With the 2024 Paris Olympics set to begin in July, a professor of computer science at MIT is convinced something else is already underway.
The creation of undetectable performance-enhancing drugs (PEDs) with the help of Artificial Intelligence (AI).
“I’m 100 percent sure that if they’re in that business (of doping), they’re using it,’’ said Manolis Kellis, the MIT professor who is a member of the computer science and artificial intelligence lab at the university. “If I were in the doping business, I would be crazy not to use generative AI right now.’’
Unlike traditional AI, which follows ‘predefined rules and patterns,’ generative AI creates ‘new and original content.” Content that could possibly include PEDs, according to Anne Carpenter, senior director of the imaging platform at Broad Institute of MIT and Harvard.
“I would say it’s practical now to attempt it,” Carpenter said. But she also said there are significant hurdles, suggesting AI is still developing as a potential tool for cheating in sports.
The use of AI for drug discovery is no longer a pipedream. The unanswered question: How soon might AI be embraced by athletes looking for new ways to cheat?
How would AI work to help athletes cheat?
The most feasible approach would be using generative AI to alter existing PEDs that trigger drug tests in a way that makes those drugs undetectable by current testing technology, according to Kellis, the MIT professor. He said it would be used to study molecular structure of the existing PEDs and determine what other molecules could be used to alter them.
He compared the process to what often happens after a pharmaceutical company comes out with a highly effective drug. Competitors attempt to create their own version of the drug by altering an atom or two to evade patents — just like AI would help alter the molecular structure of an existing PED just enough to evade detection by drug tests, Kellis said.
There is skepticism in the scientific community about whether AI is being used for pharmacological purposes in sports. Some of the reasons: No existing peer review of studies or research, the extensive testing required to prove safety and the focus on finding drugs for current incurable diseases.
But Lei Xie, a professor at Hunter College in New York who has used AI for the potential discovery of drugs for incurable diseases, said the process that would be used to alter existing PEDs is one reason he would not be surprised if it is happening now.
‘It is similar to drug repurposing (repositioning), which we have worked on for years,” Xie wrote to USA TODAY Sports by email.
Can AI be used to create PEDs?
Carpenter, the senior director of the imaging platform at Broad Institute of MIT and Harvard, said she sees the potential for AI to help create undetectable PEDs rather than relying on existing PEDs.
‘It’s not like this is futuristic technology,” she said.
But Carpenter estimated it would cost $1 billion and take 10 years to develop a PED with the required testing for FDA approval. Referring to the process of drug development, she said, “It’s not like you put data in one end and get drugs out the other side.’
But there is evidence indicating the process of drug discovery can be accelerated.
Alán Aspuru-Guzik, a professor of chemistry and computer science at the University of Toronto, helped lead a team in 2022 that in 30 days discovered a “lead candidate’’ for a potential liver cancer drug. The feat was hailed for enhanced speed in the development of drugs with the use of AI.
‘The issue about performance-enhancement is that unlike traditional drugs, the clinical trials would not be so easy to make happen,” Aspuru-Guzik wrote by email. ‘I would not recommend generating (and testing) new drugs without a fully developed clinical trial.
‘Having said so, yes, it may be possible for rogue agents to develop such drugs. Could they be not detectable by traditional tests? Sure.’
Can AI work against dopers?
WADA (World Anti-Doping Agency) has explored the use of AI as a tool to catch cheaters. Its use is inevitable, according to Dajiang Liu, director of artificial intelligence and biomedical informatics at the Penn State College of Medicine.
“More powerful AI algorithms will lead to drugs that are more difficult to be detected,’’ Liu wrote by email. ‘… As you may be aware, there is often a gap between the development of a new drug and testing procedures that can detect that. It is not surprising to me that such gap would happen to new AI-enabled drugs. At the same time, AI-driven technologies will also accelerate the development of testing procedures to identify drug use.’’
But that hasn’t stopped people in sports from moving forward with use of the technology, according to Aron D’Souza, an attorney and entrepreneur who’s trying to organize an international sports event where athletes will not be subject to drug testing. He said scientists and doctors involved in AI and PEDs have approached him about funding their projects.
Said D’Souza: ‘There will be many new performance-enhancing compounds discovered in the coming years.”