There’s one Uber ride, in par­tic­u­lar, that Beth Turn­bull re­calls clear as day. When the 21-year-old Ir­win na­tive was picked up, the driver told her that he was a pro­fes­sional SpongeBob SquarePants im­per­son­ator.

As a reg­u­lar user of ride-hail­ing ser­vices, she def­i­nitely meets some “col­or­ful char­ac­ters” that she doesn’t mind chat­ting with, said Ms. Turn­bull, an ac­counts co­or­di­na­tor at Mat­ter Com­mu­ni­ca­tions, a pub­lic re­la­tions firm in the Strip Dis­trict. But most of the time, she pre­fers si­lence in the car.

She’s not the only one.

San Fran­cisco-based Uber rolled out a new fea­ture in May, aptly named “quiet mode.” It’s a new ser­vice that al­lows rid­ers to re­quest a no-talk­ing ex­pe­ri­ence.

Uber did not re­turn a re­quest for com­ment for this story.

Quiet mode is avail­able in ev­ery city, in­clud­ing Pitts­burgh, al­though only to those us­ing Uber Black and Uber Black SUV — pre­mium ser­vices that rou­tinely cost dou­ble the price of a reg­u­lar ride.

It’s a sharp con­trast from the ear­li­est days of Uber in the early 2010s, when rid­ers were en­cour­aged to sit up front; it was the anti-taxi ser­vice, af­ter all. How­ever, that close prox­im­ity to the driver made con­ver­sa­tion a ne­ces­sity, more or less.

Many rid­ers in Pitts­burgh choose to sit in the back seat now, a way to help get a si­lent ride.

And that’s good for Uber be­cause the in-app quiet mode may cre­ate an in­cen­tive for cus­tom­ers to up­grade to a more ex­pen­sive Uber Black ve­hi­cle.

‘More than a hired hand’

Kwadwo Burgee, lead an­a­lyst for the Depart­ment of Home­land Se­cu­rity in Wash­ing­ton, D.C., has no in­ter­est in Uber’s quiet mode ser­vice.

For one thing, the 44-year-old Arling­ton, Va., res­i­dent — orig­i­nally from Phil­a­del­phia — said he was raised to keep quiet dur­ing taxi rides.

And why pay ex­tra?

Deb­o­rah Clark knew noth­ing about in­vest­ing when she re­tired four years ago from UPMC.

The only thing she knew for sure was that she never wanted to worry about her health in­sur­ance be­ing paid and that she wanted the money in her re­tire­ment ac­count to gen­er­ate enough in­come to cover the monthly pre­mium through her re­tire­ment years.

“That was the most im­por­tant thing to me,” said the 68-year-old Arling­ton woman.

Her plan was vi­a­ble enough. She had saved $107,000 in her em­ployee re­tire­ment ac­count over 36 years work­ing as an ad­min­is­tra­tor. She wanted to roll the money into a tra­di­tional IRA. She dis­cussed the plan with the PNC Bank man­ager at UPMC Pres­by­te­rian hos­pi­tal where she worked, and the man­ager ar­ranged for Ms. Clark to meet with a bank in­vest­ment ad­viser.

She re­mem­bers the meet­ing be­ing brief and to the point. The in­vest­ment ad­viser made one

tele­phone call to ask some­one how much money was needed to gen­er­ate a monthly in­come stream of $242. He hung up the phone and told her the magic num­ber: $65,475.

“There was no dis­cus­sion what­so­ever,” Ms. Clark said. “But I was com­fort­able with the fact that my health in­sur­ance was be­ing cov­ered.”

She would later find out that the in­vest­ment ad­viser rolled her re­tire­ment ac­count into a guar­an­teed life­time an­nu­ity. The an­nu­ity would pay her $299 a month for the rest of her life. But she would no lon­ger have ac­cess to the $65,475 she in­vested.

By the time Ms. Clark re­al­ized she could never with­draw any of

the money she gave the ad­viser, she said she had al­ready spent the funds she had left over — $10,000 on pre­paid fu­neral ar­range­ments for her­self, $8,000 to bury her sis­ter, and $20,000 on home ren­o­va­tions and travel.

She be­lieves she was mis­led to think her money was in a tra­di­tional stocks-and-bonds IRA that al­lowed with­draw­als.

Mean­while, her three-bed­room home still needs work to stop flood­ing in the base­ment. She has a pay­ment of $540 a month on the 30-year mort­gage she took on in 2004. She pays her bills with a $1,440 monthly So­cial Se­cu­rity check. She has no emer­gency sav­ings.

“It’s very up­set­ting,” she said.

Dar­ren Elias makes a liv­ing play­ing poker, to the tune of $7 mil­lion in tour­na­ment win­nings as of 2019. He also holds the record for the most World Poker Tour ti­tles, with four to his name.

That makes him one of the most fa­mous play­ers out there. But he says there’s one com­pet­i­tor, in par­tic­u­lar, that he wouldn’t hedge his life’s earn­ings against.

Pluri­bus is the name — and this is one Texas hold’em champ that will quickly snatch up all of your chips. That’s be­cause it’s an ar­ti­fi­cially in­tel­li­gent com­puter pro­gram de­signed to win.

Rep­re­sent­ing a sig­nifi­cant leap in AI tech, Pluri­bus has the abil­ity to beat five hu­man poker play­ers, at once.

It was de­vel­oped by Tuo­mas Sand­holm, a pro­fes­sor of com­puter sci­ence at Car­ne­gie Mel­lon Univer­sity, and Noam Brown, a re­search sci­en­tist at Face­book AI in New York City, who is also wrap­ping up his Ph.D in com­puter sci­ence at CMU.

Their aim is not to rev­o­lu­tion­ize poker, but to use it as a mile­stone for prog­ress in AI re­search. “Using games as test beds gives you a way to score how strong the AI is,” said Mr. Brown.

Mr. Elias, 32 of Med­ford, N.J.,

To ratchet up the stakes, the duo took on a new chal­lenge with Pluri­bus — the name means “many.” Texas hold’em is a six-player game.

Most ex­perts in the AI com­mu­nity, ac­a­demic and in­dus­try, thought Mr. Brown and Mr. Sand­holm could never build a sys­tem like Pluri­bus us­ing their meth­od­ol­ogy. That en­tailed a dif­fer­ent kind of ma­chine learn­ing, a method to teach an AI pro­gram how to func­tion by feed­ing

it in­for­ma­tion, called train­ing data. Once a sys­tem un­der­stands re­la­tion­ships be­tween those pieces of data, it can make pre­dic­tions on its own.

With Pluri­bus, there was no out­side data — mean­ing no in­for­ma­tion from ac­tual Texas hold’em games that hu­mans have played. Rather, the com­puter learned to mas­ter the game by play­ing against it­self and us­ing what it learned to cre­ate its own train­ing data that self-in­forms the AI