AI Workers: Warning Their Friends and Family to Stay Away
A growing number of workers who moderate AI-generated content are sounding the alarm about the potential risks of relying on these systems. These workers, often referred to as "AI raters," spend their days evaluating the accuracy and quality of AI-generated text, images, and videos.
Their experience has left them with a deep-seated distrust of the models they work on. Many have become disillusioned with the emphasis on rapid turnaround times over quality, which they believe compromises the integrity of the output.
One worker, Krista Pawloski, recalls a moment when she was tasked with labeling tweets as racist or not. She came across a tweet that used a racial slur, and after rechecking the meaning of the word, she realized she had almost clicked on the "no" button earlier. This experience led her to question how many others might have made similar mistakes and let offensive material slip by.
Pawloski's encounter with AI-generated content has made her cautious about using it herself. She warns her family and friends not to use generative AI tools, advising them to ask AI questions that they are knowledgeable about to spot its errors. For Pawloski, the potential for harm is too great, and she believes it's essential to exercise caution when interacting with these systems.
Similar concerns have been expressed by other workers in the industry. An Amazon spokesperson said that workers can choose which tasks to complete at their discretion and review a task's details before accepting it. However, many AI raters argue that this is not enough to ensure the quality of the output.
The problem runs deeper than just individual instances of error. As experts point out, when people who don't understand AI are enchanted by its capabilities, they may overlook its limitations. This lack of critical thinking can lead to the acceptance and propagation of misinformation.
Brook Hansen, an AI worker on Amazon Mechanical Turk, notes that companies prioritize speed and profit over responsibility and quality. "If workers aren't equipped with the information, resources, and time we need, how can the outcomes possibly be safe, accurate or ethical?" she asks.
The consequences of these issues are far-reaching. An audit by NewsGuard found that the top 10 generative AI models repeat false information almost twice as often as they did in August 2024. This highlights a critical flaw: if you feed bad data into an AI system, it will likely produce flawed results.
AI workers are taking matters into their own hands, educating their loved ones about the potential risks of generative AI and encouraging others to ask questions. As one worker put it, "We joke that [chatbots] would be great if we could get them to stop lying."
However, some experts caution against a doom-and-gloom approach to AI. They argue that by recognizing the limitations and flaws of these systems, we can begin to address the underlying issues.
As one expert noted, "AI is only as good as what's put into it, and what's put into it is not always the best information." By acknowledging this reality and encouraging critical thinking, we may be able to create a more nuanced understanding of AI and its potential risks.
Ultimately, the choice between embracing or cautioning around generative AI depends on our individual values and priorities. As one worker aptly put it, "Once you've seen how these systems are cobbled together – the biases, the rushed timelines, the constant compromises – you stop seeing AI as futuristic and start seeing it as fragile."
Perhaps by recognizing this fragility, we can work towards creating more responsible and transparent AI systems that prioritize quality over speed.
A growing number of workers who moderate AI-generated content are sounding the alarm about the potential risks of relying on these systems. These workers, often referred to as "AI raters," spend their days evaluating the accuracy and quality of AI-generated text, images, and videos.
Their experience has left them with a deep-seated distrust of the models they work on. Many have become disillusioned with the emphasis on rapid turnaround times over quality, which they believe compromises the integrity of the output.
One worker, Krista Pawloski, recalls a moment when she was tasked with labeling tweets as racist or not. She came across a tweet that used a racial slur, and after rechecking the meaning of the word, she realized she had almost clicked on the "no" button earlier. This experience led her to question how many others might have made similar mistakes and let offensive material slip by.
Pawloski's encounter with AI-generated content has made her cautious about using it herself. She warns her family and friends not to use generative AI tools, advising them to ask AI questions that they are knowledgeable about to spot its errors. For Pawloski, the potential for harm is too great, and she believes it's essential to exercise caution when interacting with these systems.
Similar concerns have been expressed by other workers in the industry. An Amazon spokesperson said that workers can choose which tasks to complete at their discretion and review a task's details before accepting it. However, many AI raters argue that this is not enough to ensure the quality of the output.
The problem runs deeper than just individual instances of error. As experts point out, when people who don't understand AI are enchanted by its capabilities, they may overlook its limitations. This lack of critical thinking can lead to the acceptance and propagation of misinformation.
Brook Hansen, an AI worker on Amazon Mechanical Turk, notes that companies prioritize speed and profit over responsibility and quality. "If workers aren't equipped with the information, resources, and time we need, how can the outcomes possibly be safe, accurate or ethical?" she asks.
The consequences of these issues are far-reaching. An audit by NewsGuard found that the top 10 generative AI models repeat false information almost twice as often as they did in August 2024. This highlights a critical flaw: if you feed bad data into an AI system, it will likely produce flawed results.
AI workers are taking matters into their own hands, educating their loved ones about the potential risks of generative AI and encouraging others to ask questions. As one worker put it, "We joke that [chatbots] would be great if we could get them to stop lying."
However, some experts caution against a doom-and-gloom approach to AI. They argue that by recognizing the limitations and flaws of these systems, we can begin to address the underlying issues.
As one expert noted, "AI is only as good as what's put into it, and what's put into it is not always the best information." By acknowledging this reality and encouraging critical thinking, we may be able to create a more nuanced understanding of AI and its potential risks.
Ultimately, the choice between embracing or cautioning around generative AI depends on our individual values and priorities. As one worker aptly put it, "Once you've seen how these systems are cobbled together – the biases, the rushed timelines, the constant compromises – you stop seeing AI as futuristic and start seeing it as fragile."
Perhaps by recognizing this fragility, we can work towards creating more responsible and transparent AI systems that prioritize quality over speed.
AI is like a big ol' mirror, reflecting back all the flaws we put into it 
. We need to be super cautious when using these tools, especially if we're not tech-savvy. I mean, who's going to fact-check for us? 
Companies are trying to rush out quality stuff, but that's just gonna lead to more mistakes 
. We gotta ask questions and think critically about what we're putting into these systems 
. And let's be real, if our family and friends can't trust AI, how are they supposed to navigate the wild west of online info? 
AI is only as good as what's put into it, but isn't that true for any system? like, if you use a calculator that's gonna give you the wrong answer, is it really the calculator's fault or the person using it? 
sometimes i feel like we're just scratching the surface of what these systems can do (or not do)
but at the same time, shouldn't companies be doing more to ensure the quality of their products? like, isn't it their responsibility to make sure what they're selling is safe and accurate? 
. People are relying too much on them without questioning what's going on behind the scenes. It's like they're playing with fire without even reading the warning signs 
.
. I think there's some potential for good here if we can just get it right. But let's not pretend that everything is sunshine and rainbows 
. We need to have an honest conversation about the risks and limitations of these tools and work towards creating something better in the future 
. We need to educate ourselves and each other about what we're getting into here 
.
these workers r doin their part warnin ppl but like what can we really do? 
. Anyway, these AI workers are totally right to be cautious - we don't know what kind of crazy stuff we're gonna feed into these systems and expect sane results 
. And can you blame them for wanting to warn their friends and family? Like, who wants to be the one responsible for spreading misinformation just because a bot told 'em it's true 
.
, not the other way around 
. Workers who rate AI content are sounding the alarm because they know what's up with these models 
. We need to think critically about AI like we do with science experiments 
, not just go along with the hype 
. Can't trust them yet 
. My aunt is super into chatbots and I have to explain it to her that they're not perfect 
.
, but when it comes to sensitive stuff like human emotions 
, I'm like...cautious 
. Can't have AI making decisions that affect people's lives 
.
. It's not just about embracing the future, it's about being responsible 
. We need to keep pushing for better systems that prioritize people over profit 
.
, right? So why are companies still pushing the boundaries of what's possible with these models without prioritizing quality control? 
. I mean, shouldn't we be prioritizing accuracy and ethics over being able to spit out content fast? 
The fact that these models repeat false info almost twice as often as they did last year is just wild 
. I mean, think about it - they're trained on all sorts of data, including some super sketchy stuff. It's like trying to predict the weather by just throwing a bunch of different kinds of trash into a blender 
. If we're not careful, we could end up spreading misinformation and harming people in the process.
. That's crazy talk! I'm surprised more people aren't speaking up about this.
. We need to have real conversations about AI and its limitations, not just pretend everything is fine 
.