Wada wants AI to catch drug cheats

Tech­nol­ogy could be used within two years Fo­cus on iden­ti­fy­ing sus­pi­cious pat­terns

The Daily Telegraph - Sport - - Athletics - By Tim Wig­more and Ben Rumsby in Ka­tow­ice

The World Anti-dop­ing Agency hopes to roll out ar­ti­fi­cial in­tel­li­gence tech­nol­ogy to help catch drug cheats within two years, The Daily Tele­graph can re­veal.

Wada be­lieves that us­ing AI – com­puter sys­tems that con­duct tasks nor­mally re­quir­ing hu­man in­tel­li­gence – could help tar­get tests far more ef­fec­tively and ul­ti­mately trap more of­fend­ers.

More than 300,000 sam­ples a year are an­a­lysed world­wide, but there re­main huge con­cerns about the ef­fi­cacy of test­ing.

Ar­ti­fi­cial in­tel­li­gence could be cru­cial in analysing data more ef­fec­tively, lead­ing to quicker iden­ti­fi­ca­tion of sus­pi­cious pat­terns and there­fore bet­ter-tar­geted test­ing.

The Rus­sian dop­ing scan­dal, and the re­cent case of Amer­i­can world cham­pion sprinter Chris­tian Cole­man – who com­mit­ted three where­abouts vi­o­la­tions in 12 months, but was cleared on a tech­ni­cal­ity – have high­lighted con­cerns with the cur­rent anti-dop­ing sys­tem. Prom­i­nent anti-dop­ing sci­en­tist Don Catlin has said that fewer than 10 per cent of dop­ers are caught.

“Ar­ti­fi­cial in­tel­li­gence of­fers you one big ben­e­fit: us­ing a big batch of data that the hu­man brain can­not process be­cause it’s too much for hu­man ca­pac­ity,” Dr Olivier Rabin, Wada’s di­rec­tor of science, told The Tele­graph. “We want to use ar­ti­fi­cial in­tel­li­gence for im­prov­ing our tar­get­ing ca­pac­ity. Iden­ti­fy­ing sus­pi­cious cor­re­la­tions could be an in­di­ca­tion of doped ath­letes.”

Last month, Wada an­nounced that it was fund­ing three projects to ex­plore the use of AI in dop­ing, all un­der­taken by in­de­pen­dent re­searchers. “Ath­letes are smart in how they dope, you need to be smart in how you ap­ply anti-dop­ing tests,” Rabin said.

“AI will help us in the recommenda­tion to test this par­tic­u­lar ath­lete at this par­tic­u­lar time to at­tempt to re­veal dop­ing.”

Should the re­search projects prove as suc­cess­ful as hoped, Rabin says AI could be used within 18-24 months. “It’s go­ing to help us fo­cus on the sus­pi­cious data and the sus­pi­cious ath­letes’ pro­files,” he added.

Tar­geted test­ing is far more ef­fec­tive than ran­dom test­ing, but Wada still be­lieves that not enough tests are suf­fi­ciently fo­cused. Us­ing ar­ti­fi­cial in­tel­li­gence could help spot new pat­terns typ­i­cal of dop­ers, as­sist­ing its in­ves­ti­ga­tions.

“We want to see the cor­re­la­tions that are un­known to us,” Rabin said. “There is a cor­re­la­tion that we could find be­tween some of the as­pects of the sus­pi­cious pro­file and be­tween the pop­u­la­tion of some of the ath­letes that we could iden­tify as be­ing doped. Re­veal­ing those cor­re­la­tions with AI is some­thing we would like to au­to­mat­i­cally ap­ply in our sys­tem, once val­i­dated.”

Pat­terns typ­i­cal of dop­ers could be re­vealed by AI be­ing able to an­a­lyse large sets of data quicker than hu­man be­ings, Rabin said. “You’re go­ing to look at the blood pro­file of the ath­letes over the past few years and whether the ath­lete has made a habit of chang­ing his or her where­abouts shortly be­fore the manda­tory hour for test­ing. There are dif­fer­ent sets of data you can com­bine and cor­re­late. If you have a strong cor­re­la­tion re­vealed by AI, then all the ath­letes who have those pat­terns in their pro­file, you should aim at test­ing them two hours be­fore the manda­tory win­dow.”

Rather than us­ing AI as suf­fi­cient proof to catch dop­ers, it would more likely be used to fo­cus drug test­ing on sus­pi­cious ath­letes at the times they were most likely to dope. This would build on work done by the ath­lete bi­o­log­i­cal pass­port, in­tro­duced a decade ago.

Mean­while, ge­netic drug test­ing could be in­tro­duced to sport in time for next year’s Olympics, the pres­i­dent of the In­ter­na­tional Olympic Com­mit­tee said yes­ter­day. Two months af­ter The Tele­graph re­vealed a break­through was close, Thomas Bach said it may be ready for Tokyo 2020.

He also con­firmed it could be ac­com­pa­nied by dried blood spot test­ing (DBS), pend­ing Wada ap­proval. “This new ap­proach could be a ground-break­ing method to de­tect blood dop­ing, weeks or even months af­ter it took place,” Bach said in a speech at the start of the World Con­fer­ence on Dop­ing in Sport in Ka­tow­ice. “If ap­proved by Wada, such new gene test­ing could be used al­ready at Tokyo 2020.”

‘AI can use a big batch of data that the hu­man brain can­not process’

Cleared: Chris­tian Cole­man had com­mit­ted three where­abouts vi­o­la­tions in 12 months

Newspapers in English

Newspapers from UK

© PressReader. All rights reserved.