Reading that these tweezer beams are focused with a microscope objective raises the question of how much energy is in one of these beams.

For example, what would happen if the tip of the beam made contact with a piece of paper, steel, etc.?

Also, what precautions, if any, need to be taken to get air out of the system you use these in? If vacuum, how much?

Would it be possible to build your own, e.g. something like the one shown here: https://www.youtube.com/watch?time_continue=104&v=Sq7GaO8iqu...?

this is tractor beam yes?

Note that usually you are also using the microscope to observe what you are trapping, so it' super convenient (but not easy), to use the same objective that the one you observe to send a laser through (in the opposite direction) and trap your sample. If you send your laser-beam quasi parallel it will focus and trap particula quasi-where your are observing.

(source: did my PhD with optical traps: http://matthiasphd.herokuapp.com/html/parts/part1.html#optic..., picture of trapped beed as viewd through microscope: http://matthiasphd.herokuapp.com/html/parts/part3.html#exper...)

Usually optical traps use 1064 nm NdYAG lasers, which is in the infrared. Most objectives aren't designed for infrared light, so the objectives used for optical trapping are usually customized.

Chu's prize was for laser cooling, which uses a magneto-optical trap. Magneto-optical traps confine atoms by exploiting the Zeeman effect, a spatially-varying magnetic field, and a laser with a precisely-controlled wavelength tuned to be slightly red of an electronic transition. Optical traps, on the other hand, rely on the refraction of a much larger, but still microscopic, object.

Some of Chu's work uses an optical trap, but that work did not earn his prize.

I remember as a child watching an episode of some version of battlebots, where a competitor had an articulated grabbing arm on top of his robot. The arm had, I believe, two articulating joints and then an articulating grabber/jaw at the end. To control the arm he had made a small 2d model of it and could simply move that around on a clipboard like backing, and it would map those movements to the servos in the arm. I was blown away by how intuitive and simple of an interface he had made for this contraption that I imagined would work with a complicated series of levers similar to those used by backhoe operators. I credit this in no small way with inspiring my ever-persistent need to automate as much as possible in my adult life.

EDIT: If anyone has any idea which episode of what show this was, I would be very excited to watch it again

The paper I linked describes it a bit more, but it's unfortunately behind a paywall. I'll summarise it here.

They used a camera connected to a LabVIEW interface to view the trapped particles, but they control it using a laser and a spatial light modulator (SLM), which is able to imprint a spatial pattern onto the laser beam, which is ultimately what creates the optical traps. The paper says "The trap coordinates [on the iPad] are synchronized over the wireless network with a desktop computer, which controls the SLM using our freely available LabVIEW software" (square brackets mine), i.e. they control the traps using a separate computer interfaced using LabVIEW to the SLM. The iPad is used to show the video stream and allow the user to set and manipulate the position of the traps. The camera is also connected via LabVIEW: "Using the JPEG compression available in the National Instruments Vision library, we can stream around ten frames per second from the control PC to the iPad over a wireless network (limited by available CPU power on the iPad). Up to 11 optical traps can be simultaneously dragged around, and they can be created and removed with a double tap on the screen. Double tapping with multiple fingers creates multiple optical traps simultaneously, which is very useful when trapping non-spherical objects."

I guess a more up-to-date iPad would manage more traps. The one they used (circa 2011) was probably 1st generation.