Taco Bell rethinks AI drive-through after man orders 18,000 waters

BBC ^ | August 29, 2025 | Shauna Macalumm

[The Louiswu]

Has anyone done the Touring test on any of these AI systems yet?

[telescope115]

AI is not intelligence. It basically takes known information and averages it out to try to guess what might be the answer.

It brings to mind a scene from I, Robot, https://www.youtube.com/watch?v=sOKEIE2puso

[noiseman]

AI is not intelligence. It basically takes known information and averages it out to try to guess what might be the answer.

You’re right, but it’s actually even worse than that. What is the definition of “known information?” Turning an AI large language model loose, using as its knowledge base the unfathomable tangle of facts, errors, truth, lies, and opinions of every type on the internet, and then expecting it to somehow (magic?) come up with consistently rational and accurate answers is idiotic. This drive for “general AI” is nothing less than the atheistic materialists’ pursuit of a “god” that they believe they can control, and which will require nothing of them (which is the primary drive behind atheism in general -they hate the notion that there might be a real God who outranks them).

But this materialistic mindset causes them to make false assumptions about the nature of thought, intelligence, and consciousness. They refuse to acknowledge the existence of the soul, so they must ascribe every characteristic that makes someone “human” to nothing more than biochemical and electrical processes in the brain. THAT is how we got to where we are now, with these morons believing that all it will take is a sufficiently advanced computer to equal and then surpass human thought.

That said, AI will most likely exceed human abilities at solving narrowly-defined problems, like playing chess or even driving a car (though even that task contains enough uncertainty and unpredictability that it will take longer than expected to make it reliable and safe enough). Where AI will shine is in its ability to access information from a wide variety of sensors that can give it more situational awareness than a human can achieve via his five senses, and in its incredible speed at processing incoming data. That is why it has certain advantages at driving a car (even though right now it is still far too susceptible to making a “rookie” mistake that could get the occupants killed). The AI “driver” can see in all directions, has access to sensors like night vision and radar, and can process everything its sensors see at lightning speed. This are big advantages for doing something like driving a car. But that is a far cry from trying to turn AI into an all-knowing oracle that can answer every human query.

[telescope115]

In the scene, Will Smith describes how a robot used “logic” to save him over the little girl.

[pgyanke]

I’m a skeptic and believe this example is just the tip of the iceberg. It will create a catastrophe and the inability to recover from it-for the country or some individual or individuals...

I see reliance on AI allegorically to reliance on the government. With individuals and businesses, you have recourse. With AI and the government, you don't.

[fireman15]

I went to the article and could not find any links to the Instagram video of someone ordering 18,000 waters. So, then I searched with Google and ChatGPT and asked for “link to instagram video of AI app ordering 18,000 waters at Taco Bell” which took me to Instagram and the article, but still no video despite Instagram being a video site.

I did get a link to this sort of humorous unrelated video:

https://www.youtube.com/shorts/tNK-3_H8h8U?feature=share

[dfwgator]

[CodeToad]

The thing about AI not being intelligence is that it cannot determine the most appropriate or best outcomes. It knows nothing of purpose.

[Salman]

The problem is the corporate types do not understand the technology, but have seen a lot of science fiction.

[fireman15]

OK, I found the video, but it is just stupid with a customer ordering 18,000 water cups and immediately getting handed off to an employee. Why is this even mentioned in an article, or supposed to be funny?

Here is the video on instagram:

https://www.instagram.com/p/DCE23x_RVVy/

and the same thing on Youtube:

https://www.youtube.com/shorts/DF8Pny3VTg8

[Thank You Rush]

And it’s no fun when it has to be with a business or an individual - imagine how much fun it will be when it’s AI and/or the government. It’s back to “lots of luck.”

[V_TWIN]

“87% isn’t good”

-

I certainly don’t think it is.....I was a data analyst in the aviation field, 87% was never acceptable.

[Bon of Babble]

with the aim of reducing mistakes and speeding up orders.

with the aim of reducing mistakes and speeding up orders reducing labor costs

Fixed. In California fast food workers earn $20/hour - soon to go up to $22, then $25/hour.

Result?

"Approximately 18,000 fast food jobs were lost in California following the state's minimum wage increase to $20 per hour for large chain restaurants, which took effect on April 1, 2024."

Doing the job Americans won't do, right? And to afford them a "middle class lifestyle" - according to Gavin Newsom.

[UnwashedPeasant]

Cracker Barrel would use a transgender AI.

[Tom Tetroxide]

It’s just the movie, “Idiocracy,” that has become reality with the scene of the Carls Jr.

https://www.youtube.com/watch?v=5d7SaO0JAHk

[fuzzylogic]

It’s a lot more than “averages” though. I know you must be familiar with neural networks but there’s layers and layers of NN interactions that achieves something far more special. Passing medical and Bar exams at levels above average is quite impressive - although I’d agree, for those, that does become largely training + word sequences, via vector databases. But the ability to compare and contrast complex topics, analyze pictures, deconstruct UML diagrams into requirements, use-cases, solving equations, etc.., is like having an expert consultant sitting next to me. A tool I use every day.

At some point the distinction between “thinking” at a biological level vs. “simulated in technology” is going to be hard to define. Some would argue it’s already there.

This example is nothing more than lack of effective QA testing. It’s embarrassing to the profession - but hey, it’s not FuSa related with no liability. This was just inexperience and being cheap respective to developers and testers.

It’s also inevitable. Any time there’s new technology the initial hiccups happen. Anyone thinking this is some sort of passing fad is going to be disappointed.

[Mr. K]

This isn’t a problem with AI itself, it’s a problem with a poorly programmed application.

AI used correctly would have specifically solved this type of problem.

[fireman15]

I have been playing chess against computers and computerized chess sets for many decades... I don’t think that many people called this AI back when I started. It was a good way to improve my game to the point that I typically beat pretty much everyone that wanted to play chess with me.

In the early 90s I purchased an AI suite of programs... but at that time it was nearly useless.

So, I originally thought this recent proliferation of AI apps and websites was mostly hype... until I actually started using them on a daily basis. Now I am getting vastly more useful information back from computers and phones with less effort than ever before. And various apps and websites are incredibly helpful with various types of creative endeavors as well.

It is not perfect obviously, but as an example, for fun I wanted to host several websites on an inexpensive mini-pc using WordPress all with free SSL/TLS certificates without a static IP connection to the internet. This is the type of thing that most people here would not know how to get started setting up... and without using AI to help by writing scripts, quite tedious as well.

Both Chat GPT and Grok were able to come up with solutions for this that eliminated most of the tediousness. And I learned quite a bit as well. So here is an example of a silly WordPress website with a free https certificate with content that was almost completely setup using AI and with content generated with AI hosted on a $150 computer that is hooked to the internet with a dynamic IP connection.

The computer uses just 5 to 10 watts at idle, so I just leave it on most of the time. The website is on a Virtual Machine running WordPress in Apache2 on Ubuntu Server.

https://tedskyco.info/

This is completely rudimentary but an example of how AI can be helpful setting up websites.

[I want the USA back]

Someday, maybe, people will realize that AI is not intelligent, is not sentient, doesn’t know what it’s doing, and has no conscience.

They will continue to put AI in charge of things and continue to be surprised at the outcome.

When AI is given control of something, like for example the world, it will take it over or muck up real bad.

[CodeToad]

I use the term “averages” for the common person to understand the concept. Trying to have someone to understand such concepts as quantitative segmentations and, yes, fuzzy logic *snicker*, is difficult.