A fancy way to do lighting, reflections, shadows, etc. It's the way that CG movies do that stuff so accurately. Effectively invisible rays are cast to simulate light bounce in a way that's physically accurate to real life. The techniques we use for such things currently are either inefficient, inaccurate, or are limited in some ways. For example, cube map reflections can't reflect dynamic objects like characters. Render to texture reflections are incredibly computationally expensive. And screen space reflections are neither but can only reflect objects on screen and that aren't being occluded from the camera view by another object (if an object passes between the reflective surface and the camera, or if the camera tilts to the point that an object isn't on screen anymore, the edges of the screen space reflection will disappear). Ray traced reflections are also computationally expensive, but can transfer object colors by simulating bounced light, can reflect objects that are off screen, can reflect dynamic objects, don't disappear at oblique angles like screen space reflections do, and can match the roughness of a surface accurately (so brushed metal will reflect differently than a window, which will reflect differently from a mirror, etc.). You can even get accurate reflections in incredibly tiny objects, like eyes, or a bolt, or a metal piece on a shift knob, etc. And again, ray tracing can do much more than reflections, including volumetric lighting, accurate global illumination, ambient occlusion, soft and increasingly diffuse shadows, etc. Nvidia has had dedicated ray tracing hardware in their RTX graphics cards on PC since the end of 2018, and the Series X/PS5 also include similar hardware level ray tracing capabilities. And now that it's going to be a feature in such mass market machines, we can safely expect the technology to get vastly more optimized over the next few years, so that it should become less computationally expensive.
In the GT7 trailer you can see clear examples of ray traced reflections on the sides of the trailer as they're moving up (at about 30 seconds into the trailer), on the floor of the garage as the car is heading in (at about 36 seconds... they put the driver in front of the reflection to prove that it wasn't screen space by the way... also, notice how reflective his helmet is), in the mirrors/windows, chromed pieces on cars, even the light from the gauge cluster reflecting off of the metal trim of the gauges themselves (34 seconds in)... and most importantly, you can see accurate reflections of the environment, other cars, and self reflections of your own car in your paint. Gran Turismo Sport by contrast used cube map reflections on your car. But that cube map was actually repeated on every car in the race, rather than them each getting their own reflections. You can see this by driving behind someone while passing under a banner or something like that; you'll be able to see the same banner reflect in the car(s) ahead of you when you pass under it, lol. Also, depending on the track, some reflections are really low resolution and look chunky, or update at half rate (so 30fps instead of 60fps).
We don't know if they're doing self shadowing, ambient occlusion, diffuse/soft shadows, or anything else with ray tracing yet. I wouldn't expect it, because again, it's still really computationally expensive at this point. Maybe color transfer from bounce lighting at best, and probably only a couple of bounces. But considering the performance of GT Sport is so high, and other than the lighting and ray tracing GT7 looks to largely be the same as GT Sport graphically, I think they're shooting for a high res, high fps mode, and a lower res, standard fps mode with ray tracing enabled. It's funny, because the car models in the shop section generally are the most impressive. But because the one in the trailer is reflecting nothing but a grey void, it actually comes off as one of the least impressive in the trailer.
Wall of text, but hopefully it was informative without being confusing.