An article I discussed earlier in the year in this thread - this feels much longer ago than 11 months, has had what has been called a “damning” assessment of its lack of transparency and the existential threat this creates to AI as a research field. Attached is the commentary from Nature. s41586-020-2766-y.pdf (1.16 MB)
I’m not particularly persuaded by the authors’ defensive reply to the criticism. 1. It’s not clear whether they sought permission from the hospital IRB to share further, even if only with the critics, nor is is clear why doing so would necessarily change the balance of risk v benefit (a bald assertion of the same without appropriate justification is insufficient); 2 hiding behind the unresolved liability issues is weak given there are legal options available to mitigate them, especially when it does not appear to have been used in the wild yet 3. it is a pity they are not interested in collaboration.