It’s 1997. In my first job in journalism, I’m a fact checker for New Jersey Monthly Magazine. The job is exactly as it sounds: Writers would submit their articles, and I was to make sure every fact — from the spelling of names to who’s buried in Grant’s Tomb — was correct.
I distinctly remember my first assignment, checking “Going Broke in New Jersey on $100,000 a Year,” by Sherry Sylvester. I was handed the article by an editor, got up from my desk, and logged onto the office internet.
I was then immediately told to log off the internet, as the internet was not to be trusted, all facts had to be confirmed via either phone call or library research.
I could not ask Jeeves, I could not alta any vistas, I could not excite myself. (For you youngsters who have no idea what I’m talking about, go Google it.)
But yes: The internet was not to be trusted.
Fast forward to today, and the above story is ridiculous. Imagine not being allowed to use the internet for research. Mind-boggling.
Now imagine the world 30 years from now, looking back at how we’re treating artificial intelligence today. I’m guessing all our hang-ups and concerns about it when it comes to writing and research and creation will look about as silly as me not being allowed to use the internet in 1997. (Assuming, of course, AI doesn’t go ahead and wipe humanity off the face of the planet, which is an existential column for another day.)
My point: AI is here, it’s not going anywhere, and any efforts to bottleneck it are bound to end up on the wrong side of history.
Settle down
Which is why all the hand-wringing and investigations and question-asking about DraftKings’ purported use of AI to target losing gamblers is just a giant waste of time.
The controversy stems from an examination by The New York Times of how DraftKings supposedly uses AI to figure out which customers would lose more money if offered promotions to keep betting. DraftKings denies using AI to target people based on losses or marketing to them based on signs of problem gambling.
Regulators have questions.
Massachusetts has asked every licensed sportsbook whether losses, increased betting, or other warning signs make a customer more likely to receive a promotion. Answers are due Nov. 3. Louisiana gave DraftKings a conditional, four-month license extension while examining, among other things, its AI practices. New York has proposed restrictions on AI-driven promotions and suggested bets. Michigan is reviewing issues raised by the reporting on DraftKings’ AI use and VIP program. Maine is monitoring AI use. And Maryland’s governor wants legislation banning operators from targeting people showing problem gambling patterns, with or without AI.
To be very, very, very clear: I don’t think gambling companies should prey on their customers.
Also to be clear: Fighting against use of AI is missing the point by a mile, all you regulators. Either ban all promos and come-ons and VIP programs and all of it, or realize that AI is here, it’s not going anywhere, and attempts to ban, regulate, or otherwise put the kibosh on it is like Wile E. Coyote trying to catch the Road Runner.
Not gonna happen, in other words.
And even if it were to happen, just what, exactly, are we calling AI?
View from Vegas
“AI is such a broad term, and it encompasses so many different methods and techniques,” said Kasra Ghaharian, director of research at the International Gaming Institute at UNLV. “Robotics, generative AI, vision AI, machine learning, natural language processing. There’s so many things that fall under that large umbrella term of artificial intelligence.”
Companies have been studying their customers and trying to sell them more stuff forever. Presumably, that’s still going to be allowed. So at what point does the computing get fancy enough to require a gaming commission hearing?
Ghaharian figures that if states start banning AI, whatever that means, operators could simply describe whatever they’re doing as statistical modeling and carry on.
“And how do you even police it?” Ghaharian wondered. “Are they going to be requiring independent parties to go into these companies and check what they’re doing?”
And … how far into the company would those people need to go? Someone could use AI to build a system that sorts customers by their likely response to an offer. Someone else could use it to write the email. Are both uses going to be prohibited? One? Neither, provided they call it something else?
“So I feel like probably everything is getting touched by AI in some form at this point,” he noted.
Ghaharian’s previous research into AI also found regulators and the industry weren’t even on the same page about how AI is being used. The industry reported using it primarily for cost efficiency, he said.
His suggestion is for regulators and operators to actually talk to each other about what they’re doing, which isn’t really happening.
“There is an appetite to regulate this amongst gambling regulators, and a high appetite but low knowledge of actually how it’s going to be used is a very dangerous mix, and it’s going to lead to more things like this,” he said.
Losses aren’t all the same
Meanwhile, we’re still left with the issue of gamblers getting an offer to bet another $100 when they’re having trouble paying the electric bill. To be clear once again, I don’t want that person getting that offer, and Ghaharian agrees companies shouldn’t aggressively market to customers identified as being at risk of gambling harm.
But DraftKings doesn’t have that person’s electric bill. Or, for that matter, their BetMGM activity. All it has is their DraftKings activity.
“They’re not seeing a holistic picture of someone’s wagering across different platforms,” he said. “They’re also not fully seeing a good picture of their financial life. A thousand dollars to one person is a rounding error, and $1,000 to another person can be absolutely devastating. You can’t really quantify harm by just losses.”
Beyond losses, there’s also all the other data the sportsbooks collect. But Ghaharian notes that this is far from a solved issue.
“It’s very fine and dandy and easy to say, but defining people who are at risk with behavioral data is very challenging and hard,” he said. “We don’t have a good way to do it yet.”
That seems like a useful place to spend some time and money. I’d love for sportsbooks to get better at identifying people who need to stop gambling. I’d love for those people to stop getting offers encouraging them to gamble.
And if we’re going to restrict those offers, the restrictions should apply whether a computer picked the recipient or a VIP host did it over lunch.
Because otherwise, what have we accomplished? A customer can still get the same offer, lose the same money, end up with the same problem, but everyone can feel better because no artificial intelligence was involved.
Seems like a gigantic waste of time.
