I remember finding some old MMI data books with the Fortran(?) source code for PALASM for the original PALs, back when I first started working. I think AMD swallowed MMI. The 16V8 mentioned in the original question was capable of being configured as a number of the MMI devices such as 16L8, 16R4 etc., including being pin compatible.
Ferranti, if I remember correctly, produced the ULA - uncommitted logic array, which required a custom metal layer, so one step towards an ASIC.
Signetics, MMI, AMD and a couple of others took the first steps in programmable devices (be it PROMs, PLA's or PAL/GALs). If I remember well, TI and Intel took some steps in that direction too in those early days.
Lattice Semiconductor (which spun out of AMD), Altera and Xilinx came much later.
Lattice came with the concept of 'in system programmable' with their ispLSI devices - all CPLDs. And then began the race to bigger, faster, wider ... ;)
It depends on semantics.
Signetics was the first Programmable Logic Device vendor, but MMI went smaller and cheaper with their simpler PALs, but they too lost out as others later released reprogrammable devices.
Some used registered bipolar proms, as state engines, so they too were single-package, programmable logic devices.
I agree with you on the memory being used as a PLDs except for the fact that I think the question was Who manufactured the first PLD? The first PLDs and FPGAs were essentially and basically just large blocks of memory anyhow with SW that configures them as logic devices.
This is why the big memory houses used FPGAs to verfiy their new processes.
The Other Tesla David Blaza5 comments I find myself going to Kickstarter and Indiegogo on a regular basis these days because they have become real innovation marketplaces. As far as I'm concerned, this is where a lot of cool ...