Manufacturing Broadcast TV AI Video Compression Hardware Unit, Anyone?

I was often frustrating to Google-search to find a reliable news sources whether many professional broadcast TV equipment manufacturers were open to develop a video hardware encoder unit equipped with AI video compression as a substitute to conventional base digital computing power of advanced video codecs such as HEVC, AV1, and VVC. Of course, everyone assumed like me that the AI-controlled video encoding method must ensure the processed media content should be fully compatible to older digital media devices that only support MPEG-2, MPEG-4, and/or H.264 video decoder. So what kind of computing chip prototype containing a combined RISC CPU, NPU, and video encoder hardware module should be used in order to perform that matter?

While NVIDIA that was responsible to develop a proprietary NPU chip only focuses on producing a server unit with such mentioned device targeted to AI application clients, and even though Intel, AMD, Apple, and Qualcomm were followed suit to develop their own NPU-equipped CPU, GPU, and SOC in consumer market, there would be little or no reliable information has been posted about the possible development of an "alternative" SOCs containing a combined CPU, GPU, NPU, and video encoder engine applied for professional digital TV broadcast application in order to perform AI video compression technique. Of course, not all current professional TV broadcast equipment manufacturers must always be advanced their brainstorming to produce a unit whether to include either from major computing chip brands or rather produce their own computing chips no matter of their own marketing interests.

Imagine that suppose in visionary that BBC, NHK, or any other global media broadcast science research department in separate ways of their secretive project would possibly develop an SOC prototype containing a RISC-based CPU, GPU, NPU, and legacy video encoder module. Unless either of these research company may possibly partnered with major and/or third party electronics providers as well as adopting from previous research papers, the future TV broadcast AI video encoder computing chip may worked well using a proper programming in AI neural algorithm node that must be free from bugs and even "hallucination" visual computation. The system must be closely based from MPEG-5 LCEVC previously devised by V-Nova, for instead of typical medium to slow encode time method an AI compute geometric tracing method is used to make a separate base low resolution and residual high resolution. The AI video encoded content must be always compatible to older DTV receivers containing only MPEG-2 and/or H.264 video decoder.

One of the recent example regarding the development of server hardware video encoding board was Argos developed by Google used to encode uploaded video to AV1 codec for YouTube streaming. If that's possible, why not to make a related concept for developing AI video compression encoding hardware for live DTV broadcast? Time will tell if you would ever agree on my ideas as to tolerate my expectation.