Bläddra i källkod

Published Tue Mar 19 11:46:02 AM EDT 2024: AI kills

George Jones 1 månad sedan
förälder
incheckning
be85b07237
2 ändrade filer med 162 tillägg och 0 borttagningar
  1. 81 0
      home/public/blog/content/Blog/2024-03-19.md
  2. 81 0
      home/public/blog/curious.org

+ 81 - 0
home/public/blog/content/Blog/2024-03-19.md

@@ -0,0 +1,81 @@
++++
+title = "Insurance companies are using AI.  People will die."
+author = ["George M Jones"]
+publishDate = 2024-03-19T00:00:00-04:00
+lastmod = 2024-03-19T11:43:21-04:00
+tags = ["news", "philosophy", "history", "100DaysToOffload"]
+categories = ["blog"]
+draft = false
++++
+
+So, life.  Death.  Taxes.  Insurance companies and AI.
+
+This is one very real, present way AI is going to kill people.
+Not by HAL-9000 or Borg malevolence, but with a heavy dose of
+group-think by insurance companies.
+
+<!--more-->
+
+Life has brought me rather suddenly to the place that I'm interacting
+heavily with insurance companies about a medical issue for a loved
+one.
+
+The customer web site has a chat bot, and of course the
+"cut-every-direct-costs-ignore-the-consequences" insurance company
+pushes you hard to use it.   I did.   The output was semi-coherent,
+non-resposnive and tended to just spit out canned responses that you
+would find on the customer service web site (because, to be sure,
+that's what was trained on).
+
+I can read the web site.   When I call in, it's because I've already
+done that and can't get my question answered.   I need a person who
+understands things, can explain what I'm missing, and help me resolve
+non-standard problems.    But generally the people you get only know how to
+run a call script and maybe transfer you to other people who know how
+to run different call scripts.
+
+I got one of those who I think was a real human being, but, going
+foward with text to speach/speech to text all tied to AI on the back end,
+I'm pretty sure it's going to get to the point that you can ONLY
+interact with AI.  Yay, Turning test.
+
+So the person seemed to help me, seemed concerned, and promised to
+follow up on the issue I raised.   We even got into an email exchange.
+3 weeks later, no follow-up.   No resolution.
+
+I emailed back. The emails were all answered quickly, suspiciously
+quickly. It seemed to have no reference to the email exchange 3 weeks
+earlier, even though I had replied to the earlier message. It seemed
+unable to link ideas, reason or apply any common sense. I started getting
+the same snippits-from-web-site-docs replies. It told me that they
+both had and had not received something the doctor had sent them. Yup.
+AI again.
+
+It also helpfully recommended that I set up a 3 way call with their
+billing people and the doctor involved to resolve the billing issue.
+So, the doctor is asked to sit on hold for the benefit of the
+insurance company, not spending time helping patients with medical
+issues and, not billing valuable hours while waiting on hold with the
+insurance company.   Who's running this show?   For who's benefit?
+
+It would be interesting to see the results of some blind tests, A/B
+testing, etc. where customers got access to competent people to help
+them vs. only AI. I think we could measure the results in terms of
+mortality. That sort of testing would raise ethical issues (the value
+of a human life in terms of $).
+
+Aside from raising my blood pressure (stroke anyone?), it's going to
+delay needed care, give wrong/inconsistent answers (hallucinations
+etc), and cause caregivers, patients and medical professionals alike
+to waste time and figure out how to work around it.
+
+But I'm pretty sure what's going on is mid-to-upper level people inside
+insurance companies are just on the AI-hype bandwagon, driving it all
+the way down in their organizations, further dehumanizing customer
+service people before they lay them all off, and congratulating
+themselves on how forward looking they are and what good service they
+are providing.
+
+People are going to die.
+
+\#58 of #100DaysToOffload take 3.1, <https://100daystooffload.com/>

+ 81 - 0
home/public/blog/curious.org

@@ -4376,6 +4376,87 @@ at least /try/ to add those qualities to my own life.
 
 #13 of #100DaysToOffload take 2, https://100daystooffload.com/
 *** Current Events                                  :news:philosophy:history:
+**** WIP Insurance companies are using AI.  People will die. :100DaysToOffload:
+  :PROPERTIES:
+  :EXPORT_FILE_NAME: 2024-03-19
+  :EXPORT_HUGO_PUBLISHDATE: 2024-03-19:
+  :END:
+
+
+So, life.  Death.  Taxes.  Insurance companies and AI.
+
+This is one very real, present way AI is going to kill people.
+Not by HAL-9000 or Borg malevolence, but with a heavy dose of
+group-think by insurance companies.
+
+#+hugo: more
+
+Life has brought me rather suddenly to the place that I'm interacting
+heavily with insurance companies about a medical issue for a loved
+one.
+
+The customer web site has a chat bot, and of course the
+"cut-every-direct-costs-ignore-the-consequences" insurance company
+pushes you hard to use it.   I did.   The output was semi-coherent,
+non-resposnive and tended to just spit out canned responses that you
+would find on the customer service web site (because, to be sure,
+that's what was trained on).
+
+I can read the web site.   When I call in, it's because I've already
+done that and can't get my question answered.   I need a person who
+understands things, can explain what I'm missing, and help me resolve
+non-standard problems.    But generally the people you get only know how to
+run a call script and maybe transfer you to other people who know how
+to run different call scripts.
+
+I got one of those who I think was a real human being, but, going
+foward with text to speach/speech to text all tied to AI on the back end,
+I'm pretty sure it's going to get to the point that you can ONLY
+interact with AI.  Yay, Turning test.
+
+So the person seemed to help me, seemed concerned, and promised to
+follow up on the issue I raised.   We even got into an email exchange.
+3 weeks later, no follow-up.   No resolution.
+
+I emailed back. The emails were all answered quickly, suspiciously
+quickly. It seemed to have no reference to the email exchange 3 weeks
+earlier, even though I had replied to the earlier message. It seemed
+unable to link ideas, reason or apply any common sense. I started getting
+the same snippits-from-web-site-docs replies. It told me that they
+both had and had not received something the doctor had sent them. Yup.
+AI again.
+
+It also helpfully recommended that I set up a 3 way call with their
+billing people and the doctor involved to resolve the billing issue.
+So, the doctor is asked to sit on hold for the benefit of the
+insurance company, not spending time helping patients with medical
+issues and, not billing valuable hours while waiting on hold with the
+insurance company.   Who's running this show?   For who's benefit?
+
+It would be interesting to see the results of some blind tests, A/B
+testing, etc. where customers got access to competent people to help
+them vs. only AI. I think we could measure the results in terms of
+mortality. That sort of testing would raise ethical issues (the value
+of a human life in terms of $).
+
+Aside from raising my blood pressure (stroke anyone?), it's going to
+delay needed care, give wrong/inconsistent answers (hallucinations
+etc), and cause caregivers, patients and medical professionals alike
+to waste time and figure out how to work around it.
+
+
+But I'm pretty sure what's going on is mid-to-upper level people inside
+insurance companies are just on the AI-hype bandwagon, driving it all
+the way down in their organizations, further dehumanizing customer
+service people before they lay them all off, and congratulating
+themselves on how forward looking they are and what good service they
+are providing.
+
+People are going to die.
+
+#58 of #100DaysToOffload take 3.1, https://100daystooffload.com/
+
+
 **** PUBLISHED Be kind to your call center friends    :100DaysToOffload:life:
   :PROPERTIES:
   :EXPORT_FILE_NAME: 2024-03-02