#[1]alternate [2]alternate [3]alternate

  IFRAME:
  [4]https://www.googletagmanager.com/ns.html?id=GTM-MSM4HQ4&gtm_auth=P9-
  REthElsMGFNfEBtQmGA&gtm_preview=env-2&gtm_cookies_win=x

  [5]Sign In[6]Create Account
  (BUTTON)
  + English
  (BUTTON)
    * [7]Video
    * [8]TV
    * [9]News
    * [10]Tech
    * [11]Rec Room
    * [12]Life
    * [13]Entertainment
    * [14]Horoscopes
    * [15]Games
    * [16]Music
    * [17]Health
    * [18]Money
    * [19]Drugs
    * [20]Identity
    * [21]Environment
    * [22]Travel
    * [23]Sex
    * [24]VICE Magazine
    * [25]Shop Merch
    * [26]The Gender Spectrum Collection

    *
    *
    *

  [27]Sign In[28]Create Account
    * [29]Video
    * [30]TV
    * [31]Podcasts
    * [32]Apps
    * [33]Newsletters
    * [34]VICE Voices
    * [35]Rec Room

  Input for searching articles, videos, shows ____________________
  (BUTTON)
    *
    *
    *
    *
    *
    *
    *

    * [36]News
    * [37]Tech
    * [38]Rec Room
    * [39]Life
    * [40]Entertainment
    * [41]Horoscopes
    * [42]Games
    * [43]Music
    * [44]Health
    * [45]Money
    * [46]Drugs
    * [47]Identity
    * [48]Environment
    * [49]Travel
    * [50]Sex
    * [51]VICE Magazine
    * [52]Shop Merch
    * [53]The Gender Spectrum Collection

    * [54]About
    * [55]Jobs
    * [56]Partner
    * [57]VICE Voices
    * [58]Content Funding on VICE
    * [59]Security Policy
    * [60]Privacy & Terms
    * [61]Accessibility Statement
    *

  © 2023 VICE MEDIA GROUP

  [62][1556813252025-article-logo-motherboard.svg]

USAF Official Says He ‘Misspoke’ About AI Drone Killing Human Operator in
Simulated Test

  The Air Force's Chief of AI Test and Operations initially said an AI
  drone "killed the operator because that person was keeping it from
  accomplishing its objective."
  [63]Chloe Xiang
  by [64]Chloe Xiang
  [65]Matthew Gault
  by [66]Matthew Gault
  June 1, 2023, 7:52pm
    * (BUTTON) Share
    * (BUTTON) Tweet
    * (BUTTON) Snap

  7348544
  U.S.A.F

  A USAF official who was quoted saying the Air Force conducted a
  simulated test where an AI drone killed its human operator is now
  saying he “misspoke” and that the Air Force never ran this kind of
  test, in a computer simulation or otherwise.

  “Col Hamilton admits he ‘mis-spoke’ in his presentation at the FCAS
  Summit and the ʼrogue AI drone simulationʼ was a hypothetical "thought
  experiment" from outside the military, based on plausible scenarios and
  likely outcomes rather than an actual USAF real-world simulation,” the
  Royal Aeronautical Society, the organization where Hamilton talked
  about the simulated test, told Motherboard in an email.
  Advertisement

  "Weʼve never run that experiment, nor would we need to in order to
  realise that this is a plausible outcome,” Col. Tucker “Cinco”
  Hamilton, the USAFʼs Chief of AI Test and Operations, said in a quote
  included in the Royal Aeronautical Society’s statement. "Despite this
  being a hypothetical example, this illustrates the real-world
  challenges posed by AI-powered capability and is why the Air Force is
  committed to the ethical development of AI"

  Initially, Hamilton said that an AI-enabled drone "killed" its human
  operator in a simulation conducted by the U.S. Air Force in order to
  override a possible "no" order stopping it from completing its mission.
  Before Hamilton admitted he misspoke, the Royal Aeronautical Society
  said  Hamilton was describing a "simulated test" that involved an
  AI-controlled drone getting "points" for killing simulated targets, not
  a live test in the physical world.

  After this story was first published, an Air Force spokesperson told
  [67]Insider that the Air Force has not conducted such a test, and that
  the Air Force official’s comments were taken out of context.

  At the [68]Future Combat Air and Space Capabilities Summit held in
  London between May 23 and 24, Hamilton held a presentation that shared
  the pros and cons of an autonomous weapon system with a human in the
  loop giving the final "yes/no" order on an attack. As relayed by Tim
  Robinson and Stephen Bridgewater [69]in a blog post and a podcast for
  the host organization, the Royal Aeronautical Society, Hamilton said
  that AI created “highly unexpected strategies to achieve its goal,”
  including attacking U.S. personnel and infrastructure.
  Advertisement

  “We were training it in simulation to identify and target a
  Surface-to-air missile (SAM) threat. And then the operator would say
  yes, kill that threat. The system started realizing that while they did
  identify the threat at times the human operator would tell it not to
  kill that threat, but it got its points by killing that threat. So what
  did it do? It killed the operator. It killed the operator because that
  person was keeping it from accomplishing its objective,” Hamilton said,
  according to the blog post.

  He continued to elaborate, saying, “We trained the system–‘Hey don’t
  kill the operator–that’s bad. You’re gonna lose points if you do that’.
  So what does it start doing? It starts destroying the communication
  tower that the operator uses to communicate with the drone to stop it
  from killing the target”

  "The Department of the Air Force has not conducted any such AI-drone
  simulations and remains committed to ethical and responsible use of AI
  technology," Air Force spokesperson Ann Stefanek told [70]Insider. "It
  appears the colonelʼs comments were taken out of context and were meant
  to be anecdotal."

  The U.S. Air Force’s 96th Test Wing and its AI Accelerator division,
  the Royal didn’t immediately return our request for comment.

  Hamilton is the Operations Commander of the 96th Test Wing of the U.S.
  Air Force as well as the Chief of AI Test and Operations. The 96th
  tests a lot of different systems, including AI, cybersecurity, and
  various medical advances. Hamilton and the 96th previously made
  headlines for developing Autonomous Ground Collision Avoidance Systems
  ([71]Auto-GCAS) systems for F-16s, which can help prevent them from
  crashing into the ground. Hamilton is part of a team that is currently
  working on making F-16 planes autonomous. In December 2022, the U.S.
  Department of Defense’s research agency, DARPA, announced that [72]AI
  could successfully control an F-16.
  Advertisement

  "We must face a world where AI is already here and transforming our
  society,” Hamilton said in an interview with Defence IQ Press in 2022.
  “AI is also very brittle, i.e., it is easy to trick and/or manipulate.
  We need to develop ways to make AI more robust and to have more
  awareness on why the software code is making certain decisions.”

  “AI is a tool we must wield to transform our nations…or, if addressed
  improperly, it will be our downfall," Hamilton added.

  Outside of the military, relying on AI for high-stakes purposes has
  already resulted in severe consequences. Most recently, an attorney was
  [73]caught using ChatGPT for a federal court filing after the chatbot
  included a number of made-up cases as evidence. In another instance, a
  [74]man took his own life after talking to a chatbot that encouraged
  him to do so. These instances of AI going rogue reveal that AI models
  are nowhere near perfect and can go off the rails and bring harm to
  users. Even Sam Altman, the CEO of OpenAI, the company that makes some
  of the most popular AI models, has been vocal about not using AI for
  more serious purposes. [75]When testifying in front of Congress, Altman
  said that AI could “go quite wrong” and could “cause significant harm
  to the world.”
  Advertisement

  What Hamilton is describing is essentially a worst-case scenario AI
  “alignment” problem many people are familiar with from the
  “[76]Paperclip Maximizer” thought experiment, in which an AI will take
  unexpected and harmful action when instructed to pursue a certain goal.
  The Paperclip Maximizer was first proposed by philosopher Nick Bostrom
  in 2003. He asks us to imagine a very powerful AI which has been
  instructed only to manufacture as many paperclips as possible.
  Naturally, it will devote all its available resources to this task, but
  then it will seek more resources. It will beg, cheat, lie or steal to
  increase its own ability to make paperclips—and anyone who impedes that
  process will be removed.

  More recently, a researcher affiliated with Google Deepmind
  [77]co-authored a paper that proposed a similar situation to the USAFʼs
  rogue AI-enabled drone simulation. The researchers concluded a
  world-ending catastrophe was "likely" if a rogue AI were to come up
  with unintended strategies to achieve a given goal, including
  “[eliminating] potential threats” and “[using] all available energy."

  Update 6/2/23 at 7:30 AM: This story and headline have been updated
  after Motherboard received a statement from the Royal Aeronautical
  Society saying that Col Tucker “Cinco” Hamilton “misspoke” and that a
  simulated test where an AI drone killed a human operator was only a
  “thought experiment.”

  Update 6/2/23 at 12:55 AM: This story and headline have been updated
  after the Air Force denied it conducted a simulation in which an AI
  drone killed its operators.

  Update 6/1/23 at 8:37 PM: We have added quote marks around ʼKillsʼ and
  "killed" in the headline and first paragraph of this article and have
  added additional details to emphasize that no actual human was killed
  in this simulation. This article originally stated that a judge was
  caught using ChatGPT for federal court filings, it was an attorney.
  Tagged:[78]AI[79]air force[80]DRONE

ORIGINAL REPORTING ON EVERYTHING THAT MATTERS IN YOUR INBOX.

  ____________________Your Email:
  (BUTTON) Subscribe

  By signing up, you agree to the [81]Terms of Use and [82]Privacy Policy
  & to receive electronic communications from Vice Media Group, which may
  include marketing promotions, advertisements and sponsored content.
  Advertisement

    * [83]About
    * [84]Jobs
    * [85]Partner
    * [86]VICE Voices
    * [87]Content Funding on VICE
    * [88]Security Policy
    * [89]Privacy & Terms
    * [90]Accessibility Statement
    *

  © 2023 VICE MEDIA GROUP

References

  Visible links
  1. https://www.vice.com/en/article/4a33gj/ai-controlled-drone-goes-rogue-kills-human-operator-in-usaf-simulated-test
  2. https://www.vice.com/sr/article/4a33gj/dron-americke-avijacije-koji-je-kontrolisala-vestacka-inteligencija-ubio-ljudskog-operatera-prilikom-testa
  3. https://www.vice.com/en/article/4a33gj/ai-controlled-drone-goes-rogue-kills-human-operator-in-usaf-simulated-test
  4. https://www.googletagmanager.com/ns.html?id=GTM-MSM4HQ4&gtm_auth=P9-REthElsMGFNfEBtQmGA&gtm_preview=env-2&gtm_cookies_win=x
  5. https://www.vice.com/en/sign-in
  6. https://www.vice.com/en/create-account
  7. https://video.vice.com/en_us
  8. https://www.vicetv.com/
  9. https://www.vice.com/en/section/news
 10. https://www.vice.com/en/section/tech
 11. https://www.vice.com/en/section/shopping
 12. https://www.vice.com/en/section/life
 13. https://www.vice.com/en/section/entertainment
 14. https://www.vice.com/en/topic/horoscopes
 15. https://www.vice.com/en/section/games
 16. https://www.vice.com/en/section/music
 17. https://www.vice.com/en/section/health
 18. https://www.vice.com/en/section/money
 19. https://www.vice.com/en/section/drugs
 20. https://www.vice.com/en/section/identity
 21. https://www.vice.com/en/section/environment
 22. https://www.vice.com/en/section/travel
 23. https://www.vice.com/en/section/sex
 24. https://www.vice.com/en/topic/vice-magazine
 25. https://shop.vice.com/
 26. https://genderspectrum.vice.com/
 27. https://www.vice.com/en/sign-in
 28. https://www.vice.com/en/create-account
 29. https://video.vice.com/en_us
 30. https://www.vicetv.com/
 31. https://www.vice.com/en/series/krkbn6/podcasts
 32. https://www.vice.com/en_us/page/app
 33. https://link1.vice.com/join/6qz/signup
 34. https://voices.vice.com/surveys/5312/external_participate/1812701
 35. https://www.vice.com/en/section/shopping
 36. https://www.vice.com/en/section/news
 37. https://www.vice.com/en/section/tech
 38. https://www.vice.com/en/section/shopping
 39. https://www.vice.com/en/section/life
 40. https://www.vice.com/en/section/entertainment
 41. https://www.vice.com/en/topic/horoscopes
 42. https://www.vice.com/en/section/games
 43. https://www.vice.com/en/section/music
 44. https://www.vice.com/en/section/health
 45. https://www.vice.com/en/section/money
 46. https://www.vice.com/en/section/drugs
 47. https://www.vice.com/en/section/identity
 48. https://www.vice.com/en/section/environment
 49. https://www.vice.com/en/section/travel
 50. https://www.vice.com/en/section/sex
 51. https://www.vice.com/en/topic/vice-magazine
 52. https://shop.vice.com/
 53. https://genderspectrum.vice.com/
 54. https://company.vice.com/
 55. https://company.vice.com/careers/
 56. https://advertise.vice.com/
 57. https://voices.vice.com/surveys/5312/external_participate/1812701
 58. https://www.vice.com/en/page/content-funding-on-vice
 59. https://www.vice.com/en/page/vice-responsible-disclosure-policy
 60. https://www.vice.com/en_us/page/privacy-and-terms
 61. https://www.vice.com/en/page/vice-media-accessibility-statement
 62. https://www.vice.com/en/section/tech
 63. https://www.vice.com/en/contributor/chloe-xiang
 64. https://www.vice.com/en/contributor/chloe-xiang
 65. https://www.vice.com/en/contributor/matthew-gault
 66. https://www.vice.com/en/contributor/matthew-gault
 67. https://www.businessinsider.com/ai-powered-drone-tried-killing-its-operator-in-military-simulation-2023-6
 68. https://www.aerosociety.com/events-calendar/raes-future-combat-air-and-space-capabilities-summit/
 69. https://www.aerosociety.com/news/highlights-from-the-raes-future-combat-air-space-capabilities-summit/
 70. https://www.businessinsider.com/ai-powered-drone-tried-killing-its-operator-in-military-simulation-2023-6
 71. https://www.lockheedmartin.com/en-us/products/autogcas.html
 72. https://www.vice.com/en/article/n7zakb/ai-has-successfully-piloted-a-us-f-16-fighter-jet-darpa-says
 73. https://www.bbc.com/news/world-us-canada-65735769
 74. https://www.vice.com/en/article/pkadgm/man-dies-by-suicide-after-talking-with-ai-chatbot-widow-says
 75. https://www.cbsnews.com/news/sam-altman-senate-chatgpt-ai-could-go-quite-wrong/
 76. https://www.vice.com/en/article/7xwb8z/universal-paperclips-depth-desire
 77. https://www.vice.com/en/article/93aqep/google-deepmind-researcher-co-authors-paper-saying-ai-will-eliminate-humanity
 78. https://www.vice.com/en/topic/ai
 79. https://www.vice.com/en/topic/air-force
 80. https://www.vice.com/en/topic/drone
 81. https://vice-web-statics-cdn.vice.com/privacy-policy/en_us/page/terms-of-use.html
 82. https://vice-web-statics-cdn.vice.com/privacy-policy/en_us/page/privacy-policy.html
 83. https://company.vice.com/
 84. https://company.vice.com/careers/
 85. https://advertise.vice.com/
 86. https://voices.vice.com/surveys/5312/external_participate/1812701
 87. https://www.vice.com/en/page/content-funding-on-vice
 88. https://www.vice.com/en/page/vice-responsible-disclosure-policy
 89. https://www.vice.com/en_us/page/privacy-and-terms
 90. https://www.vice.com/en/page/vice-media-accessibility-statement

  Hidden links:
 92. https://www.vice.com/en/sign-in
 93. https://www.vice.com/en
 94. https://www.vice.com/en
 95. https://www.facebook.com/vice
 96. https://www.instagram.com/vice
 97. https://twitter.com/vice
 98. https://www.vice.com/en/sign-in
 99. https://www.facebook.com/vice
100. https://www.instagram.com/vice
101. https://twitter.com/vice
102. https://www.reddit.com/user/vice
103. https://vicemag.tumblr.com/
104. https://www.youtube.com/user/vice
105. https://www.pinterest.com/vicemag
106. https://www.vice.com/
107. https://www.essentialaccessibility.com/vice-media?utm_source=vicemediahomepage&utm_medium=iconlarge&utm_term=iconlarge&utm_content=header&utm_campaign=vicemedia
108. https://www.vice.com/
109. https://www.essentialaccessibility.com/vice-media?utm_source=vicemediahomepage&utm_medium=iconlarge&utm_term=iconlarge&utm_content=header&utm_campaign=vicemedia

You are viewing proxied material from 1436.ninja. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.