1248
(June 2024) AI Drive Thru's Failed But The Rollout Continued (europe.pub)
submitted 2 weeks ago* by to c/fuck_ai@lemmy.world
143 comments fedilink hide all child comments
 

McDonald's is removing artificial intelligence (AI) powered ordering technology from its drive-through restaurants in the US, after customers shared its comical mishaps online.

A trial of the system, which was developed by IBM and uses voice recognition software to process orders, was announced in 2019.

It has not proved entirely reliable, however, resulting in viral videos of bizarre misinterpreted orders ranging from bacon-topped ice cream to hundreds of dollars' worth of chicken nuggets.

McDonald's told franchisees it would remove the tech from the more than 100 restaurants it has been testing it in by the end of July, as first reported by trade publication Restaurant Business.

you are viewing a single comment's thread
view the rest of the comments
[–] 6 points 2 weeks ago (11 children)

If you cannot get the right burger from an Al customer service bot, it is not an appropriate tool for education, tutoring, or any other high-stakes task.

It's not an appropriate tool for those things, no. But this is a fallacious statement. These companies are not utilizing good AI models in their drive through software, they're using the cheapest, most bare-minimum models they can possibly get their hands on. But even if they use good models and do solve AI food ordering, even if they do manage to make it as good as a human for the task, it will still not be an appropriate tool for education or tutoring.

  • permalink
  • fedilink
  • source
  • hideshow 11 child comments
    •
  • [–] 3 points 2 weeks ago (6 children)

    It's not a fallacious statement. You're correct that it would still not be an appropriate tool, but that statement doesn't claim it would be.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 2 points 2 weeks ago (5 children)

    It suggests that it would. It is suggesting that the quality of the models' ability to correctly understand a fast food order indicates its fitness to be used in teaching, but I would say that even a model which can flawlessly take a cheeseburger order—even a model which can flawlessly mimic the action of teaching—wouldn't be appropriate for the task. The quality of its output is unrelated to its fitness to the purpose.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 2 points 2 weeks ago (4 children)

    That last sentence is interesting to me. Care to expand further?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 2 points 2 weeks ago

    Not Op, but fitness for the purpose of education is as follows: being human.

    A non-human cannot teach to be a human to others, just like a non-cheetah cannot teach s cheetah how to be a cheetah. It applies to every social animal...

    Emotions, hormones, dealing with our body's health, that's all stuff that ai model will never be able to understand and teach correctly.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 2 weeks ago (1 child)

    The other two answers are excellent: only a human can teach anyone how to be a human, and AIs are structurally incapable of understanding any concept sufficiently to teach it.

    I might further add: the current structure of AI as a wholly for-profit enterprise adds a level of cynicism and a susceptibility to corruption that even a for-profit human educator is not subject to; that is to say, do we want any corporation (really, any single entity) to be responsible, not only for the structure of education, but for its entire expression?

    Even if a hypothetical perfect AI were completely benevolent, and owned by the people, it would still represent a bottleneck of thought and idea that would heterogenize our society. But we don't live in that world, so it's almost a certainty that education would be for sale to the highest bidder. You think historical revisionism is bad now? What if BMW could pay OpenAI to downplay the role of Germany in World War II?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 0 points 2 weeks ago

    Isn't that a rejection of the premise you set up though? If you are describing a flaw in that an AI isn't capable of teaching a human, then 1) it isn't flawless, and 2) it's circular in that an AI isn't capable because it isn't capable. Would you apply the same the same logic to something like a philosophical zombie?

    Not engaging with the other 2 paragraphs because they are irrelevant to the original comment.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 2 weeks ago

    (I'm not OP)

    If we accept that facts are justified true beliefs (as per epistemology), then someone needs to have a belief (a proposition) which corresponds to reality and a justification to link the two (otherwise it is merely a correct guess). Looking at it like this we can see why AI is unfit for the purpose of teaching, which needs to be factual.

    I assume he could also mean the fundamental problem of induction. For the purpose of teaching, it is required to be truth-apt, but AI can at best be truth-like. Just as it can't tell apart instructions from data it can't tell apart facts from fabrication due to its stochastic nature.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 2 weeks ago (2 children)

    They don't want to replace teachers.

    They want to replace mandatory reporters.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •