Chatbot Fiascos: 2024’s Funniest and Most Cringe-Worthy Failures
From $1 car sales to glue-covered pizzas, these AI blunders had us laughing—and cringing—all year long.
Good morning, everyone!
We wanted to start the year doing something we all don't do enough: look back and think. Or think in general—just do more thinking. ;)
We came up with a series of cool ideas to reflect on, starting today with the top AI missteps of 2024. In the next one, we'll cover the year's biggest wins.
Buckle up, and let's get started (in no particular order)!
Mistral: From Hype to Silence
Mistral's 2023 ended so promisingly (primarily thanks to MoEs) that it was on our radar for the year. Yet Mistral struggled to stand out in 2024. We wonder if Mistral will be back in the race with top labs.
Claude Computer Use: All Hype, No Substance?
Claude Computer Use was supposed to be the next big thing in AI assistants. People loved it; there was hype everywhere. We remained skeptical (or pessimistic, whichever you see fit, as usual). Many users found it glitchy, inconsistent, and requiring a steep learning curve. It struggled to deliver a smooth and reliable experience. It turns out that sophisticated tech without usability doesn't win hearts. Who uses Anthropic’s Computer Use today?
GPT-5? Nope. Meet GPT-4o
OpenAI skipped GPT-5 entirely, rolling out GPT-4o instead. Was it a strategic pivot toward Artificial Superintelligence (ASI)? Maybe. But let's be honest—it felt more like a stall. GPT-4o didn't bring the leap we were waiting for. Critics weren’t impressed, but OpenAI insists it’s part of the long game.
Sora Turbo: Cherry-Picked and Choppy
Sora Turbo wowed us on launch day with carefully curated demos. But in real-world use? Meh. Inconsistent results and lackluster performance in video generation made the hype feel hollow. Video generation didn’t seem to have progressed much this year, at least to the end user. Are you using it yourself? Did you get good results?! We’d love to know!
AGI: Still a Dream, ASI Takes the Stage
AGI didn’t happen in 2024. Again. Now the focus has shifted to Artificial Super Intelligence (ASI)—probably because if you can’t build a general AI but still want investment money, why not aim for something even more out of reach? In any case, OpenAI seems quite confident they can…
AI Wearables: Fashion Fails
Rabbit R1 and Human Pin promised futuristic convenience. Instead, they delivered clunky designs and confusing interfaces. It turns out that slapping current AI models onto a gadget is not useful. Still, the progress on Meta’s Ray-Ban glasses seems quite promising! It can record videos (sadly, just up to 3 mins) or do live translations with its bone-conduction speakers, which is quite cool but pricy, too!
AI Robots: Collecting Dust
AI-powered robots were anticipated to revolutionize various sectors in 2024. While advancements were made, widespread practical adoption remained limited. Boston Dynamics' Spot robot, for instance, was deployed in mining environments to enhance safety and efficiency, yet its integration of advanced AI capabilities is quite minimal.
Overall, many robots remained confined to research labs or served as high-priced marketing tools, with practical, everyday robots still not becoming a reality. Even Elon Musk couldn’t deliver real-speaking robots in their closed event, simply hiring people to talk in a microphone through the robot! We hope to see more AI-powered robots progress in 2025.
AI Scams and Misinformation: Out of Control
From deepfakes to scams, 2024 was a playground for AI misuse. Fraudsters got smarter, misinformation spread faster, and the public trust in AI took a serious hit. As predicted 4 years ago, generated images and videos become harder and harder to detect. We are not as excited to see what will come next vs. in 2020, though! This is becoming quite scary!
Chatbot Disasters: A Comedy of Errors
Here’s a small rapid-fire list of notable failure cases we’ve witnessed this year, as those are important to remember and work on if we want to diverge from them:
Air Canada: The bot promised a discount that didn’t exist. They tried arguing that the virtual assistant was solely responsible for its own actions. Nope, doesn’t work like that!
Chevrolet: Chatbot sold a car for $1. Yep, and they had to honor it!
Google's Gemini AI: Told people to add glue to pizza. Bon appétit?
Meta's AI: Denied historical events and changed history.
Character.AI: Chatbots encouraging (really) harmful behavior. Lawsuits followed.
That’s it! No tenth one. We are ending on an odd number!
Stay tuned for our next edition, where we'll spotlight the top AI triumphs of 2024. Spoiler: it’s a bit more positive and uplifting!
P.S. Here’s a 10th: To all big tech companies that burn WAY too much money with little return only to stay in the race of AI.