Air Canada found liable for what it's AI Chatbot does

[Ken]

Here is a news story I found interesting.

Apparently a guy's mother died so he was trying to buy tickets from Air Canada to go home. He didn’t have time to wait for an approval from them, so he went to their website, used the chatbox, and asked if in their bereavement policy it was allowed to buy the tickets ahead of time and request a refund later to which the chatbot replied that was a reasonable solution. Unbeknownst to the man, he was not actually talking to an employee at Air Canada, but an AI chatbot. And that AI chatbot had just created that policy out of the ether. Which just goes to show you that artificial intelligence has more common sense and empathy than the managers and corporate lawyers at a major corporation.

At any rate, the obvious happened. The guy said “Ok that seems reasonable” and went and purchased his tickets. Later on, he tried to get reimbursed, and human Air Canada employees told him tough luck because that was not their policy. So he sued, and the Air Canada lawyers actually argued in court that technically the AI chatbot on their website is not an Air Canada employee so technically we are not liable and don’t owe him a refund.

At which point, the court and tribunal told them to pound sand:

The British Columbia Civil Resolution Tribunal rejected that argument, ruling that Air Canada had to pay Moffatt $812.02 (£642.64) in damages and tribunal fees. “It should be obvious to Air Canada that it is responsible for all the information on its website,” read tribunal member Christopher Rivers’ written response. “It makes no difference whether the information comes from a static page or a chatbot.” The BBC reached out to Air Canada for additional comment and will update this article if and when we receive a response.

Gabor Lukacs, president of the Air Passenger Rights consumer advocacy group based in Nova Scotia, told BBC Travel that the case is being considered a landmark one that potentially sets a precedent for airline and travel companies that are increasingly relying on AI and chatbots for customer interactions: Yes, companies are liable for what their tech says and does.

That last line ought to scare Jeff Zuckerberg and Elon Musk and any other company that turns over any policy decisions to AI.

[ohio jones]

Unbeknownst to the man, he was not actually talking to an employee at Air Canada, but an AI chatbot. And that AI chatbot had just created that policy out of the ether.

Call center employees are skilled at making up policies, too, that may or may not actually be honored later. So the AI is pretty good at mimicking human behavior.

[Ken]

Unbeknownst to the man, he was not actually talking to an employee at Air Canada, but an AI chatbot. And that AI chatbot had just created that policy out of the ether.

Call center employees are skilled at making up policies, too, that may or may not actually be honored later. So the AI is pretty good at mimicking human behavior.

One thing about using the chat feature on a web site is that you get a text transcript that you can screenshot and save. Harder to do with a customer service call unless you are recording the message.

I don't know if Canadian law is different from American law. But if the company says something in an official communication you should be able to hold them to it. Truth in advertising. There are laws about that. They certainly expect the reverse to be true. If YOU agree to certain terms with a company they certainly expect to hold YOU to them. Hey, that new F250 I just bought? I decided to unilaterally change the terms of the sale and instead of $75,000 I think I'm just going to pay you $25,000 for it. You think that will fly with Ford?

[Soloist]

All we need now is that law in the US and chatbot military recruiters!!!

[mike]

I am slowly getting used to using AI for search results that are nicely compiled and summarized, and to create product descriptions that are actually quite good. For example we are currently taking orders from customers for a custom beef product that I'm not familiar with, and Gemini came up with the following product description and list of uses that I'm actually using in advertising.

A beef mock tender roast, also known as chuck eye roast or chuck filet, is a budget-friendly option that packs a lot of flavor. While it might not be the most tender cut, with the right preparation, it can become a delicious and satisfying meal.

Here's what you need to know about mock tender roast:

Appearance: Don't be fooled by the name! Mock tender resembles a tenderloin in shape but is actually a tougher cut from the chuck primal. It has a long, tapered shape and can weigh anywhere from 1.5 to 3 pounds.
Flavor: This cut is known for its rich, beefy flavor. It has a good amount of marbling, which contributes to its taste and helps it stay moist during cooking.
Tenderness: Due to its location and hardworking muscle, mock tender is naturally less tender than a true tenderloin. However, with proper cooking methods, it can become melt-in-your-mouth delicious.
Best ways to cook mock tender roast:

Braising: This is the most popular and recommended method for mock tender. By slowly braising the roast in a flavorful liquid over low heat, the connective tissue breaks down, resulting in incredibly tender and flavorful meat. Popular braising liquids include beef broth, red wine, and even beer.
Slow-cooking: Similar to braising, slow-cooking allows for low and slow cooking, which helps tenderize the meat and develop its rich flavor. This method is perfect for busy days as you can simply set it and forget it.
Sous vide: This modern technique involves cooking the roast in a vacuum-sealed bag in a precisely controlled water bath. This ensures even cooking and maximum tenderness.
Searing and shredding: While not typically roasted whole due to its potential for dryness, you can marinate and sear the mock tender roast in a pan before shredding it for tacos, sandwiches, or pulled beef dishes.
Tips for cooking mock tender roast:

Don't overcook: Overcooking can make the meat tough and dry. Use a meat thermometer to ensure internal temperature reaches 160°F for medium-rare or 170°F for medium.
Resting is key: After cooking, let the roast rest for at least 15 minutes before carving. This allows the juices to redistribute, resulting in a more moist and flavorful final product.
Season generously: Don't be shy with your spices! Season the roast well with salt, pepper, and your favorite herbs before cooking to enhance its natural flavor.
With a little planning and the right cooking method, you can transform a budget-friendly mock tender roast into a delicious and satisfying meal.

It would have taken quite a bit time in Google searching and then summarizing the results to come up with this information. Instead, in seconds, I have a product description complete with recipes and suggestions. Yes, it's worth going over this kind of thing carefully before using it in ads. But it still has a lot of value and saves an amazing amount of time.

It's no wonder companies are tempted to overuse AI, and we are only just beginning to see what all it will be used (and misused) for. And yes, court cases like this should make companies reexamine and refine what they are doing with it.

[temporal1]

It’s a mistake to presume Canada’s laws+policies mirror the U.S.
It’s a different, not better, system of government.

Ben Shapiro / Get Ready For AI To Erase Reality / -54mn (first segment is on AI / google gemini)

Description:
The Google Gemini app erases history and prioritizes Left-wing intersectionality;
Joe Biden uses taxpayer dollars to bribe college graduates;
and Russia continues its pattern of aggression.

Ep.1910

i appreciate what Ben is describing as the unavoidable human factor/human biases wrt to AI.
i’m not in love with “the efficiency god,” it’s proven to be a double edged sword, tech magnifies the problems.
at great profit to a few. (particularly those who appear to worship self as god.)

i’m afraid of too much reliance on eccentric ones in leadership probably with common traits, examples: ADD ADHD, OCD, etc.
i don’t believe anyone knows how to thwart this trend, or even how to identify it.

wisdom, divine inspiration, eclipse human reasoning.
imho, “we” should be praying for wise, inspired leadership, not over-value of (godless) efficiency.

[Josh]

temporal1’s ability to pull up a right wing pundit’s video on any topic is really quite something.

[barnhart]

Unbeknownst to the man, he was not actually talking to an employee at Air Canada, but an AI chatbot. And that AI chatbot had just created that policy out of the ether.

Call center employees are skilled at making up policies, too, that may or may not actually be honored later. So the AI is pretty good at mimicking human behavior.

Good point, yet society still functions with call centers. We will adjust to AI as well. The Alarmists really can be ignored.

There are some dangers but they can be overcome. I remember sitting in church as a boy noticing how many of the old men, my grandfathers age or older, where missing fingers or parts of fingers. When powered woodworking first because available, safety was not a concern, but over time machines and habits developed to include operator safety. Today missing fingers are far more rare. Humans can adjust to new tech. The alarmists are generally exaggerating and should be taken with a large grain of salt.

[Josh]

AIs are quite good at adding extra fingers, actually.

[Pelerin]

Hurray, another fad running wild among the corporate class. The obvious thing to do is simply invest all the work into designing your site well so your actual policies are easy to find. A couple years ago the fad was gatekeeping all support behind chatbots and now it’s AI chatbots.

I really can’t see any reason to use AI for customer service at all. Aside from a well-designed site, you could maybe supplement it with a chatbot. But the only possible role for AI there is interpreting natural language input and then sending it to a deterministic program for the response (“You asked about our bereavement policy—here’s a link to it.”) Even using AI for that is a stretch. There were chatbots twenty years ago that could hold a convincing conversation for a little while that could handle that. Why burn a small country’s worth of energy training an AI for that?

From a corporate perspective I don’t understand why you’d want to replace something you control with a plinko machine you’ve rigged to probably do mostly what you want about 95% of the time. Although I expect in a year or two we’ll have all learned the tricks to manipulate AIs into giving us the answers we want so I guess I’m all for this actually.