Artificial intelligence built on mountains of potentially confusing information has created a real risk of automated discrimination, but is there a way to teach those machines too?
This question is about some really quick stuff. During ChatGPT, AI will make more decisions for healthcare providers, banks or lawyers, using anything scraped from the Internet as a source.
So the intelligence of AI is only as good as the world from which it emerges, as it is likely to be full of wisdom, wisdom and value, as well as hatred, prejudice and diatribes.
“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. Ensuring that technology accurately reflects human diversity is not just a political choice.
Other uses of AI, such as facial recognition, have seen companies find themselves in hot water with authorities for discrimination. This is the case against Rite-Aid, a drug store in the United States, where in-store cameras falsely identify shoppers, especially women and people of color, as shoplifters. stores, according to the Federal Trade Commission.
“I made a mistake”
ChatGPT’s generative AI model, which can create a human-level opinion in a few seconds, opens up new opportunities to make mistakes, experts are worried.
AI giants are well aware of the problem, fearing that their models will lead to bad behavior or reflect the extremes of Western society when their users are the world.
“We have people asking us from Indonesia or the United States,” Google CEO Sundar Pichai said, explaining why requests for images from doctors or lawyers would be hard to prove. different ethnic groups. But these considerations can reach absurd levels and lead to virulent accusations of excessive political correctness.
That’s what happened when Google’s Gemini image generator spat out an image of German soldiers from World War II that absurdly included a black man and an Asian woman. “It’s clear that the offense is overusing … where it shouldn’t have been applied. It’s a bug and we’re wrong,” Pichai said. Both sashi luccioni, of the scientificness of the scientific breast and “apologize that there is a technical process doing wrong.”
If the produce is “in harmony with the employee who wants something to” but he said.
Jayden Pogger, a non-unpleasant product manager or any negative or any negative or nothing. For now at least, it is up to humans to ensure that AI produces anything that fits or meets their expectations.
“Cooked” misery
But given the frenzy around AI, this is no easy task. Hug Face has about 600,000 AI or machine learning models on its platform.
Luccioni said, “In a few weeks, a new model emerges and we have a kind of snobbery so that we can only try to analyze and record negative thoughts or negative behavior.” One method that is being developed is called algorithmic disgorging that will allow engineers to remove the content, without destroying the entire model.
But there are serious doubts that it can really work. Another way would be to “encourage” the model going in the right direction, its “beauty”, “rewarding the good and the bad,” said Ram Sriharsha, CTO at Pinecone.
Pinecone specializes in integrated generation (or RAG), a process where the model generates information from a reliable and standardized source. For Weaver of the Texas Opportunity & Justice Incubator, these “beautiful” efforts to correct negative stereotypes are “the outline of our hopes and dreams of a better future.”
But bias “is also what it means to be human and, as such, it’s also built into AI,” he said.
