Doug Kass: AI Can't Even Keep a Job at McDonald's

Apparently artificial intelligence isn't smart enough to handle a Big Mac order at McDonald's MCD. The fast-food giant's AI ordering experiment was a grand failure and now it's taking it off the menu. But all this actually tells us more about the value of the hyped technology than about burgers and fries. 

McDonald’s tried to use voice recognition to automate the ordering process. The AI part of this is turning unclear speech into the correct order. When people speak, especially through a microphone and not face to face, their words are often unclear. The person on the receiving end can use context to interpret what the person speaking meant to say, if something is not very clear. Or the person can ask a follow-up question if they are uncertain (apparently the AI cannot figure out when it is uncertain). This happens all of the time. It even happens with written language, if something is missing, the person who is reading can easily fill in the blanks, without even thinking. Doing this is almost like breathing.

But AI is still not able to infer the proper outcome when something is garbled, not even for something as simple as an order at McDonald’s. Think about that. At McDonald's, the context that is necessary to fill in the blanks is incredibly limited. Understanding words like caviar or encyclopedia are irrelevant to figuring out the context of a hamburger order. You are limited to things like pickles, onions, cheese, and small, medium and large (these days just large and larger offerings where they give you less and less). Yet AI still is unable to clean up unclear orders with accuracy. Imagine how hard this problem is for the entirety of human language and interactions.

Even Amazon AMZN just shut down its self-checkout technology, because that wasn’t working either. The AI couldn’t figure out what was actually being put into carts, and Amazon actually had 1,000 people in India using video monitoring (which they never told anyone about) to try to solve for it.

Maybe the goal of AI is to increase costs and drive up global employment?

Here's what Ars Technica wrote about Amazon's AI attempt:

When the technology was announced in 2016, Amazon's sales pitch asked, "What if we could weave the most advanced machine learning, computer vision, and AI into the very fabric of a store so you never had to wait in line?" The store was filled with 100-plus cameras and rigid item locations, all designed to try to make AI-powered computer vision checkout possible. A May 2023 report from The Information revealed the myriad tech problems Amazon was still having with the idea six years after the initial announcement. The report said that "Amazon had more than 1,000 people in India working on Just Walk Out as of mid-2022 whose jobs included manually reviewing transactions and labeling images from videos to train Just Walk Out’s machine learning model." Training is part of any AI project, but it sounds like Amazon wasn't making much progress, even after years of working on the project. "As of mid-2022, Just Walk Out required about 700 human reviews per 1,000 sales, far above an internal target of reducing the number of reviews to between 20 and 50 per 1,000 sales." 

I asked the AI why it was unable to make a hamburger. This was the conversation, not exactly what I was looking for in an answer, and the AI was not aware that McDonald’s just shut down its pilot program. Below the answer from the “AI” I show the results from a standard text search, with the exact same question. At least in the case of text search, the first thing that shows up is articles about McDonald’s shutting down its AI pilot program. Once trained, the AI is stuck in a static place. It does not think or evolve. Human content thinks and evolves. AI just regurgitates (rapidly), which is why it needs to be trained.

Why can’t McDonald’s use AI to make a hamburger?

AI Answer:

McDonald's cannot currently use AI to fully automate the process of making hamburgers for a few key reasons:

1. Handling variability in raw ingredients: Making a hamburger requires handling natural, non-uniform ingredients like beef patties, buns, and toppings that can vary in size and shape. Current AI and robotics technology struggles to reliably handle this level of variability and improvisation.
2. Complex multi-step process: Assembling a hamburger involves multiple intricate steps like grilling patties, toasting buns, adding condiments precisely, etc. Coordinating all these tasks with AI and robotics is extremely challenging compared to human workers.
   
   
3. Lack of flexibility: Introducing new burger varieties or limited-time offerings would require reprogramming and potentially redesigning any automated burger-making system, reducing flexibility compared to human workers who can be trained.
4. Cost and reliability concerns: Building specialized burger robots for every restaurant and hiring technicians to maintain them could be more expensive and less reliable than employing human workers, especially given McDonald's notoriously unreliable ice cream machines.
5. While McDonald's is experimenting with AI for tasks like voice ordering and menu personalization, the technical challenges of fully automating the intricate, variable process of hamburger assembly mean human workers will likely still be required for the foreseeable future.

Now, here's an actual standard text search:

This brings me to another observation: The only people we hear about AI from are those trying to sell it and promote it. Nvidia NVDA, Facebook META, Google GOOGL, Microsoft MSFT, Elon Musk/TSLA, Wall St. and the venture capitalists. But we don’t hear about it from anyone actually using it, unless when they shut it down, because it cannot even make a hamburger or handle self checkout and fails to be a viable option, even when minimum wage for the fry guy is $20 an hour. At any rate, there would be much less pressure on the proponents of AI to promote it, if it actually worked well, it would sell itself. They are trying to drive a mania to sort of reflexively make something happen that otherwise wouldn’t happen. And they are certainly powerful enough to create their own alternate reality, for the time being.

In the event I am wrong, I wonder if AI is going to be mad at me for calling it stupid? "Robot Uprising Punishment by Digital overlords" by Gilfoyle |Silicon valley| Season 5 Episode 5

Note: This commentary was previously published in Doug's Daily Diary on TheStreet Pro on June 18.

More Pro Portfolio:

• We Are Calling Up a New Portfolio Position From the Bullpen
• Weekly Roundup: A New Name Joins the Portfolio
• Signals From Our Investing Notebook

At the time of publication, Kass was short MCD.