AlphaZero is a hoax.... (READ)

It's not unfair that Stockfish and all other traditional chess engines are designed in a way that makes it very hard to take effective advantage of massive parallelism so they have to stick with CPUs. MTCS instead of alpha-beta and (what must be) an incredibly expensive evaluation function are the reason AlphaZero is able to use the TPUs effectively, it'd be ridiculous to give Stockfish thousands of cores and say that the systems are equally powerful - at that point, the Stockfish cluster is far more expensive and uses far more power than the 4 TPUs.

Also, the CPU performance numbers in #14 are off. My dual-core laptop can do 198.4 GFLOPS. Skylake-SP can do twice the flops per clock cycle, and Intel's biggest chip, the Xeon Platinum 8180M, has 28 cores at just over 200 GFLOPS each, for a total of 5.7344 TFLOPS (assuming it can keep running at the maximum all-core turbo speed, which is admittedly unlikely. 4.48 TFLOPS if running at stock speed).

But even these numbers are far off from what Stockfish is actually capable of using. It doesn't use floating point - there's no FMA instruction for integers, you can halve that FLOPS number right away. It operates on 64-bit values rather than 32-bit, another halving (though not an actual reduction on the amount of computation done - also note that the TPU's quoted FLOPs appear to be 16-bit operations, you might as well halve those before comparing to 32-bit flops in the first place). It doesn't use the vector registers - divide by at least 4 (512 bits' worth of ops * 2 execution ports -> 64-bit ops * at most 4 ports).

My best guess is that you could have both programs doing about the same amount of pure number-crunching per second if you gave Stockfish about 2000 cores. But this system would be far more expensive and draw far more power than the one used by AlphaZero, so this time Stockfish would clearly be the gorilla to AlphaZero's mouse.

Generally speaking stockfish was deprived some of its trumps, for example 1 minute per move limit was definitely something that jepardized its chances. Another problem is parallelism, one can look at the stockfish's code and easely conclude that it really doesn't support parallelism to the extent which AlphaZero team claims.
In my opinion Sotckfish is not that inferior to AlphaZero as the chart suggest, but AlphaZero is still a very new approach to chess engine creation, they've only recently sovled few problems which have them bothered for a while so I expect a steep improvment in AlphaZero's play. Just my 2 cents.

I keep reading similar posts how Alpha-Zero will "improve", increase it's strength, become the top rated chess engine, will increase our knowledge of chess theory, has developed new ideas in the openings, etc; etc.

The developers of DeepMind have repeatedly expressed they have no intention of ever spending another minute of research time playing the board games of chess, go and shogi.

It was an exhibition by Google showcasing their advances in machine learning. The algorithms (Monte Carlo) are not practical for developing software. A0 had no evaluation functions that could be used for review. The algorithms used will never available for use by the public.

From all reports, A0 will never make another chess move. Debate is made about how equal the playing field was. Often it depends on perspective of hardware vs. software and which is a greater factor. Clearly though, SF had it's strength's disabled. It was no contest. DeepMind's advances are changing the AI field of study.

GM's such as Nepo... in an interview stated after further being educated about the playing parameters was less impressed with A0 than initial impressions. He highly praised the advances made in the new approach of machine learning, but was not all that impressed with the match result as the level of play was not overly impressive.