TfL's AI Tube Station experiment is amazing and slightly terrifying
- “the system could apparently identify up to 77 different ‘use cases’ – though only eleven were used during trial. This ranges from significant incidents, like fare evasion, crime and anti-social behaviour, all the way down to more trivial matters, like spilled drinks or even discarded newspapers.” ⤴️
- “in the “safeguarding” bucket of use-cases, the AI was programmed to alert staff if a person was sat on a bench for longer than ten minutes or if they were in the ticket hall for longer than 15 minutes, as it implies they may be lost or require help.” ⤴️
- “They instead trained the system to spot people with both arms raised in the air – because this is thought to be a “common behaviour” linked to acts of aggression” ⤴️
- “attempts by the AI to spot “aggressive behaviour”.” ⤴️
- “if they are faced with someone behaving violently, and are unable to reach their radio to call for help, the staff themselves can simply raise their arms to trigger an alert to their colleagues.” ⤴️
- “TfL manually scrubbed through and tagged “several hours” of CCTV footage to train the system to teach it what fare evasion looks like.” ⤴️
- “TfL had to go back and reconfigure it to avoid counting kids as fare evaders – and they did it by automatically disregarding anyone shorter than the ticket gates.” ⤴️
- “It could be possible for augmented reality software to guide cleaning and maintenance staff to the spill that needs mopping up or the lightbulb that needs changing. Or when you plan a journey on Google Maps, it could warn you to avoid changing at Stratford to avoid the West Ham fans upset after a six-nil loss” ⤴️
- “Need to build a business case to install a lift? Now you can get an exact count of the number of people passing through with wheelchairs, prams and over-sized luggage.” ⤴️
- “It would be trivial from a software (if not legal) perspective to train the cameras to identify, say, Israeli or Palestinian flags – or any other symbol you don’t like. The system could be used to surveil staff, and work them even harder, by literally keeping a by-the-second count of their idle time while on shift. And of course, the black-box AI training data could turn out the be flawed, perhaps unfairly or disproportionately identifying fare evaders with certain skin colours.” ⤴️
- “So this is just a thing that exists in the world now. And even if we wanted to stop it being used, to do so would be just as virtually impossible as inventing it felt in the first place.” ⤴️
- “the system could apparently identify up to 77 different ‘use cases’ – though only eleven were used during trial. This ranges from significant incidents, like fare evasion, crime and anti-social behaviour, all the way down to more trivial matters, like spilled drinks or even discarded newspapers.” ⤴️
- “in the “safeguarding” bucket of use-cases, the AI was programmed to alert staff if a person was sat on a bench for longer than ten minutes or if they were in the ticket hall for longer than 15 minutes, as it implies they may be lost or require help.” ⤴️
- “They instead trained the system to spot people with both arms raised in the air – because this is thought to be a “common behaviour” linked to acts of aggression” ⤴️
- “attempts by the AI to spot “aggressive behaviour”.” ⤴️
- “if they are faced with someone behaving violently, and are unable to reach their radio to call for help, the staff themselves can simply raise their arms to trigger an alert to their colleagues.” ⤴️
- “TfL manually scrubbed through and tagged “several hours” of CCTV footage to train the system to teach it what fare evasion looks like.” ⤴️
- “TfL had to go back and reconfigure it to avoid counting kids as fare evaders – and they did it by automatically disregarding anyone shorter than the ticket gates.” ⤴️
- “It could be possible for augmented reality software to guide cleaning and maintenance staff to the spill that needs mopping up or the lightbulb that needs changing. Or when you plan a journey on Google Maps, it could warn you to avoid changing at Stratford to avoid the West Ham fans upset after a six-nil loss” ⤴️
- “Need to build a business case to install a lift? Now you can get an exact count of the number of people passing through with wheelchairs, prams and over-sized luggage.” ⤴️
- “It would be trivial from a software (if not legal) perspective to train the cameras to identify, say, Israeli or Palestinian flags – or any other symbol you don’t like. The system could be used to surveil staff, and work them even harder, by literally keeping a by-the-second count of their idle time while on shift. And of course, the black-box AI training data could turn out the be flawed, perhaps unfairly or disproportionately identifying fare evaders with certain skin colours.” ⤴️
- “So this is just a thing that exists in the world now. And even if we wanted to stop it being used, to do so would be just as virtually impossible as inventing it felt in the first place.” ⤴️