Skip to Content
logologo
AI Incident Database
Open TwitterOpen RSS FeedOpen FacebookOpen LinkedInOpen GitHub
Open Menu
Discover
Submit
  • Welcome to the AIID
  • Discover Incidents
  • Spatial View
  • Table View
  • List view
  • Entities
  • Taxonomies
  • Submit Incident Reports
  • Submission Leaderboard
  • Blog
  • AI News Digest
  • Risk Checklists
  • Random Incident
  • Sign Up
Collapse
Discover
Submit
  • Welcome to the AIID
  • Discover Incidents
  • Spatial View
  • Table View
  • List view
  • Entities
  • Taxonomies
  • Submit Incident Reports
  • Submission Leaderboard
  • Blog
  • AI News Digest
  • Risk Checklists
  • Random Incident
  • Sign Up
Collapse

Report 732

Associated Incidents

Incident 4127 Report
All Image Captions Produced are Violent

Loading...
MIT researchers use Reddit to create the first "psychopath AI"
cnet.com · 2018

MIT Media Lab

Meet Norman.

He's not your everyday AI. His algorithms won't help filter through your Facebook feed or recommend you new songs to listen to on Spotify.

Nope -- Norman is a "psychopath AI", created by researchers at the MIT Media Lab as a "case study on the dangers of artificial intelligence gone wrong when biased data is used in machine learning algorithms."

The researchers set Norman up to perform image captioning, a deep-learning method that generates a textual description of an image, then plugged him into an unnamed subreddit known for its graphic imagery surrounding death.

Then they had Norman explain a range of Rorschach inkblots, comparing the answers from their psychopathic AI with that of your friendly, neighbourhood "standard AI". Although Norman was originally unveiled on April 1, those answers are no joke -- they're highly disturbing.

Where a standard AI sees "a group of birds sitting on top of a tree branch" (awww!), Norman, our HAL-9000-esque death-machine, sees "a man electrocuted to death" (ahhh!). Where the standard AI sees "a close up of a wedding cake on a table", Norman, our malicious AI robokiller sees "a man killed by speeding driver".

The researchers didn't "create" Norman's "psychopathic" tendencies, they just helped the AI on its way by only allowing it to see a particular subset of image captions. The way Norman describes the Rorschach inkblots with simple statements does make it seem like it is posting on a subreddit.

But why even create a psychopath AI?

The research team aimed to highlight the dangers of feeding specific data into an algorithm and how that may bias or influence its behaviour.

That starts to make me wonder -- don't the MIT team at Boston Dynamics constantly push and poke and annoy their running, jumping and door-opening robot creations?

Are we doomed to be overrun by four-legged robo-hell-beasts? Let's hope not.

Tech Enabled: CNET chronicles tech's role in providing new kinds of accessibility.

Blockchain Decoded: CNET looks at the tech powering bitcoin -- and soon, too, a myriad of services that will change your life.

Read the Source

Research

  • Defining an “AI Incident”
  • Defining an “AI Incident Response”
  • Database Roadmap
  • Related Work
  • Download Complete Database

Project and Community

  • About
  • Contact and Follow
  • Apps and Summaries
  • Editor’s Guide

Incidents

  • All Incidents in List Form
  • Flagged Incidents
  • Submission Queue
  • Classifications View
  • Taxonomies

2024 - AI Incident Database

  • Terms of use
  • Privacy Policy
  • Open twitterOpen githubOpen rssOpen facebookOpen linkedin
  • e1b50cd