It's really crazy that ATC isn't 99% computerized. It's exactly the kind of problem which is trivial for a computer program with all the necessary information, but incredibly stressful for a human.
I'd argue against the triviality, but even if we accept that premise, the exact challenge is "with all the necessary information".
Human ATC operators integrate a huge variety of information including voice comms to build a situational awareness of not just the current position and velocities of planes, but also their stated intentions.
Yes, with sufficient coordination you could push all of that over into the cockpit and have the pilots input their intentions directly. We all know that voice transcription is a bit hit and miss (though the phraselogy used would probably be quite helpful), and certainly typing is... not optimal for emergencies.
This aligns with basically all questions about automation in aviation (including in the cockpit). Automated systems can generally out perform human operators in probably the 99.99+% of typical cases. However a human with full situational awareness is still critical in the remaining situations. The challenge is that:
* For a human to have proper awareness, they must be meaningfully engaged. They must actually be having some real agency and effect on the system.
* Neither full human nor full automation provides the degree of safety and performance that we currently desire. Somehow we must integrate the two systems together.
In this case, planes coming from 04R (per the article) were crossing 04L. All the normal crossings of 04L I see in this PDF are equipped with the lights.
The same reason why ACH transfers still happen by transferring large files to a FTP server once a day or why San Francisco's train routing system still uses 5.25-inch floppy disks or why social security payments are calculated by a system running COBOL. It has always worked this way, and there's no political will or public pressure to change it.
