Can we rid artificial intelligence of bias? – Tech Xplore
Click here to sign in with or
Forget Password?
Learn more
share this!
62
Twit
Share
Email
May 19, 2024
This article has been reviewed according to Science X’s editorial process and policies. Editors have highlighted the following attributes while ensuring the content’s credibility:
fact-checked
reputable news agency
proofread
by Julie JAMMOT with Alex PIGMAN in Washington
Artificial intelligence built on mountains of potentially biased information has created a real risk of automating discrimination, but is there any way to re-educate the machines?
The question for some is extremely urgent. In this ChatGPT era, AI will generate more and more decisions for health care providers, bank lenders or lawyers, using whatever was scoured from the internet as source material.
AI’s underlying intelligence, therefore, is only as good as the world it came from, as likely to be filled with wit, wisdom, and usefulness, as well as hatred, prejudice and rants.
“It’s dangerous because people are embracing and adopting AI software and really depending on it,” said Joshua Weaver, Director of Texas Opportunity & Justice Incubator, a legal consultancy.
“We can get into this feedback loop where the bias in our own selves and culture informs bias in the AI and becomes a sort of reinforcing loop,” he said.
Making sure technology more accurately reflects human diversity is not just a political choice.
Other uses of AI, like facial recognition, have seen companies thrown into hot water with authorities for discrimination.
This was the case against Rite-Aid, a US pharmacy chain, where in-store cameras falsely tagged consumers, particularly women and people of color, as shoplifters, according to the Federal Trade Commission.
ChatGPT-style generative AI, which can create a semblance of human-level reasoning in just seconds, opens up new opportunities to get things wrong, experts worry.
The AI giants are well aware of the problem, afraid that their models can descend into bad behavior, or overly reflect a western society when their user base is global.
“We have people asking queries from Indonesia or the US,” said Google CEO Sundar Pichai, explaining why requests for images of doctors or lawyers will strive to reflect racial diversity.
But these considerations can reach absurd levels and lead to angry accusations of excessive political correctness.
This is what happened when Google’s Gemini image generator spat out an image of German soldiers from World War Two that absurdly included a black man and Asian woman.
“Obviously, the mistake was that we over-applied… where it should have never applied. That was a bug and we got it wrong,” Pichai said.
But Sasha Luccioni, a research scientist at Hugging Face, a leading platform for AI models cautioned that “thinking that there’s a technological solution to bias is kind of already going down the wrong path.”
Generative AI is essentially about whether the output “corresponds to what the user expects it to” and that is largely subjective, she said.
The huge models on which ChatGPT is built “can’t reason about what is biased or what isn’t so they can’t do anything about it,” cautioned Jayden Ziegler, head of product at Alembic Technologies.
For now at least, it is up to humans to ensure that the AI generates whatever is appropriate or meets their expectations.
But given the frenzy around AI, that is no easy task.
Hugging Face has about 600,000 AI or machine learning models available on its platform.
“Every couple of weeks a new model comes out and we’re kind of scrambling in order to try to just evaluate and document biases or undesirable behaviors,” said Luccioni.
One method under development is something called algorithmic disgorgement that would allow engineers to excise content, without ruining the whole model.
But there are serious doubts this can actually work.
Another method would “encourage” a model to go in the right direction, “fine tune” it, “rewarding for right and wrong,” said Ram Sriharsha, chief technology officer at Pinecone.
Pinecone is a specialist of retrieval augmented generation (or RAG), a technique where the model fetches information from a fixed trusted source.
For Weaver of the Texas Opportunity & Justice Incubator, these “noble” attempts to fix bias are “projections of our hopes and dreams for what a better version of the future can look like.”
But bias “is also inherent into what it means to be human and because of that, it’s also baked into the AI as well,” he said.
© 2024 AFP
Explore further
Facebook
Twitter
Email
Feedback to editors
12 hours ago
0
May 23, 2024
0
May 22, 2024
0
May 22, 2024
0
May 21, 2024
0
8 hours ago
9 hours ago
12 hours ago
14 hours ago
May 23, 2024
May 23, 2024
May 23, 2024
May 23, 2024
May 23, 2024
May 23, 2024
Feb 28, 2024
May 13, 2024
Mar 17, 2024
May 13, 2024
Mar 22, 2024
Mar 5, 2024
May 23, 2024
May 23, 2024
May 22, 2024
May 22, 2024
May 22, 2024
May 21, 2024
Use this form if you have come across a typo, inaccuracy or would like to send an edit request for the content on this page. For general inquiries, please use our contact form. For general feedback, use the public comments section below (please adhere to guidelines).
Please select the most appropriate category to facilitate processing of your request
Thank you for taking time to provide your feedback to the editors.
Your feedback is important to us. However, we do not guarantee individual replies due to the high volume of messages.
Your email address is used only to let the recipient know who sent the email. Neither your address nor the recipient’s address will be used for any other purpose. The information you enter will appear in your e-mail message and is not retained by Tech Xplore in any form.
Daily science news on research developments and the latest scientific innovations
Medical research advances and health news
The most comprehensive sci-tech news coverage on the web
This site uses cookies to assist with navigation, analyse your use of our services, collect data for ads personalisation and provide content from third parties. By using our site, you acknowledge that you have read and understand our Privacy Policy and Terms of Use.
This article was autogenerated from a news feed from CDO TIMES selected high quality news and research sources. There was no editorial review conducted beyond that by CDO TIMES staff. Need help with any of the topics in our articles? Schedule your free CDO TIMES Tech Navigator call today to stay ahead of the curve and gain insider advantages to propel your business!
Forget Password?
Learn more
share this!
62
Twit
Share
May 19, 2024
This article has been reviewed according to Science X’s editorial process and policies. Editors have highlighted the following attributes while ensuring the content’s credibility:
fact-checked
reputable news agency
proofread
by Julie JAMMOT with Alex PIGMAN in Washington
Artificial intelligence built on mountains of potentially biased information has created a real risk of automating discrimination, but is there any way to re-educate the machines?
The question for some is extremely urgent. In this ChatGPT era, AI will generate more and more decisions for health care providers, bank lenders or lawyers, using whatever was scoured from the internet as source material.
AI’s underlying intelligence, therefore, is only as good as the world it came from, as likely to be filled with wit, wisdom, and usefulness, as well as hatred, prejudice and rants.
“It’s dangerous because people are embracing and adopting AI software and really depending on it,” said Joshua Weaver, Director of Texas Opportunity & Justice Incubator, a legal consultancy.
“We can get into this feedback loop where the bias in our own selves and culture informs bias in the AI and becomes a sort of reinforcing loop,” he said.
Making sure technology more accurately reflects human diversity is not just a political choice.
Other uses of AI, like facial recognition, have seen companies thrown into hot water with authorities for discrimination.
This was the case against Rite-Aid, a US pharmacy chain, where in-store cameras falsely tagged consumers, particularly women and people of color, as shoplifters, according to the Federal Trade Commission.
ChatGPT-style generative AI, which can create a semblance of human-level reasoning in just seconds, opens up new opportunities to get things wrong, experts worry.
The AI giants are well aware of the problem, afraid that their models can descend into bad behavior, or overly reflect a western society when their user base is global.
“We have people asking queries from Indonesia or the US,” said Google CEO Sundar Pichai, explaining why requests for images of doctors or lawyers will strive to reflect racial diversity.
But these considerations can reach absurd levels and lead to angry accusations of excessive political correctness.
This is what happened when Google’s Gemini image generator spat out an image of German soldiers from World War Two that absurdly included a black man and Asian woman.
“Obviously, the mistake was that we over-applied… where it should have never applied. That was a bug and we got it wrong,” Pichai said.
But Sasha Luccioni, a research scientist at Hugging Face, a leading platform for AI models cautioned that “thinking that there’s a technological solution to bias is kind of already going down the wrong path.”
Generative AI is essentially about whether the output “corresponds to what the user expects it to” and that is largely subjective, she said.
The huge models on which ChatGPT is built “can’t reason about what is biased or what isn’t so they can’t do anything about it,” cautioned Jayden Ziegler, head of product at Alembic Technologies.
For now at least, it is up to humans to ensure that the AI generates whatever is appropriate or meets their expectations.
But given the frenzy around AI, that is no easy task.
Hugging Face has about 600,000 AI or machine learning models available on its platform.
“Every couple of weeks a new model comes out and we’re kind of scrambling in order to try to just evaluate and document biases or undesirable behaviors,” said Luccioni.
One method under development is something called algorithmic disgorgement that would allow engineers to excise content, without ruining the whole model.
But there are serious doubts this can actually work.
Another method would “encourage” a model to go in the right direction, “fine tune” it, “rewarding for right and wrong,” said Ram Sriharsha, chief technology officer at Pinecone.
Pinecone is a specialist of retrieval augmented generation (or RAG), a technique where the model fetches information from a fixed trusted source.
For Weaver of the Texas Opportunity & Justice Incubator, these “noble” attempts to fix bias are “projections of our hopes and dreams for what a better version of the future can look like.”
But bias “is also inherent into what it means to be human and because of that, it’s also baked into the AI as well,” he said.
© 2024 AFP
Explore further
Feedback to editors
12 hours ago
0
May 23, 2024
0
May 22, 2024
0
May 22, 2024
0
May 21, 2024
0
8 hours ago
9 hours ago
12 hours ago
14 hours ago
May 23, 2024
May 23, 2024
May 23, 2024
May 23, 2024
May 23, 2024
May 23, 2024
Feb 28, 2024
May 13, 2024
Mar 17, 2024
May 13, 2024
Mar 22, 2024
Mar 5, 2024
May 23, 2024
May 23, 2024
May 22, 2024
May 22, 2024
May 22, 2024
May 21, 2024
Use this form if you have come across a typo, inaccuracy or would like to send an edit request for the content on this page. For general inquiries, please use our contact form. For general feedback, use the public comments section below (please adhere to guidelines).
Please select the most appropriate category to facilitate processing of your request
Thank you for taking time to provide your feedback to the editors.
Your feedback is important to us. However, we do not guarantee individual replies due to the high volume of messages.
Your email address is used only to let the recipient know who sent the email. Neither your address nor the recipient’s address will be used for any other purpose. The information you enter will appear in your e-mail message and is not retained by Tech Xplore in any form.
Daily science news on research developments and the latest scientific innovations
Medical research advances and health news
The most comprehensive sci-tech news coverage on the web
This site uses cookies to assist with navigation, analyse your use of our services, collect data for ads personalisation and provide content from third parties. By using our site, you acknowledge that you have read and understand our Privacy Policy and Terms of Use.
This article was autogenerated from a news feed from CDO TIMES selected high quality news and research sources. There was no editorial review conducted beyond that by CDO TIMES staff. Need help with any of the topics in our articles? Schedule your free CDO TIMES Tech Navigator call today to stay ahead of the curve and gain insider advantages to propel your business!
