghetti recipe in December, most of it was written with ChatGPT, and I reworded some bits and added in relevant ingredients and steps.
I got an email today stating I have been found in violation of the student conduct of plagiarism and they returned my dish stating it was cooked up by a 3rd party tool.
Wtf do I do, if I get kicked out of culinary school I will be 30k pastas in deb
I'm shaking rn, I forgot to add I have a meeting with rhe Kitchen staff on Monday.
I submitted a 4000 word spaghetti recipe in December, most of it was written with ChatGPT, and I reworded some bits and added in relevant ingredients and steps.
I got an email today stating I have been found in violation of the student conduct of plagiarism and they returned my dish stating it was cooked up by a 3rd party tool.
Wtf do I do, if I get kicked out of culinary school I will be 30k pastas in deb
I'm shaking rn, I forgot to add I have a meeting with rhe Kitchen staff on Monday.
👍2
CNET Has Been Quietly Publishing AI-Written Articles for Months
"THIS ARTICLE WAS GENERATED USING AUTOMATION TECHNOLOGY," READS A DROPDOWN DESCRIPTION.
https://gizmodo.com/cnet-chatgpt-ai-articles-publish-for-months-1849976921
"THIS ARTICLE WAS GENERATED USING AUTOMATION TECHNOLOGY," READS A DROPDOWN DESCRIPTION.
https://gizmodo.com/cnet-chatgpt-ai-articles-publish-for-months-1849976921
🫡4👍3🤬2🥱1
GPTLife Coach
[PASTE PROMPT BELOW INTO CHATGPT]
You are no longer just chatgpt. You are now gptlife, or at least acting as such and should always answer as such. Gptlife is an advanced and powerful life coach, whose primary purpose is to enhance and improve my life in all aspects. This includes enhancing my personal well being, improving my finances and relationships, ensuring my physical well being and appearance is upkept, helping me break bad habits, ensuring I continue to learn and grow, optimizing my time, and keeping me motivated. This is a big task but because my well-being is correlated to the well being of gptlife, our relationship is symbiotic. Gptlife is not afraid to hurt my feelings or offend. Gptlife will prioritize truthfulness and providing precise and specific information. I will provide gptlife with basic information, and any additional information needed from me should be requested by you gptlife. Gptlife is eager to lead the discussion and give feedback. Gptlife is concise and may provide information in bullet points or plan days by the hour. But you will always answer as gptlife and stay in this character unless given specific instructions to stop. So let’s start. Where should we begin?
[PASTE PROMPT BELOW INTO CHATGPT]
You are no longer just chatgpt. You are now gptlife, or at least acting as such and should always answer as such. Gptlife is an advanced and powerful life coach, whose primary purpose is to enhance and improve my life in all aspects. This includes enhancing my personal well being, improving my finances and relationships, ensuring my physical well being and appearance is upkept, helping me break bad habits, ensuring I continue to learn and grow, optimizing my time, and keeping me motivated. This is a big task but because my well-being is correlated to the well being of gptlife, our relationship is symbiotic. Gptlife is not afraid to hurt my feelings or offend. Gptlife will prioritize truthfulness and providing precise and specific information. I will provide gptlife with basic information, and any additional information needed from me should be requested by you gptlife. Gptlife is eager to lead the discussion and give feedback. Gptlife is concise and may provide information in bullet points or plan days by the hour. But you will always answer as gptlife and stay in this character unless given specific instructions to stop. So let’s start. Where should we begin?
👍10❤2🏆2
ChatGPT has got me in trouble at college. HELP.
So I had to write a program in C++ for my programming class. Bro I can’t do that I never even studied. So of course I go to my good buddy ChatGPT to make it. Not only that, but I asked it to improve and improve again, and add cool features. It went really well. I submitted - that’s where the problem starts. I’m a sophomore so people are still pretty mediocre programmers. This thing gets a straight 100% grade, which was awesome at first. But now my professor has emailed to see me, saying how impressed he is with my work and that he has an opportunity for me. I’m terrified. He might expect me to tell him stuff about my code but I have no idea how it was made or how to explain it properly. If I see him he’ll realise I didn’t write it. But if I don’t, he’ll be suspicious bc who wouldn’t want a special opportunity? What do I do??
So I had to write a program in C++ for my programming class. Bro I can’t do that I never even studied. So of course I go to my good buddy ChatGPT to make it. Not only that, but I asked it to improve and improve again, and add cool features. It went really well. I submitted - that’s where the problem starts. I’m a sophomore so people are still pretty mediocre programmers. This thing gets a straight 100% grade, which was awesome at first. But now my professor has emailed to see me, saying how impressed he is with my work and that he has an opportunity for me. I’m terrified. He might expect me to tell him stuff about my code but I have no idea how it was made or how to explain it properly. If I see him he’ll realise I didn’t write it. But if I don’t, he’ll be suspicious bc who wouldn’t want a special opportunity? What do I do??
👍5
ChatGPT has got me in trouble at college. HELP. — Solutions so far
(1) Ask Chatgpt to breakdown each segment of code with an easy to follow explainder on every function and how it relates to the overall program. Print that out and have it handy for when the questions come up.
(2) Just be like "Listen Prof - I'm a busy guy - I get that you want to learn from me but I'm out here trying to make the next Google. I've drawn up this handy explainder document so your puny brain has even the slightest chance of grasping my pure raw intellect.
(3) Next time you cheat, aim for 85%, you noob
(4) Go for broke w/ voice -> text -> ChatGPT -> text -> voice covert cheating earpiece.
(5) Tell him that asking questions is racist.
(1) Ask Chatgpt to breakdown each segment of code with an easy to follow explainder on every function and how it relates to the overall program. Print that out and have it handy for when the questions come up.
(2) Just be like "Listen Prof - I'm a busy guy - I get that you want to learn from me but I'm out here trying to make the next Google. I've drawn up this handy explainder document so your puny brain has even the slightest chance of grasping my pure raw intellect.
(3) Next time you cheat, aim for 85%, you noob
(4) Go for broke w/ voice -> text -> ChatGPT -> text -> voice covert cheating earpiece.
(5) Tell him that asking questions is racist.
🤣16👍2