← AItopiaOrAImageddon?

The calculation

+1: every standing finding weighed by the share of the planet it reached, mean +1.46 per person-exposure over 90 entries, rounded. Worst standing loss −5 (Integrated Joint Operations Platform (Xinjiang)), best standing benefit +2 (Learned credit scorecards (Fair Isaac / FICO Score)). The rule is weighted, in SCALE.md, the owner's to strike.

All 90 standing: losses −69 over 49 entries, benefits +63 over 41, sum −6. Strike Integrated Joint Operations Platform (Xinjiang) and the level reads +1, set by Samagra Vedika (Telangana welfare entity resolution). Strike Learned credit scorecards (Fair Isaac / FICO Score) and the level reads +1, set by HNC/FICO Falcon payment-card fraud detection. 32 entries stand nowhere: no people count or no severity ruling (the impacts).

−80−60−40−20+20+40+60+800Standing losses, summedStanding benefits, summedThe levelEverything standing, summed1950: everything standing 0 = losses 0 over 0 + benefits 0 over 0; the level 01992: everything standing −3 = losses −3 over 2 + benefits 0 over 0; the level −11997: everything standing −5 = losses −5 over 3 + benefits 0 over 0; the level −11999: everything standing −7 = losses −7 over 4 + benefits 0 over 0; the level −12009: everything standing −8 = losses −10 over 7 + benefits +2 over 1; the level +22012: everything standing −6 = losses −10 over 7 + benefits +4 over 2; the level +22014: everything standing −7 = losses −11 over 8 + benefits +4 over 2; the level +22015: everything standing −2 = losses −13 over 10 + benefits +11 over 6; the level +22016: everything standing +3 = losses −14 over 11 + benefits +17 over 9; the level +22017: everything standing +1 = losses −17 over 12 + benefits +18 over 10; the level +22018: everything standing −3 = losses −22 over 15 + benefits +19 over 11; the level +22019: everything standing −16 = losses −40 over 22 + benefits +24 over 14; the level +22020: everything standing −11 = losses −41 over 23 + benefits +30 over 18; the level +22021: everything standing −12 = losses −45 over 27 + benefits +33 over 20; the level +22022: everything standing −7 = losses −44 over 28 + benefits +37 over 24; the level +22023: everything standing −2 = losses −49 over 32 + benefits +47 over 30; the level +22024: everything standing 0 = losses −52 over 35 + benefits +52 over 33; the level +22025: everything standing −4 = losses −63 over 45 + benefits +59 over 38; the level +12026: everything standing −6 = losses −69 over 49 + benefits +63 over 41; the level +119501960197019801990200020102020−80−60−40−200+20+40+60+80−6
Since 1950. Losses summed (red), benefits summed (blue), everything (the line), the level (amber, dashed). Hover a dot.
fromsideSRrungentry
2019lossS3R3−5Integrated Joint Operations Platform (Xinjiang)
2019lossS2R3−4Samagra Vedika (Telangana welfare entity resolution)
2017lossS3R1−3Integrated Joint Operations Platform (Xinjiang)
1992lossS2R1−2AT&T Voice Recognition Call Processing (HMM/word-spotting speech recognition)
1997lossS2R1−2USPS handwritten address recognition (learned word recognisers, CEDAR/SUNY Buffalo)
1999lossS2R1−2USPS Remote Computer Reader / OCR handwritten address recognition
2018lossS1R5−2Learned credit scorecards (Fair Isaac / FICO Score)
2018lossS3R0−2Uber ATG automated driving system
2019lossS3R0−2Meituan / Ele.me algorithmic dispatch and delivery-time estimation
2022lossS2R1−2naviHealth nH Predict (UnitedHealth)
2023lossS2R1−2Lavender (IDF target-marking system)
2023lossS3R0−2Police face recognition (Detroit PD, DataWorks Plus)
2025lossS2R1−2Generative AI cited as a reason for layoffs
2025lossS2R1−2Generative AI in fraud (deepfake video, voice cloning, synthetic text)
2026lossS3R0−2Police facial recognition (Clearview AI, DataWorks Plus and other vendors)
2026lossS1R5−2TikTok automated content moderation
1992lossS1R2−1IRS Discriminant Function (DIF) audit selection
2007lossS1R4−1Learned credit scorecards (Fair Isaac / FICO Score)
2007lossS1R3−1Credit-based insurance scores (auto insurance)
2009lossS2R0−1Automated licence plate reader optical plate recognition (multi-vendor)
2014lossS1R4−1Automated payment-card fraud scoring (issuer authorisation screening)
2015lossS1R2−1Computer-aided detection in screening mammography
2015lossS1R4−1Credit scorecards — the people the scorecard cannot see
2016lossS2R0−1COMPAS recidivism risk score
2018lossS1R1−1Shenzhen traffic-police facial recognition for jaywalking
2019lossS1R2−1AFR Locate (South Wales Police live facial recognition)
2019lossS2R3−1Kenyan digital-lender credit scoring (M-Shwari, Tala, Branch and others)
2019lossS1R0−1Brazilian state police facial recognition (Bahia, Rio de Janeiro, Santa Catarina, Paraíba, Ceará)
2020lossS1R3−1Meituan / Ele.me algorithmic dispatch and delivery-time estimation
2021lossS2R1−1Dutch childcare-benefits risk classification model
2021lossS2R0−1Epic Sepsis Model
2021lossS1R0−1ShotSpotter / SoundThinking gunshot detection (Chicago Police Department)
2021lossS1R4−1Learned credit scorecards (Fair Isaac / FICO Score)
2023lossS2R0−1Tesla Autopilot / Autosteer
2023lossS1R2−1Generative AI in writing-related freelance occupations
2024lossS2R0−1Deepfake video-conference fraud (Arup, Hong Kong)
2024lossS1R3−1Google Ads LLM-driven policy enforcement
2024lossS1R0−1Uber dynamic (upfront) pricing and work allocation
2025lossS1R0−1LLM coding assistants (Claude 3.5/3.7 in Cursor Pro)
2025lossS2R0−1Flock Safety automated licence plate reader network (Johnson County, Texas abortion search)
2025lossS1R2−1CBP Traveler Verification Service (biometric entry-exit facial comparison)
2025lossS1R3−1ChatGPT
2025lossS1R2−1ChatGPT
2025lossS1R3−1Metropolitan Police live facial recognition (NEC NeoFace Live)
2025lossS2R0−1Flock Safety automated licence plate reader network
2026lossS1R0−1Delhi Police facial recognition system
2026lossS2R0−1Flock Safety automated licence plate reader network
2019lossS1R30Facebook ad delivery optimisation
2025lossS0R70Spectre: the tax on a trainable branch predictor
2015benefitS1R4+1Netflix recommender system
2017benefitS2R0+1ICRISAT/Microsoft AI Sowing App
2018benefitS2R0+1IDx-DR autonomous diabetic retinopathy detection
2019benefitS2R1+1Deep-learning diabetic retinopathy screening in China's Metabolic Management Centres
2020benefitS2R1+1Qure.ai qXR chest X-ray AI for tuberculosis
2020benefitS2R1+1Novissi emergency cash transfers targeted by a phone-and-satellite poverty model (Togo)
2021benefitS2R0+1TREWS sepsis early warning system
2022benefitS1R0+1GitHub Copilot
2022benefitS2R0+1AI-supported mammography screening (MASAI trial, Transpara)
2022benefitS2R1+1Viz LVO stroke triage (Viz.ai)
2022benefitS3R0+1CAD4TB chest X-ray screening on ultra-portable X-ray, Nigeria
2023benefitS1R0+1ChatGPT (writing tasks, randomised experiment)
2023benefitS1R0+1Generative AI conversational assistant in a call centre
2024benefitS1R1+1Computer-aided polyp detection in colonoscopy (CADe)
2025benefitS2R0+1Rentosertib (ISM001-055), AI-discovered TNIK inhibitor
2025benefitS2R1+1Flock Safety automated licence plate reader network
2025benefitS1R0+1Metropolitan Police live facial recognition (NEC NeoFace Live)
2026benefitS2R0+1Waymo Driver
2026benefitS1R0+1Bahia (SSP-BA) facial recognition network
2007benefitS1R5+2Learned credit scorecards (Fair Isaac / FICO Score)
2012benefitS1R6+2HNC/FICO Falcon payment-card fraud detection
2015benefitS1R5+2eBay Machine Translation (eMT)
2015benefitS1R5+2Credit scorecards (FICO and successors)
2015benefitS2R3+2M-Shwari algorithmic credit scoring (Safaricom / CBA, Kenya)
2016benefitS1R5+2Google Translate (statistical machine translation)
2016benefitS1R5+2Google Translate
2016benefitS1R5+2Google Translate
2019benefitS1R6+2Learned email spam filtering (Gmail)
2019benefitS1R6+2Gmail spam, phishing and malware classification
2020benefitS1R6+2Google Maps ETA prediction (DeepMind graph neural networks)
2020benefitS2R2+2Infervision / Nankai CT screening for COVID-19 pneumonia
2021benefitS1R5+2Learned credit scorecards (Fair Isaac / FICO Score)
2023benefitS2R3+2AlphaFold 2 / AlphaFold Protein Structure Database
2023benefitS1R5+2ChatGPT
2023benefitS2R2+2ARDA (Google/Verily Automated Retinal Disease Assessment), India
2023benefitS2R4+2Kenyan digital-lender credit scoring (M-Shwari, Tala, Branch and others)
2024benefitS1R7+2Visa Advanced Authorization
2024benefitS1R5+2Google/DeepMind AI riverine flood forecasting (Flood Hub)
2025benefitS1R7+2Perceptron branch prediction in CPUs
2025benefitS1R5+2ChatGPT
2026benefitS1R5+2TikTok automated content moderation
Every entry standing in 2026, by rung. An entry stands from the later of its start and its count's date until a remedy ends it or the owner strikes it.
regionentriesworstbestweighted
US37−2Learned credit scorecards (Fair Isaac / FICO Score)+2Learned credit scorecards (Fair Isaac / FICO Score)+1.48
CN8−5Integrated Joint Operations Platform (Xinjiang)+2Infervision / Nankai CT screening for COVID-19 pneumonia+1.50
IN5−4Samagra Vedika (Telangana welfare entity resolution)+2ARDA (Google/Verily Automated Retinal Disease Assessment), India+1.50
UK4−1AFR Locate (South Wales Police live facial recognition)+1Metropolitan Police live facial recognition (NEC NeoFace Live)+1.50
other:Kenya3−1Kenyan digital-lender credit scoring (M-Shwari, Tala, Branch and others)+2M-Shwari algorithmic credit scoring (Safaricom / CBA, Kenya)+1.50
other:Brazil2−1Brazilian state police facial recognition (Bahia, Rio de Janeiro, Santa Catarina, Paraíba, Ceará)+1Bahia (SSP-BA) facial recognition network+1.50
EU2−2TikTok automated content moderation+2TikTok automated content moderation+1.48
other:Netherlands1−1Dutch childcare-benefits risk classification model−1Dutch childcare-benefits risk classification model+1.50
other:Palestine (Gaza)1−2Lavender (IDF target-marking system)−2Lavender (IDF target-marking system)+1.50
other:HK1−1Deepfake video-conference fraud (Arup, Hong Kong)−1Deepfake video-conference fraud (Arup, Hong Kong)+1.50
other:SE1+1AI-supported mammography screening (MASAI trial, Transpara)+1AI-supported mammography screening (MASAI trial, Transpara)+1.50
other:NG1+1CAD4TB chest X-ray screening on ultra-portable X-ray, Nigeria+1CAD4TB chest X-ray screening on ultra-portable X-ray, Nigeria+1.50
other:TG1+1Novissi emergency cash transfers targeted by a phone-and-satellite poverty model (Togo)+1Novissi emergency cash transfers targeted by a phone-and-satellite poverty model (Togo)+1.50
Each region on its own line: its own entries' worst and best, and the mean weighed by people over its own entries plus the world-wide ones. One country is not the world; a country's cage stays visible here.
yearlevelworstbestlossesnbenefitsnsum
195000000000
1992−1−20−3200−3
1997−1−20−5300−5
1999−1−20−7400−7
2007+2−2+2−96+21−7
2009+2−2+2−107+21−8
2012+2−2+2−107+42−6
2014+2−2+2−118+42−7
2015+2−2+2−1310+116−2
2016+2−2+2−1411+179+3
2017+2−3+2−1712+1810+1
2018+2−3+2−2215+1911−3
2019+2−5+2−4022+2414−16
2020+2−5+2−4123+3018−11
2021+2−5+2−4527+3320−12
2022+2−5+2−4428+3724−7
2023+2−5+2−4932+4730−2
2024+2−5+2−5235+52330
2025+1−5+2−6345+5938−4
2026+1−5+2−6949+6341−6
Every year something changed.