I want to use this opportunity to shill possibly the best history of science ever written: The Eighth Day of Creation [1], which describes the history of structural biology, including Watson’s various contributions. He comes across as a precocious asshole, not without talent but with a stronger eye towards self-advancement.
I am adjacent to the field, have read old perspectives, and have worked closely with some of that milieu's students, so that I have gotten my share of gossip from octogenerians who still pick sides in all of this. To spread some of that gossip, one opinion worth mentioning is that the only "real genius" among that group (including Franklin and Wilkins) was actually Crick, and that Watson was precocious but that his real brilliance was clinging to him. It's probably worth mentioning that being a 30 something doing a PhD seems to be a big advantage, though, especially if it's after a decade of doing physics research.
Edit: Watson is also personally responsible for convincing one of the most unethical and conniving scientists I know to go into science rather than medicine, so I have additional reasons to be suspicious, given assholes propagate assholes. If you're a Crick, for God's sake, stop taking pity, and don't tolerate Watsons even if you feel bad for them or they treat you in particular very well, have some standards and be a Stoner.
> Watson is also personally responsible for convincing one of the most unethical and conniving scientists I know to go into science rather than medicine
it is also of note that Watson repeatedly made public statements, starting in 2007, asserting that he believed Black people are inherently less intelligent than white people, attributing this to genetics, a claim broadly denounced by scientists and the public as racist and scientifically invalid.
That's a pretty epic title. And the cover art reminds me fondly of those textbooks from my past that were somehow extremely dry yet captivating at the same time.
Will check this out and see how it measures up to my favorite book on the topic, The Gene: An Intimate History [1]
He spoke at an event hosted by my company, once. He was pissed at Alec Baldwin, and devoted some time to calling him names. We were all looking at each other, going “WTF?”. He was supposed to be talking about using our microscopes, which never came up. It was a lot more like listening to Grandpa complaining about “kids these days,” after getting into the schnapps.
Have a friend that retired from CSH, a few years ago. Watson was a familiar presence, there; even after his Fall From Grace, which came about 20 years late. He used to live like a prince, on campus. Not sure if he was still there, before he went into hospice.
Most folks had a lot of difficulty with him, but he was a money magnet. They put up with his stuff, because he was such a good fundraiser.
It’s amazing how forgiving we can be, when money is to be made.
To my knowledge this connection was first noted in 2021 in https://arxiv.org/abs/2107.03006 (page 5). We wanted to do text diffusion where you’d corrupt words to semantically similar words (like “quick brown fox” -> “speedy black dog”) but kept finding that masking was easier for the model to uncover. Historically this goes back even further to https://arxiv.org/abs/1904.09324, which made a generative MLM without framing it in diffusion math.
It goes further back than that. In 2014, Li Yao et al (https://arxiv.org/abs/1409.0585) drew an equivalence between autoregressive (next token prediction, roughly) generative models and generative stochastic networks (denoising autoencoders, the predecessor to difussion models). They argued that the parallel sampling style correctly approximates sequential sampling.
In my own work circa 2016 I used this approach in Counterpoint by Convolution (https://arxiv.org/abs/1903.07227), where we in turn argued that despite being an approximation, it leads to better results. Sadly being dressed up as an application paper, we weren't able to draw enough attention to get those sweet diffusion citations.
Yeah, that's the first formal reference I remember as well (although, BERT is probably the first thing NLP folks will think of after reading about diffusion).
[1] https://www.cshlpress.com/default.tpl?cart=17625586661954464...