Alex Choi
@aschoi.bsky.social
student @georgemasoncs.bsky.social. NLP with GMNLP. optimistic and curious.
I mean, isn't it possible that LLMs could just be better at reading scientific abstracts? Not necessarily better at predicting neuroscience results?
November 29, 2024 at 9:45 PM
I mean, isn't it possible that LLMs could just be better at reading scientific abstracts? Not necessarily better at predicting neuroscience results?
And as an extension to this, I've heard anecdotally that debugging "questionable" AI generated code can take just as long as writing it from scratch.
November 27, 2024 at 8:10 PM
And as an extension to this, I've heard anecdotally that debugging "questionable" AI generated code can take just as long as writing it from scratch.
And in this report, its buried a bit, but in the section "towards effective AI support for developers" there's mention of concern amongst developers about the potential vulnerabilities using AI to generate code could introduce.
www.microsoft.com/en-us/resear...
www.microsoft.com/en-us/resear...
Generative AI in Real-World Workplaces - Microsoft Research
This report presents the most recent findings of Microsoft’s research initiative on AI and Productivity, which seeks to measure and understand the productivity gains associated with LLM-powered produc...
www.microsoft.com
November 27, 2024 at 8:10 PM
And in this report, its buried a bit, but in the section "towards effective AI support for developers" there's mention of concern amongst developers about the potential vulnerabilities using AI to generate code could introduce.
www.microsoft.com/en-us/resear...
www.microsoft.com/en-us/resear...
First, I just wanted to say thank you for taking the time to respond.
So, here is one paper that I would say defends the notion of "stochastic parrot"
x.com/RTomMcCoy/st...
So, here is one paper that I would say defends the notion of "stochastic parrot"
x.com/RTomMcCoy/st...
x.com
x.com
November 27, 2024 at 8:06 PM
First, I just wanted to say thank you for taking the time to respond.
So, here is one paper that I would say defends the notion of "stochastic parrot"
x.com/RTomMcCoy/st...
So, here is one paper that I would say defends the notion of "stochastic parrot"
x.com/RTomMcCoy/st...
You're definitely more well connected than someone in my position, but the assumption of "of course they can" isnt quite the shared and dominant sense that I've been getting.
November 27, 2024 at 4:58 PM
You're definitely more well connected than someone in my position, but the assumption of "of course they can" isnt quite the shared and dominant sense that I've been getting.
Either one works! Given my current profile, it probably makes sense to be in the responsible AI starter pack. (I'm working with security folks right now so I guess that's just been more at the forefront of my thoughts.)
November 24, 2024 at 6:31 PM
Either one works! Given my current profile, it probably makes sense to be in the responsible AI starter pack. (I'm working with security folks right now so I guess that's just been more at the forefront of my thoughts.)
Hi Stephen! Would I be able to join this 2nd privacy and security starter pack? Thank you!
November 24, 2024 at 4:40 AM
Hi Stephen! Would I be able to join this 2nd privacy and security starter pack? Thank you!
Hi! Could I join in on this starter pack? Thank you!
November 24, 2024 at 4:27 AM
Hi! Could I join in on this starter pack? Thank you!
Could I be included here as well? Thank you!
November 24, 2024 at 4:23 AM
Could I be included here as well? Thank you!
thru the framework of Topic Modeling w/ and w/out LLM assistance.
Our findings indicate positive efficiency gains but also a worrisome anchoring bias that colors the final analysis.
We hope that our study can add meaningful evidence to this ongoing global dialogue. Read the paper for details!
3/3
Our findings indicate positive efficiency gains but also a worrisome anchoring bias that colors the final analysis.
We hope that our study can add meaningful evidence to this ongoing global dialogue. Read the paper for details!
3/3
November 19, 2024 at 4:01 AM
thru the framework of Topic Modeling w/ and w/out LLM assistance.
Our findings indicate positive efficiency gains but also a worrisome anchoring bias that colors the final analysis.
We hope that our study can add meaningful evidence to this ongoing global dialogue. Read the paper for details!
3/3
Our findings indicate positive efficiency gains but also a worrisome anchoring bias that colors the final analysis.
We hope that our study can add meaningful evidence to this ongoing global dialogue. Read the paper for details!
3/3
our paper examines the integration of LLMs into specialized expert workflows, and explores both the potential benefits & worrisome pitfalls of this emerging partnership.
We asked International Policy experts to analyze and annotate lengthy AI Policy interviews,
2/3
We asked International Policy experts to analyze and annotate lengthy AI Policy interviews,
2/3
November 19, 2024 at 4:01 AM
our paper examines the integration of LLMs into specialized expert workflows, and explores both the potential benefits & worrisome pitfalls of this emerging partnership.
We asked International Policy experts to analyze and annotate lengthy AI Policy interviews,
2/3
We asked International Policy experts to analyze and annotate lengthy AI Policy interviews,
2/3
Hi, I'm at George Mason University and published to emnlp 2024 main with a paper related to computational social science. Can you add me to this starter pack as well? Thank you!
November 18, 2024 at 3:23 PM
Hi, I'm at George Mason University and published to emnlp 2024 main with a paper related to computational social science. Can you add me to this starter pack as well? Thank you!