Read on Twitter
Remember when @lexfridman authored a study showing that Tesla Autopilot didn't negatively impact the "functional vigilance" of drivers? The one Elon touted in this tweet. Weirdly, that study seems to have disappeared from any website bearing MIT's name. 
https://twitter.com/elonmusk/status/1114168824599842817
https://twitter.com/elonmusk/status/1114168824599842817
But wait, it gets even weirder: the paper in question, entitled "The Human Side of Tesla Autopilot" isn't even listed at http://lexfridman.com ! The author's own website! It's one thing if the paper doesn't meet MIT's standards, but they don't appear to even meet Lex's
This actually isn't that surprising. I'm no academic, but I found a wide variety of issues with the paper.
I wrote this @thedrive: https://www.thedrive.com/tech/27359/mit-study-showing-high-functional-vigilance-among-autopilot-users-comes-with-massive-caveats
I recorded this with @TheAutonocast: http://www.autonocast.com/blog/2019/4/18/140-unpacking-the-mit-autopilot-study
And this, with @missy_cummings: https://soundcloud.com/user-155427377/episode-2-dr-missy-cummings-on
I wrote this @thedrive: https://www.thedrive.com/tech/27359/mit-study-showing-high-functional-vigilance-among-autopilot-users-comes-with-massive-caveats
I recorded this with @TheAutonocast: http://www.autonocast.com/blog/2019/4/18/140-unpacking-the-mit-autopilot-study
And this, with @missy_cummings: https://soundcloud.com/user-155427377/episode-2-dr-missy-cummings-on
I wasn't alone in this criticism. Others, most notably @missy_cummings blasted the study... and Fridman's potential conflicts of interests. There was even a rumor going around that these included Fridman's (apparently unsuccessful) efforts to be hired by Tesla around this time.
I want to be clear: I don't know the whole story here, although I'd love it if a journalist tracked it down. The reason it matters is that this study was so uncritically reported on. BusinessInsider, Bloomberg, Electrek and others all have stories citing it like it's science.
Here's the real reason to dig into this: after apparently removing Fridman from the research, actual MIT researchers found the exact opposite of what he'd found! Actually, Autopilot does increase distraction!
Paper: https://dl.acm.org/doi/10.1145/3409120.3410644
Reporting: https://www.cnn.com/2020/09/24/cars/tesla-autopilot-study/index.html
Paper: https://dl.acm.org/doi/10.1145/3409120.3410644
Reporting: https://www.cnn.com/2020/09/24/cars/tesla-autopilot-study/index.html
There are two more reasons to investigate this: first, this paper is relevant to policy and regulation. NHTSA says that ADAS that contribute to "predictable abuse" may be deemed defective and recalled. Fridmans' paper is cited as "proof" AP doesn't do so. https://www.federalregister.gov/documents/2016/09/23/2016-23010/nhtsa-enforcement-guidance-bulletin-2016-02-safety-related-defects-and-automated-safety-technologies
The second is that this paper is just one utterly debunked "study" showing positive safety impacts from AP use. Most notably, the NHTSA study saying AP reduced crashes by 40%: https://www.thedrive.com/tech/26455/nhtsas-flawed-autopilot-safety-study-unmasked
But also, every other "data point" Musk has touted: https://arstechnica.com/cars/2018/05/sorry-elon-musk-theres-no-clear-evidence-autopilot-saves-lives/
But also, every other "data point" Musk has touted: https://arstechnica.com/cars/2018/05/sorry-elon-musk-theres-no-clear-evidence-autopilot-saves-lives/
This long pattern of utterly misleading "studies" and "data," from Fridman, NHTSA, and Tesla itself suggests exactly the kind of influence campaign that Musk regularly suggests the oil industry (among other enemies) engages in to attack Tesla. Maybe it's all a big coincidence...
But when every positive "study" and "data point" you can find has been debunked, and turn out to have emerged from such mysterious and suspicious circumstances, that's worth looking into. If any of my implications here turn out to be wrong, I'm happy to retract and apologize.
Thanks for making the time for this trip down memory lane. I'd ask Lex about this myself, but he blocked me (and others, including @AlexRoy144 and Anima Anandkumar) as soon as I started to criticize the paper in question. Another suspicious data point right there, IMO.
~FIN~
~FIN~
