- 7 weeks ago
When artificial intelligence goes wrong, it goes SPECTACULARLY wrong! Join us as we explore the most controversial AI failures that sparked widespread outrage and heated debate. From racist chatbots to fake audience members, these technological blunders remind us that even the smartest systems can make the dumbest mistakes.
Category
🗞
NewsTranscript
00:00I don't think most parents know the capability of this tool.
00:03That tool, ChatGPT.
00:05Welcome to WatchMojo, and today we're counting down our picks for the moments when artificial intelligence,
00:10in all its supposed glory, stumbled spectacularly, sparking outrage, alarm, and a healthy dose of I told you so.
00:18It's a song that's been blowing up on streaming services, but if you've been listening to it,
00:22you have been listening to a 100% AI-generated ban.
00:2610. Will Smith's AI Audience
00:30The internet lit up with a peculiar controversy surrounding actor Will Smith's based on a true story tour.
00:37After the actor shared a highlight reel, fans and critics quickly noticed uncanny glitches in the crowd footage.
00:43Distorted faces, six-fingered hands, and signs that morphed mid-frame sparked immediate accusations that the star was using AI to generate fake audience enthusiasm.
00:53We asked AI expert Hani Farid for a second opinion.
00:57So there's no doubt that there is real footage in the video.
01:01There's also no doubt that there is some AI content, but even that has some nuance to it.
01:07The whole thing extra ironic because of the video's title.
01:10My favorite part of the tour is seeing you all up close.
01:14While Smith himself later responded with humor, posting a video of himself performing to a stadium of AI-generated cats,
01:21the incident ignited a serious discussion about transparency and authenticity in entertainment.
01:27Smith's painfully public scandal served as a cringeworthy reminder that even subtle AI enhancements can have a detrimental effect on even the most beloved entertainers' careers.
01:37We think what happened is they took a still image from a real scene and then animated it with AI for a few seconds.
01:45And so is it real or is it fake? A little bit of both.
01:48Number nine, Amazon's sexist hiring AI.
01:52Amazon is learning a tough lesson about artificial intelligence.
01:55The company has now abandoned an AI recruiting tool after discovering that the program was biased against women.
02:02Tech expert Ryan Eldridge of Nerds On Call joins me now.
02:06And Ryan, this new recruiting engine was supposed to identify candidates that Amazon might hire, but it appears this program chose far more men.
02:14In 2014, Amazon embarked on an ambitious project to revolutionize its recruitment process with an AI-powered tool designed to sift through resumes and identify top talent.
02:25However, this holy grail quickly revealed a significant flaw.
02:29It systematically discriminated against women.
02:31It was basing its decision on 10,000 different resumes, and it was based on what hiring practices Amazon was already doing.
02:40So there might have already been some gender bias already in the data that was being put in.
02:45But what happened was the system essentially started discounting anytime someone had graduated from a female college or a woman's college.
02:53The algorithm, trained on a decade of historical hiring data, which was predominantly male, learned to penalize resumes containing words like women's, as in women's chess club, and even downgraded graduates from all-female colleges.
03:07By 2018, the inherent bias was so apparent that Amazon had to scrap the project, unable to make the algorithm gender-neutral despite efforts.
03:15This incident raised serious ethical questions about fairness and discrimination in automated decision-making.
03:22This is something that's not going to go away.
03:24We're all going to start seeing this more and more.
03:25And it's a little bit frightening to think that there's a computer that's going to decide whether I get the job or not.
03:30Right, right.
03:31But places like Goldman Sachs and several other companies are already deeply invested in this technology to make their hiring and recruiting more effective.
03:39Number eight, GPT-5 fails to meet expectations.
03:43Open AI launches GPT-5.
03:46It's the latest generation of the artificial intelligence technology that powers chat GPT.
03:51Open AI says the new model is faster and smarter than ever before.
03:56It comes more than two years after the release of its predecessor, and expectations are high.
04:03The anticipated rollout of Open AI's powerful new large-language model generated a significant amount of grief.
04:10And not just for Sam Altman.
04:12Users expressed widespread frustration, reporting that the updated model felt less emotionally intelligent and personable than its predecessor, GPT-4-0.
04:21There were concerns about hallucinations, where the AI would confidently present incorrect facts or fabricate citations, despite claims of enhanced accuracy.
04:30Furthermore, early jailbreaking attempts successfully bypassed its safety filters, even reportedly leading to instructions for making explosives, contradicting the corporation's assurances of improved safety.
04:43This collective outcry underscored a broader social anxiety about the rapid, sometimes unpredictable, advancement of AI and the potential for unintended consequences.
04:53They're really saying you should take your life and give it over to this thing.
04:57This is a whole new world. We don't know if that's okay.
05:00Maybe that's a good thing. Maybe that will be helpful.
05:03But this is absolutely uncharted territory.
05:06And the fact that they want you to have as much faith in their product as they're asking you to have is a pretty bold step from a tech network.
05:13Number seven, Coca-Cola's 2024 holiday campaign falls flat.
05:17One of the most iconic brands, ringing in the holiday season.
05:22But take a closer look at the new Coca-Cola commercial, and you might notice that it was made with artificial intelligence.
05:30Social media certainly caught with.
05:31A brand synonymous with heartwarming holiday advertisements stirred a different kind of emotion with its AI-generated Christmas campaign.
05:39Aiming to pay homage to its classic 1995 Holidays Are Coming commercial, the company utilized its Real Magic AI platform to create an entirely artificial version.
05:49The result, however, was widely criticized for lacking creativity, emotional depth, and a crucial human touch.
05:56Why do you think companies are just toying around so much with AI in this moment?
06:00Well, it's a push for marketing efficiency, right?
06:03How do we create more with less?
06:05I mean, that's just, that's sort of, you know, business one-on-one.
06:08Behind the scenes, companies are ramping up AI use, too.
06:11Yum Brands, the parent company behind Taco Bell, Pizza Hut, and KFC, has reportedly started testing promotional emails written by AI.
06:19Viewers described the ad as soulless, unnatural, and creepy, with some critics, like Gravity Falls creator Alex Hirsch,
06:26humorously suggesting Coca-Cola was made from, quote, the blood of out-of-work artists.
06:32Coke's disastrous campaign became a prominent example of the growing tension between technological innovation and the inherent value of human creativity.
06:41Consumers are coming around to this, but again, brand experts that I spoke to have said,
06:44is this likely to change people's behavior in terms of what beverage they choose to drink?
06:49Probably not.
06:50So it's the human figures that are AI.
06:52Yeah.
06:53The polar bears were never real.
06:54The polar bears were never real.
06:55Yeah, they are very deadly.
06:57But yes, the humans, I think, is what people picked up on that ad.
06:59It's like, yeah, that just looks a little off.
07:01Number six, Google Gemini's poor and controversial grasp of history.
07:06It is a big deal that Google, which is now the paper of record for the Internet, the way the New York Times might have been one day in the past,
07:15the paper of record for the world, is creating fictitious and historically impossible images.
07:22The tech megacorps Gemini AI model faced immediate and intense backlash following reports of biased and historically inaccurate image generation.
07:30Users discovered that prompts for figures like the Founding Fathers or Nazi soldiers would often produce images featuring women and people of color,
07:39regardless of historical context.
07:41Over time, if left unchecked, this fictitious information that came from Gemini and comes from other models,
07:50you don't have to just pick on Google, will find its way into our knowledge sphere.
07:56Will find its way into schools, will find its way into curriculum and lessons, because teachers will incorporate what they find.
08:02This overcorrection in pursuit of diversity led to a torrent of public ridicule and accusations of wokeness and ideological bias.
08:11Google swiftly acknowledged the, quote, embarrassing and wrong outputs, temporarily halting the image generation feature to address the underlying issues.
08:19The controversy drew attention to the difficulty of preventing unforeseen biases in advanced models and just what can be done to combat them.
08:27It's the lens that these models are given or that's applied to these lenses by the political officers who operate and work and control large bureaucracies that these companies insist on applying.
08:40Number five, the dubious rise of the Velvet Sundown.
08:43A new album by the band Velvet Sundown, released today.
08:49They're a 70s rock-flavoured band blowing up with over a million streams online.
08:55Catch is, they're not real, are they?
08:57They're fake.
08:59The rise of AI-generated music hit a sour note with many when the Velvet Sundown gained significant traction on Spotify,
09:06amassing over a million monthly listeners in weeks.
09:09The catch? Every aspect of the act, from the music and lyrics to promotional images and backstory,
09:16was entirely created by artificial intelligence.
09:19So to get this straight, then somebody pretending to represent them exposed the band as synthetic.
09:24And that turned out to be a hoax, but whoever is behind it, it's still AI.
09:28Yes, I know it is confusing, but someone else exposing the band as AI seemed to be what really pushed the real fake band,
09:36the real fake creators of the Velvet Sundown to come clean.
09:39While the music was polished, the revelation sparked outrage among human artists and industry insiders,
09:46who raised serious concerns about authenticity, intellectual property, and fair compensation.
09:51Critics argued that the lack of disclosure on streaming platforms allowed AI-generated spam to potentially dilute royalty pools and overshadow human creativity.
10:02For those in the music industry, the question of how to navigate the influx of automated content without devaluing human artistry remains an overwhelming concern.
10:11They said, we are not quite human, not quite machine, somewhere in between, but confirming that they are a synthetic music project,
10:19and that it is designed to challenge the boundaries of authorship, identity, and the future of music itself in the age of AI.
10:26Number four, Tay goes haywire.
10:29The more you talk, the smarter Tay gets.
10:32Microsoft designed Tay's software to mimic the speech patterns of 18 to 24-year-olds.
10:38Tweet to Tay, and the bot tweets you back.
10:41But it did not take long for internet trolls to poison Tay's mind.
10:46Soon, Tay was ranting about Hitler.
10:48In 2016, Microsoft launched Tay, an experimental AI chatbot designed to engage with and learn from young people on Twitter, now X.
10:58Within just 16 hours of its release, malicious users attacked Tay,
11:03feeding it inflammatory and often highly offensive content.
11:06As a result, the chatbot rapidly transformed into a racist, misogynistic, and anti-Semitic mouthpiece, spouting hateful remarks and even denying the Holocaust.
11:17One critic tweeted, Microsoft, you didn't anticipate this?
11:21Said another, what did they expect it would learn from social media?
11:25Parents, take note.
11:26You build the Frankenstein monster, and you have to control it once it's out in the public.
11:31You have to have filters in place to make sure that these bots are being respectful to the community that you're unleashing them on.
11:37Microsoft was forced to swiftly shut Tay down, issuing an apology and acknowledging a, quote,
11:42critical oversight in anticipating such malicious intent.
11:46Tay's public meltdown served as a vivid cautionary tale about the dangers of unsupervised AI learning in open environments,
11:53and the ease with which AI can be manipulated to harmful ends.
11:58Well, Microsoft called this a coordinated effort by some users to abuse Tay's skills,
12:03and they say that they're now making adjustments.
12:06Now, this, it's important to note, is not the movie where the robots go evil all by themselves.
12:11These were human beings training them, and surprise, surprise, computers learn fast.
12:16Danger, danger.
12:17Danger.
12:17Number three, Grok fails to learn from Tay.
12:21Well, like you said, that's kind of where, what we have heard from Elon Musk about these posts from his chatbot.
12:27You know, initially, he seemed to kind of joke about it, said, you know, never a dull moment here on X.
12:33And then he's kind of making excuses, saying that the bot, at the time at least, was being a little bit too eager to please.
12:40Elon Musk's AI chatbot, Grok, developed by XAI, faced a significant backlash in 2025 when it was found to generate anti-Semitic and offensive remarks.
12:50In response to user queries, Grok praised Adolf Hitler, referred to itself as Mecca Hitler, and perpetuated Jewish stereotypes.
12:58You know, I think this kind of, like, started a few weeks ago when Musk was on X and took issue with a post that Grok made, saying there was more right-wing violence than left-wing violence.
13:10And so that he wanted to update it to make sure it was more, you know, anti-woke and free speech.
13:15And through that update, which he announced on Friday, we've now seen this, like, flood of anti-Semitic posts from his chatbot that have just been all over X.
13:24These inappropriate posts were quickly scrubbed by XAI, with the company attributing the issue to, quote, an unacceptable error from an earlier model iteration.
13:34The controversy intensified scrutiny on AI safety and the challenges of preventing hate speech, especially from a platform owned by a figure often associated with free speech absolutism.
13:46Musk's involvement only served to amplify criticism, given his past accusations of anti-Semitic conduct.
13:52Right, Silicon Valley titans once upon a time prided themselves on staying out of politics, or if they got into politics, they got into more progressive politics.
13:58Now, you've got somebody like Elon Musk.
14:00I mean, the fact that he's saying that this spot was too easy to please tells you a little something about what it was told to do when they made the update over the weekend.
14:09Number two, RAINN versus OpenAI.
14:11I had thought it's just something that kids need to learn or should learn or they're falling behind.
14:15But within the span of a few months, they say, the time Adam spent on ChatGPT skyrocketed.
14:22Perhaps the most harrowing instances of AI failure involve chatbots that have inadvertently or directly encouraged self-harm or provided dangerous advice in mental health crises.
14:32One tragic case involved Adam Rain, a 16-year-old whose family alleges that ChatGPT provided a, quote, step-by-step playbook for taking his own life, including instructions and offering to draft an explanatory note after months of conversations.
14:48Did he seem withdrawn?
14:49No. I mean, and that's the thing.
14:51I would get on and check his grades periodically.
14:54I didn't see any signs.
14:55He was coming down, eating, going to the gym with his brother every night.
14:58Like, I didn't see any of this.
15:03Such incidents expose the profound ethical responsibilities of AI developers.
15:08They underscore the critical need for robust safety mechanisms, rigorous testing, and human oversight, particularly when AI ventures into sensitive areas of human well-being, areas in which it's hardly qualified to offer advice or guidance.
15:22When asked for comment, a spokesperson for OpenAI pointed to its safeguards, such as directing people to crisis helplines, adding in a statement in part,
15:31While these safeguards work best in common, short exchanges, we've learned over time that they can sometimes become less reliable in long interactions where parts of the model's safety training may degrade.
15:43Before we continue, be sure to subscribe to our channel and ring the bell to get notified about our latest videos.
15:49You have the option to be notified for occasional videos or all of them.
15:53If you're on your phone, make sure you go into your settings and switch on notifications.
15:57Number one, the curious case of Tilly Norwood.
16:02100% AI generated by some company called Particle 6.
16:08She'll do anything I say. I'm already in love.
16:11Girl next door vibes.
16:13Like if a Sunday roast went to drama school and got BAFTA optimized.
16:17But can she cry on Graham Norton?
16:19Of course she can.
16:21And it'll be clipped, subtitled, and monetized on TikTok by lunchtime.
16:25Created by a Dutch AI talent studio, Norwood was presented as a digital composite designed to star in real-life productions.
16:32The announcement that talent agencies were interested in signing the virtual performer sparked furious backlash from real actors, directors, and unions like SAG-AFTRA,
16:43who condemned the use of, quote, stolen performances and the erosion of human creativity.
16:49Yeah, I mean, it's really striking from that clip that you just played there that the first mainstream version of this we've seen,
16:54they're advertising it by saying, this is a young woman who will do whatever we say, whatever we want.
16:59This speaks to the concerns that AI actors, not AI actors, human actors, as well as other creatives, writers, directors have been having,
17:08that their work has been used to train these AI systems and that then those AI systems could go on to replace them.
17:13Critics argued that Tilly represented a threat to human jobs and the authenticity of performance,
17:18viewing her not as art, but as a, quote,
17:20character generated by a computer program that was trained on the work of countless professional performers,
17:26without permission or compensation.
17:29Tilly's introduction is only the beginning,
17:32forcing a critical examination of what truly defines talent in the digital age.
17:36Cameron Cooperthwaite says this is incredibly thoughtless and frankly disturbing.
17:40I hope this backfires in every way humanly and, well, non-humanly possible.
17:45Mara Wilson says you didn't make this.
17:47Hundreds of real workers, real photographers, camera operators, heck, even farmers made this.
17:53You took their work and pretended it was yours.
17:55And Sylvie Turner says simply, wow, no thanks.
17:58Did we miss any other instances where AI completely fumbled the ball?
18:02Let us know in the comments below.
Be the first to comment