I think people would be much more inclined to trust self driving tech in these scenarios than in congested traffic with complex intersections and people unexpectedly darting this way and that with little to no warning.
144 karma · joined July 16, 2017
I think people would be much more inclined to trust self driving tech in these scenarios than in congested traffic with complex intersections and people unexpectedly darting this way and that with little to no warning.
I understand that this is a pretty cynical view, and to be fair all the Googlers I've spoken with are true believers in their tech (perhaps a little too much). But don't underestimate the power of ego and financial incentives to constrain ethical/moral calculus.
So what about an algorithm that has been given the regulatory green light and fails under some edge case or adversarial attack? How about something like the Tesla situation, where they advertise "self-driving" capabilities while tucking in some liability limiting fine print about how the driver should be alert and ready to take over control at all times?
Both the MDs and the insurance companies are rife with a number of rotten incentives and bad actors that contribute to expensive medical care without quality commensurate with costs.
In fact, the entire policy debate over risk-based vs fee-for-service payment models is essentially the payers and providers arguing over who gets to rip you off.
Last time I checked, the NY Times was still publishing as the NY Times and placing their writers' names on the byline. When someone catches them printing up counterfeit copies of the Wall Street Journal designed to discredit we can talk about some sort of equivalence.
Whether something is white (e.g. RT, Voice of America, etc), grey (e.g. swift boat vets for truth, astroturfed citizens for X), or black hat propaganda (e.g. Black Matters, @TEN_GOP) depends on whether you represent yourself as (a) yourself, (b) a neutral party, or (c) your adversary.
So yeah, the problem is one of policing identity rather than messages. In other words, how do we make it more difficult for people to get catfished?
This is pretty much exactly the plot of Robocop. https://en.wikipedia.org/wiki/RoboCop
100% agree. I think this can be further broken down into the questions:
(1) What are the problems caused by extreme gerrymandering?
(2) How will efficiency gap minimization alleviate these problems?
(3) What assumptions are we making in (2)? How likely are they to be true? And how robust is the solution to their falsification?
(4) How could this all go catastrophically wrong?
My intuition is that efficiency gap minimization is an incomplete solution and probably should be only one of several factors considered.
Positive:
- Adds the voice of a civic minded, non-profit entity to the board. Maybe less brutal than pure capitalism?
- Less risk of Frankenstein's monster (e.g. dual use adware)
Negative:
- Govt is now married to them, and the tech graveyards are littered with the corpses of behemoths who lost their edge.
Same thing applies here. In one case the questions are being asked to provide treatment the patient is paying for. In the other it's being done to collect data for a publication.
Putting a patient at risk to treat them (e.g. chemotherapy) is very different than putting them at risk to get data for a publication or NDA. There's no reason what's tolerable in one context should be in the other.
(B) Is the lack of risk really self-evident? To everyone? Or just to you? Thus the question who should decide? And by what process do they decide?
The question is who decides and how they make the decision? Surely it should involve someone other than the investigator, and have clearly defined rules so people know what to expect. Hence the IRB.
https://en.wikipedia.org/wiki/Belmont_Report
(B) Is experimentation on suicidal patients really minimal risk? Maybe, maybe not. But failure of the investigator to show they have thought about it is a huge red flag.
(1) The study title doesn't work at the top of the form? Change it! Surely there is some other arrangement of English words that would satisfy both the author and IRB.
(2) They leave the risk section blank because minimal risk, and then complain about the requirement for pens because the patients might stab themselves if given pens. Looks to me like they didn't spend much effort thinking about risks to the patient. If they had just mentioned the pen stabby thing as a risk factor, and given "don't use pens" as a risk minimization plan, they wouldn't have had a problem.
(3) They poopoo the training because Hitler, and then get upset when they unknowingly violate a personnel rule that is most definitely covered in any decent IRB/GCP training.
I could go on, but every new thing I enumerate makes me angrier. Life is too short. Look. IRB applications suck, but human experimentation is serious business.
The rules aren't there to make the study easy, the are there to (a) look out for the patient (respect), (b) evaluate risk and benefits of the study (beneficence), and (c) ensure that the benefits are distributed to the group bearing the risk (justice).