A firm providing AI drive-thru tech to fast food chains actually relies on human workers to take orders 70% of the time::Presto Automations recently admitted that most of the orders taken by its AI drive-thru chatbot are actually assisted by off-site human workers.
One would think that by now, these companies would have built up enough training data to no longer require human intervention?
Is their existing “AI” tech just your usual old chatbot, except with a STT and TTS so it’s usable at a drive thru? The article only mentions that they started recently using ChatGPT to assist with speech recognition… so unless I missed it, there’s no mention of their current tech using LLMs at all - just another company trying to climb on board the AI hype train 🤦♂️
Presto said that off-site workers based in places like the Philippines that assist the chatbots will becoming [sic] increasingly expensive, Bloomberg reported.
Good. People in countries who aren’t so well off shouldn’t be exploited as cheap & disposable call center labor IMO.
Not using a LLM is clearly a mistake.
I bet if you provided ChatGPT the menu along with the spoken text it would figure it out no problem.
I guess the STT is a bigger issue.
The McDonalds here had an AI prompt for like a week. I don’t care because all I need to do is say the number for my mobile order and it was faster. But everyone over 30 would be screaming and yelling shit about “who are you”, “what’s happening”, “am I supposed to talk now?”. I still get stuck behind old people that struggle with actual humans at the drive thru.
General technological competence is so far behind what can be offered to consumers. People are the bottle neck, look at bear proof trash can designs. And I don’t think it’s getting better like it was. With the internet now packaged into 2 click apps, the majority of kids are just doing that instead of getting into FOSS and Linux like the majority of the early 2000s internet users.
You realize that Millennials are over 30, and spent their entire lives speed running through the most significant changes, year over year, of the digital age, right?
Millenials are actually somewhat the exception because we actually needed to use computers. Generally speaking it got worse because every fucking thing is abstracted away from consumers.
http://www.coding2learn.org/blog/2013/07/29/kids-cant-use-computers/
Yes, I’m aware of those trends, but I don’t think it’s as relevant in the context of “smart” or “AI” system user interactions. Younger generations have grown up with the touchscreen/voice interface - which is the primary driver of the specific problems you’re alluding to.
So in this context, I think Gen Z, Gen Y, and Millennials are on equal footing when they each individually make the rational decision to either smash their head, or a baseball bat, into an AI run McDonald’s Drive-Thru.
Im 24 and AI cant recognize my fucking accent, and I dont like suppressing it. I want to go full ooga booga caveman and chuck a spear through them.
I’m generally pro-automation if it can increase efficiency, but McDonalds’ ordering AI is terrible. It had issues understanding their buy one get one for $1 deal and then one time I ordered a “bacon McCrispy” which was an item right there on the menu but what I got was a plain McCrispy and a side order of bacon in a breakfast container. They need to send their AI back to training. I’d really just prefer kiosks at the drive thru like they have inside. Voice is the worst way to interact with a computer IMO, but maybe that’s just because most implementations suck. Voice is too open ended though, a kiosk can provide exactly what options are available and as long as it has full set of customization options I don’t think that open endedness benefits anyone.
Also, over 30? Millenials grew up on the Internet for the most part. I’m 34 and grew up with computers and Internet. It was our parents’ generation that fails to understand tech.
Well this kid also thinks that most people on the Internet in the early 2000s were using FOSS and Linux. He doesn’t know what’s going on.
People were mostly using email the way we use social media: sending pictures, dumb chain emails, chatting. Instead of Instagram, you would just go to the comment section of a magazine or newspaper and post your inane ramblings there.
Yeah, email was the social media before social media especially in the 90’s and early 2000’s. I got into Linux in 2005 and that’s also around the time Digg and Reddit started growing. I was never into social media like Myspace and Facebook but I spent a ton of my high school days on Digg.
That said I realize most millenials didn’t get into FOSS or Linux but we did use computers a ton regardless, and smartphones were available by high school/college for most of us.
it takes me about 3 times as long to use an AI bot because I have to ask it a bunch of silly questions first to try and fuck with it.
“Yeah, let me get uhhhh the McTrangle with a side of blubblub, and do you guys do the Krango geep still?”
I’m scared that if they start including all our shit posting in their model data, it will roast me back for saying dumb stuff.
If it doesn’t meet the user’s needs or expectations, then the system is wrong, not the user.
You’re not wrong, but damn I sure would love the user to be less of an idiot. My job would be so much easier
I’m the “kid who’s good with technology” despite being almost 40. Most of the problems I deal with are people who don’t know their passwords.
I should have listened when a younger coworker told me “never let them know that you can fix the printer.”
I used to think like that, but recently i’ve begun to realize that companies need to take some responsibility for their dumb designs.
Have you ever had to deal with SBL (Sign in Before Logon) on Windows? All the uses have to do is enter their username and password into a box, assuming they know what their password is, and windows will automatically log them in and connect them to the corporate VPN.
But it’s quite easy to do it wrong, now they are shown when they start how to do it but they still do it wrong all the time, and if you do it wrong you have to shut the computer down and start again. Uses can be idiots but Microsoft are bigger idiots. Why is it possible to do it wrong?
There’s a fine line between your job being easier and your job being redundant, though
I expect your system to know what I mean when I click this button and tell it to delete everything, of course I don’t want to delete it I just wrote it! It should save it before deleting it so that I still have it after.
I trust you’ll fix this bug in the next version
Remember when people were conspiracy theorists for saying AI would replace a ton of people’s jobs?
Remember when they promised us advances in technology would mean we would work less while enjoying a life of leisure?
Ted Kaczynski remembered
Maybe he’s better at remembering murder
They were right. The problem is it’s a big club and you ain’t in it. We ain’t in it.
deleted by creator
The article states 30% of those jobs are lost. And the 70% that are working are training the models. It is very normal that they start with a lot of oversight, manual intervention and hypercare, they are likely training the models to little my little reduce the amount of people. I don’t work in this company, but in a similar one and I don’t want to think how many people lost their job because of what we did.
AI has already replaced ton of jobs, after all it can be used as a form of automation. Media is over hyping it’s current capabilities, but this is moving forward.
The article states 30% of those jobs are lost.
And good riddance, too! The fewer lives wasted doing menial fast-food jobs, the better.
Edit: imagine simping for minimum-wage shit-jobs. Y’all are like crabs in a bucket.
Only if there is another, better, source of income.
Please let it be UBI, please let it be UBI… 🤞🤞
No? Maybe I just ignored them because that’s such a dumb take. Anything that allows someone to do more work than they could before will lead to job losses, and that includes the vast majority of technological innovation.
The main exception I can think of is domestic appliances, but even that is questionable. Few “jobs” were lost because housekeeping is mostly unpaid labor. Even then, the idea that married women should be able to afford to stay home and take care of chores is gone. I think the automation of domestic work has a lot to do with why it’s now the norm for both partners in a marriage to work outside the home.
No? Everyone has known ai will do things like this since way before anyone can remember - it’s a staple of golden era science fiction for example, referenced in cartoons like the original Jetsons, and part of every single tired doomists anti technology speel since the Victorian era.
I guess there were and still are people who lack the ability to understand progress and seem to live continually in the delusional belief that we’re at the end of history and nothing else will ever change - there’s always people who’ll take any position that lets them avoid thinking too deeply.
This is the best summary I could come up with:
A restaurant tech firm that provides AI drive-thru chatbots acknowledged that over 70% of its orders are actually assisted by off-site human workers, according to an SEC filing in November.
The company even announced a collaboration with OpenAI in March 2023 to utilize ChatGPT to enhance its drive-thru voice assistant features for more natural and human-like interactions.
However, in the recent SEC filing, the company said it actually largely relies on human agents — or what it calls “humans-in-the-loop” — to assist the AI in receiving orders.
“As we continue to improve our AI accuracy and further deploy Presto Voice across store locations, we believe that the percentage of orders that do not require any human agent intervention will reach 30% or better.”
Restaurant chains across the US are increasingly relying on AI to take orders and eventually reduce the amount of human labor needed.
Wendy’s is one such company that is automating its drive-thru service and training its chatbot to recognize customer lingo when taking orders and suggesting better offers.
The original article contains 370 words, the summary contains 172 words. Saved 54%. I’m a bot and I’m open source!
