Sign up to save your podcasts
Or
Share In-Ear Insights: Responsible AI Part 2, Managing Bias
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
In-Ear Insights: Responsible AI Part 2, Managing Bias
In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris tackle the issue of bias in AI, particularly managing bias in large language models. Discover real-life examples of how bias manifests and the potential consequences for businesses, including reputational damage and the reinforcement of harmful stereotypes. You will learn about a critical framework for responsible AI development and deployment, emphasizing the importance of accountability, fairness, and transparency. Tune in to gain actionable strategies for mitigating bias in AI and promoting ethical practices within your organization.
Watch the video here:
Can’t see anything? Watch it on YouTube here.
Listen to the audio here:
https://traffic.libsyn.com/inearinsights/tipodcast-responsible-ai-part-2-bias.mp3
Download the MP3 audio here.
[podcastsponsor]
Machine-Generated Transcript
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.
Christopher S. Penn – 00:00
In this week’s In-Ear Insights, this is part two of our series on responsible AI. This time, we’re going to talk about—particularly one of the biggest problems of generative AI, one of the biggest risks for our companies—is bias. Bias in the way AI systems and models work, bias in how they produce outputs. And, Katie, actually, just very recently, you were working on some content with Kelsey for our website using the generative models, and it spit out some stuff that was fairly sexist, at least from what I heard. So, do you want to talk about what happened there?
Katie Robbert – 00:39
Yeah. It’s no secret that we use large language models to assist with our writing. We’re a small company, but we’ve done a lot of work to create custom large language models based on our previous writing, based on our data. So we’re not just trying to pull something generic from any old open model. What happened, though? Because it was supposed to be writing as me, Katie Robbert, with a feminine name, the language model recognized Katie as a woman. And so, one of the things that I—Katie, in real life—do in my writing is I give anecdotes and examples to really help the content resonate with people.
Katie Robbert – 01:27
And so what the large language model did was it made decisions on its own about including anecdotes and examples, and everything it included was about fashion, shoe shopping, closets, going to the mall, keeping up on the latest clothing trends, which anyone who knows me knows that I basically dress one step above homeless. And so those are not things that are important to me, nor are they examples that I have ever given in my writing. I’ve not talked about cleaning out my closet to fit in the latest designer fashions.
It’s kind of disappointing that the large language model—despite having probably years’ worth of examples of my writing with no examples of those specific anecdotes—made those choices and assumptions about me because I have a feminine name, that those would be things
View all episodes
By Trust Insights
5
99 ratings
In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris tackle the issue of bias in AI, particularly managing bias in large language models. Discover real-life examples of how bias manifests and the potential consequences for businesses, including reputational damage and the reinforcement of harmful stereotypes. You will learn about a critical framework for responsible AI development and deployment, emphasizing the importance of accountability, fairness, and transparency. Tune in to gain actionable strategies for mitigating bias in AI and promoting ethical practices within your organization.
Watch the video here:
Can’t see anything? Watch it on YouTube here.
Listen to the audio here:
https://traffic.libsyn.com/inearinsights/tipodcast-responsible-ai-part-2-bias.mp3
Download the MP3 audio here.
[podcastsponsor]
Machine-Generated Transcript
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.
Christopher S. Penn – 00:00
In this week’s In-Ear Insights, this is part two of our series on responsible AI. This time, we’re going to talk about—particularly one of the biggest problems of generative AI, one of the biggest risks for our companies—is bias. Bias in the way AI systems and models work, bias in how they produce outputs. And, Katie, actually, just very recently, you were working on some content with Kelsey for our website using the generative models, and it spit out some stuff that was fairly sexist, at least from what I heard. So, do you want to talk about what happened there?
Katie Robbert – 00:39
Yeah. It’s no secret that we use large language models to assist with our writing. We’re a small company, but we’ve done a lot of work to create custom large language models based on our previous writing, based on our data. So we’re not just trying to pull something generic from any old open model. What happened, though? Because it was supposed to be writing as me, Katie Robbert, with a feminine name, the language model recognized Katie as a woman. And so, one of the things that I—Katie, in real life—do in my writing is I give anecdotes and examples to really help the content resonate with people.
Katie Robbert – 01:27
And so what the large language model did was it made decisions on its own about including anecdotes and examples, and everything it included was about fashion, shoe shopping, closets, going to the mall, keeping up on the latest clothing trends, which anyone who knows me knows that I basically dress one step above homeless. And so those are not things that are important to me, nor are they examples that I have ever given in my writing. I’ve not talked about cleaning out my closet to fit in the latest designer fashions.
It’s kind of disappointing that the large language model—despite having probably years’ worth of examples of my writing with no examples of those specific anecdotes—made those choices and assumptions about me because I have a feminine name, that those would be things
More shows like In-Ear Insights from Trust Insights
View all
The Artificial Intelligence Show
171 Listeners
AI Security Podcast
4 Listeners