It's interesting and maybe come popular for long distance travel like cruise control but the whole notion proceeds from a false assumption.........people don't like driving.
Demand would rise/fall.
A pleasant country drive isn't the same thing as driving at 5:04pm in a multilane parking lot with a 1 hour traffic-related delay.
Ridehails could go up considerably, the more congestion there is, in a self-feedback loop, especially if carpool lanes become much more widespread. 407 is also much cheaper by carpool -- you can breeze from Hamilton to Markham in less than 1 hour at peak for just, say, $5 toll charge per person.
It can take in many forms. One possible form this could take is -- congestion is unplesant so you'd opt-in to pick up paying passengers (opt-in ridehail apps might be pre-installed by car manufacturers, to lower your monthly car bill, or get a cut of your ridehail income) so you can drive carpool lanes & use 407 cheaper. Get home faster, make driving more fun, make driving cheaper, despite having to drive passengers. This is often done under the table (I promise to pay your gas) in many countries, apps would just make this more popular, quicker nearby pickup of a vetted 4.5-star or 5-star rated passenger, easier to do, and efficient.
There's solo ridehails, and carpooled ridehails, which can straddle various gap between solo driving and a full size bus. If ridehail carpools show up in less than 5 minutes after pressing a button, the demand will definitely show up especially if time-estimating is very accurate (including accounting for complexities of picking up/dropping off people).
Whether it's a quick hop between Liberty Village and Queen's University, or a longer hop like Hamilton-Markham.
The pickup/dropoff delays could eventually become pretty accurate (yet quite optimized, only a few minutes longer than non-carpooled) in future trip time estimates. Google Maps successfully developed accurate time estimate even for complex stop-and-go routes because of cloud knowledge of previous GPS trips. It would be accurate even if people would be on the wrong side of a divided road, and apps can automatically tell people to walk one block to lock in a 50% carpool time/cost savings (to prevent car pool diversion away from straight line), including accurate estimating of the pedestrian's walking time (plus safety margin) including having to walk to intersections to get across a divided road. Undoubtedly, algorithms for optimized carpools (a straighter line with quick pickups/dropoffs) will get better and speed things up for everyone. If people can see a carpool will be faster than public transit, shorter walk than to a bus stop, and much cheaper than driving, and easy as pressing a button, the market can be there.