Get Premium
Dark mode theme is available exclusively for premium users. Learn more about the benefits of subscribing.
No fees, cancel anytime.
Dark Mode 
No fees, cancel anytime 
213
Company Bought An AI Machine To Answer Internal Questions, And It Malfunctions So Bad It’s Funny
Lately, we’ve been hearing how AI is going to take over the world. Scientists are trying to create various models that could be implemented in many areas of life. And their effort isn’t fruitless. Artificial intelligence is already implemented in quite a lot of places. For instance, in the workplace, AI models are “hired” to do certain jobs, like answering questions for employees. Unfortunately, it doesn’t always work the way it’s coded, just like in today’s story.
More info: Reddit
RELATED:
Some say that one day, artificial intelligence might take over the world. But some of the current AI models prove that day is not today
Image credits: Hatice Baran (not the actual photo)
A company bought an expensive AI model to help employees with any questions they might have
Image credits: Ankit Rainloure (not the actual photo)
Image credits: Matheus Bertelli (not the actual photo)
Image credits: Alterokahn
But the AI model ended up malfunctioning so badly that instead of employees asking it questions, they asked questions about it, as the model started giving made-up information
To justify layoffs, the OP’s company purchased an expensive ChatGPT-based AI chat model with no name to answer internal questions the employees had.
A few well-known examples of AI models meant for answering questions are Google Assistant, Amazon Alexa, and Apple Siri. Of course, it should be noted that these examples aren’t extensive by any means, and with the recent AI craze, there are a lot of new models coming into work nearly every day.
Coming back to the story, the information given to the AI was very specific, so the model started telling people it wasn’t sure about the results. Shortly after it started, employees began to simply “Google It.” And that’s not even the end. According to the post’s author, now the model simply makes stuff up and ignores the original questions.
The OP talked with the engineers in charge of the model, and they explained that the model has a list of 5-10 top-rated documents it bases its answers on. If none of these documents have a straightforward answer, the AI model starts “hallucinating” and mismatching pieces together.
Part of it could be solved by providing gap-filling documentation, but the problem with this is that the troubles appear due to the way the questions are being asked. And since the company has employees all around the world, it’s possible some employees aren’t forming questions in English. So, the company is left stuck with a self-hating AI model.
Some of the netizens found this situation funny and even made jokes like, “This AI entered its teenage years,” or they expressed how they felt kind of jealous of it, as it can simply tell people to Google the information they need. A few were just simply interested in what made the model malfunction this way.
Image credits: Lukas (not the actual photo)
But AI models can be used for way more than just answering questions. In fact, AI is so ingrained in our lives that we don’t even realize it. For example, nearly all social media platforms use AI in some capacity. When you see recommendations of accounts you would like or simply when the platforms display your feed according to what you’re interested in.
Basically, it’s the same thing with Netflix — it uses AI to improve its content recommendations. With online stores, AI-based algorithms are used to predict, for example, what customers might want to buy. Again, the AI field is already so broad, and it’s growing every day, so it’s hard to keep track of every single thing in our lives that AI is in charge of.
Yet, experts voice that AI shouldn’t be trusted blindly — after all, it’s still a machine. It can be unpredictable, act immorally, or the algorithms it’s based on can be unclear.
In the future, it’s possible that artificial intelligence will evolve enough to overcome these flaws, but then other challenges will arise, such as the models potentially taking away jobs from people. These models even do this right now, or at least attempt to. As we can see in the Reddit story, the AI was bought to justify layoffs, but in the future, the possibility of this might increase.
AI is both useful and unhelpful in many areas, depending on what or by whom it’s being used. At least, right now, it isn’t capable of avoiding burnout, so we can probably feel safe in our jobs, at least for a little while. And what awaits in the future even AI itself can’t predict right now.
Some internet folks thought this whole situation was hilarious, while others were interested in why such a smart technology started acting so stupidly
Poll Question
Thanks! Check out the results:
Newsletter
Subscribe to Access
Exclusive Polls
By entering your email and clicking Subscribe, you're agreeing to let us send you customized marketing messages about us and our advertising partners. You are also agreeing to our Privacy Policy.
Thank you! You've successfully subscribed to newsletters!
What do you think ?
Sonja 8 months ago
People don't understand that AI isn't really intelligent. It's just an algorithm propped on a database. It doesn't understand the data it's given. It doesn't know what any of the data is, it just follows it's protocols and if those reach a dead end, it randomises. On the surface it appears smart, but as soon as you start working with it, you quickly reach it's limits, and that will never change. Even if you could upload all the data from a human brain into its database, that wouldn't change a thing. Even a golden retriever actually knows more and is smarter than any given AI. They just can't make it 'alive'. It can't understand anything. And you can't program understanding. It's the difference between rare data and true, organic intelligence. No matter how much you know, that's not truly a measure of your intelligence. There are hard limits on what a computer can do, and we've reached them years ago. Every new 'breakthrough' of the last years was unreliable and turned out as mere fluff
John Harrison 8 months ago
This entire paragraph misunderstands what AI is.
Load More Replies...
kath morgan 8 months ago
“AI” language models routinely make stuff up because they’re just advanced predictive text engines and don’t actually *know* anything. If you try to use it to retrieve information you’re not capable of fact checking yourself, you’re always at risk of receiving nonsense.
Sonja 8 months ago
People don't understand that AI isn't really intelligent. It's just an algorithm propped on a database. It doesn't understand the data it's given. It doesn't know what any of the data is, it just follows it's protocols and if those reach a dead end, it randomises. On the surface it appears smart, but as soon as you start working with it, you quickly reach it's limits, and that will never change. Even if you could upload all the data from a human brain into its database, that wouldn't change a thing. Even a golden retriever actually knows more and is smarter than any given AI. They just can't make it 'alive'. It can't understand anything. And you can't program understanding. It's the difference between rare data and true, organic intelligence. No matter how much you know, that's not truly a measure of your intelligence. There are hard limits on what a computer can do, and we've reached them years ago. Every new 'breakthrough' of the last years was unreliable and turned out as mere fluff
John Harrison 8 months ago
This entire paragraph misunderstands what AI is.
Load More Replies... kath morgan 8 months ago
“AI” language models routinely make stuff up because they’re just advanced predictive text engines and don’t actually *know* anything. If you try to use it to retrieve information you’re not capable of fact checking yourself, you’re always at risk of receiving nonsense.
Related on Bored Panda
42
14