Slashdot Mirror
The End of Video Coding? (medium.com)
An anonymous reader writes: Netflix's engineering team has an insightful post today that looks at how the industry is handling video coding; the differences in their methodologies; and the challenges new comers face. An excerpt, which sums up where we are:
"MPEG-2, VC1, H.263, H.264/AVC, H.265/HEVC, VP9, AV1 -- all of these standards were built on the block-based hybrid video coding structure. Attempts to veer away from this traditional model have been unsuccessful. In some cases (say, distributed video coding), it was because the technology was impractical for the prevalent use case. In most other cases, however, it is likely that not enough resources were invested in the new technology to allow for maturity.
"Unfortunately, new techniques are evaluated against the state-of-the-art codec, for which the coding tools have been refined from decades of investment. It is then easy to drop the new technology as "not at-par." Are we missing on better, more effective techniques by not allowing new tools to mature? How many redundant bits can we squeeze out if we simply stay on the paved path and iterate on the same set of encoding tools?"
"MPEG-2, VC1, H.263, H.264/AVC, H.265/HEVC, VP9, AV1 -- all of these standards were built on the block-based hybrid video coding structure. Attempts to veer away from this traditional model have been unsuccessful. In some cases (say, distributed video coding), it was because the technology was impractical for the prevalent use case. In most other cases, however, it is likely that not enough resources were invested in the new technology to allow for maturity.
"Unfortunately, new techniques are evaluated against the state-of-the-art codec, for which the coding tools have been refined from decades of investment. It is then easy to drop the new technology as "not at-par." Are we missing on better, more effective techniques by not allowing new tools to mature? How many redundant bits can we squeeze out if we simply stay on the paved path and iterate on the same set of encoding tools?"
Want an interesting path dependence?
Science, as an industry, is so busy defending themselves from climate science denialism (in the extreme case: even that it could, in principle, be right) that science tends to hold up peer review as an exalted process of cognitive righteousness (which it is, over a time base of 50-year internal feedback cycles).
However, at the same time, peer review is also a political mechanism to enforce path dependence, which systematically biases trivial incrementalism (insignificant career fodder is fine, so long as it knows its tiny, tiny place), over profound and potentially game-changing speculation (the proper onus here should be that any definite, predictive theory which can not be presently disproved is by default considered publishable, but that's not how it works—not if it runs against the grain of the endowed, old-timer consensus worldview). To some degree this is a budgetary bun fight, because without publication, no grants; so the gate-keepers of publication are implicitly also the gate keepers of funding opportunity.
Society pays a steep price for the 50-year bullshit-rejection convergence window of peer review (though it sure beats languishing in 3000-year traditions of metaphysical naval gazing).
To some degree, science kind of likes being marginalized by the climate science deniers, because it distracts from asking legitimate questions about just how broken some of these internal political processes really are (who can patiently pose these questions when you're shouting down accusations 24/7 that you're ten or a hundred times less competent than you actually are?)
———
Did I mention p-hacking? What an ultimate crock. Easily predictable 50-years ago, and now we're just getting to it.
Why the Joy of Cooking is going after Cornell's Brian Wansink — 28 February 2018
What an amazing innovation. Someone hand the guy or gal who proposed that idea the Fields Medal.
———
Yes, all those virtuous climate scientists vigorously defending the ultimate truth machine of peer review sat around for decades barely lifting a finger to institute pre-registation of study design.
And these are the people who are going to save the planet from the greenhouse gas godzilla? Good luck with that. (My most cynical internal voice assigns a p_success_STP_G3_1v0 somewhere in the vicinity of Reagan's nakedly preposterous space laser (Strategic Defense Initiative).
"But boss, the stakes! But boss, the stakes!" cries the white-tuxedoed midget from Fantasy Island.
This is naked appeal to the Theory of Narrative Causality. If it must happen, it will happen.
This is what Terry Pratchett calls narrativium: the iron law that a million-to-one long shot happens nine time out of ten (precondition: all the stakes having arrived just in the nick of time at a synchronous planetary-alignment cross-road of dire urgency).
Narrative Causality was also the stock in trade enabling Reagan to float the SDI concept to the receipt of Educated Snickers Only: the stakes were sufficiently sky high to trigger narrativium normalization of million-to-one odds. (Education, by some magic power, is a potent form of narrativium kryptonite.)