@rampageJackson
No, AlphaZero came up with the first Neural Network implementation for chess, which was based on expected game results of the moves, but the NNUE thing was completely unrelated technology developed by a Shogi programmer, and the technology was so strong that simply replacing Stockfish's eval with NNUE made it stronger already.
That programmer joined the Stockfish team just to give them the gift of this code, which increased Stockfish's development speed ten-fold for about two months (not exaggerating, it was the improvements of 20 months in 2 months.) This technology is based on score prediction with more depth, that is, the engine learns to predict at low depth what it'll show on bigger depths.
Then someone came with the idea of using the higher depth of that to make a new net that at low depth can predict what that net would show at high depth. And then another which does that to that one, so there were high improvements just by training nets like this.
That's the main source of Fat Fritz 2's strength, all Albert Silver did was gather training resources to train a net and press the go button, and wait. And independent testers showed that his efforts weren't even better than the Stockfish's net at the time (you read this correctly, people would be better buying Fat Fritz 2 and throw its engine away and put the Stockfish version they used instead and get better analysis, which means Albert's efforts weakened the engine!)
The problem was that version couldn't be had by... the layman, that's why they had to accelerate the release of Stockfish 13 to show clear superiority (and nowadays Stockfish 14 leaves 13 in the dust, anyway.)
What Chessbase did was nothing wrong, I actually commend them on being the first because the Stockfish's license allows and entices it. The problem was THE WAY they did it. Claiming it was something new, whilst they only changed 12 lines of code (to accomodate Albert's efforts of doubling the net's size), and being very vague about the engine's origins, and withholding the net they sell commercially (while putting on the available source code a dummy net which is weaker and doesn't allow people to build Fat Fritz 2 on their own.)
It's basically plagiarism, all they needed to do was give full credit and accept all they're giving you is a different net for a free open source engine (anybody saying they also give GUI, bells and whistles forget they sell much cheaper products with the GUI, bells and whistles - basically people are much better grabbing Komodo Dragon at this point, because at least it's not something free yet.)
The irony - They didn't do it the right way because they feared it'd hurt sales. It's scammy to sell something pretending to be something else. But the scandal might end up killing their sales.
The irony of the irony - all publicity is good publicity, the bigger this gets, the more people are going to know they exist, they're gonna look them up, they're gonna see they sell stuff, and they're gonna buy. Maybe they're geniuses doing this on purpose so people know about them.
@rampageJackson
No, AlphaZero came up with the first Neural Network implementation for chess, which was based on expected game results of the moves, but the NNUE thing was completely unrelated technology developed by a Shogi programmer, and the technology was so strong that simply replacing Stockfish's eval with NNUE made it stronger already.
That programmer joined the Stockfish team just to give them the gift of this code, which increased Stockfish's development speed ten-fold for about two months (not exaggerating, it was the improvements of 20 months in 2 months.) This technology is based on score prediction with more depth, that is, the engine learns to predict at low depth what it'll show on bigger depths.
Then someone came with the idea of using the higher depth *of that* to make a new net that at low depth can predict what *that* net would show at high depth. And then another which does that to *that one*, so there were high improvements just by training nets like this.
That's the main source of Fat Fritz 2's strength, all Albert Silver did was gather training resources to train a net and press the go button, and wait. And independent testers showed that his efforts weren't even better than the Stockfish's net at the time (you read this correctly, people would be better buying Fat Fritz 2 and throw its engine away and put the Stockfish version they used instead and get better analysis, which means Albert's efforts weakened the engine!)
The problem was that version couldn't be had by... the layman, that's why they had to accelerate the release of Stockfish 13 to show clear superiority (and nowadays Stockfish 14 leaves 13 in the dust, anyway.)
What Chessbase did was nothing wrong, I actually commend them on being the first because the Stockfish's license allows and entices it. The problem was THE WAY they did it. Claiming it was something new, whilst they only changed 12 lines of code (to accomodate Albert's efforts of doubling the net's size), and being very vague about the engine's origins, and withholding the net they sell commercially (while putting on the available source code a dummy net which is weaker and doesn't allow people to build Fat Fritz 2 on their own.)
It's basically plagiarism, all they needed to do was give full credit and accept all they're giving you is a different net for a free open source engine (anybody saying they also give GUI, bells and whistles forget they sell much cheaper products with the GUI, bells and whistles - basically people are much better grabbing Komodo Dragon at this point, because at least it's not something free yet.)
The irony - They didn't do it the right way because they feared it'd hurt sales. It's scammy to sell something pretending to be something else. But the scandal might end up killing their sales.
The irony of the irony - all publicity is good publicity, the bigger this gets, the more people are going to know they exist, they're gonna look them up, they're gonna see they sell stuff, and they're gonna buy. Maybe they're geniuses doing this on purpose so people know about them.