At AWS re:Invent this week, Amazon launched AWS AI Service Cards, a form of responsible AI documentation meant to help customers better understand the AI services offered by the cloud computing leader.
According to Gartner analyst Svetlana Sicular, the AWS AI Service Cards are a signal that Amazon is making moves to catch up with its competitors when it comes to embracing responsible AI.
In the past, Amazon “denied responsible AI” and fell behind its competitors, including Microsoft and Google, in addressing responsible AI issues, Sicular told VentureBeat.
But Amazon has “a very good history of catching up when they put their mind and the effort and resources to catching up,” she said. “They realized that this is serious and in everybody’s interest to have transparency, because otherwise they won’t be competitive.”
AWS AI Service Cards offer responsible AI repository
According to an Amazon blog post, the AWS AI Service Cards “provide customers with a single place to find information on the intended use cases and limitations, responsible AI design choices, and deployment and performance optimization best practices for our AI services.” The company said it addresses issues including fairness and bias, explainability, robustness, governance, transparency, privacy and security.
The first three service cards made available are for software dealing with sensitive demographic issues, including Amazon Textract – AnalyzeID, Amazon Transcribe – Batch (English-US).and Amazon Rekognition – Face Matching. The third card addresses issues related to the controversial tool which made headlines in 2019 when Amazon contested a study saying that the technology struggled to identify the gender of individuals with darker skin tones. After the 2020 murder of unarmed Black man George Floyd during an arrest, the company issued a moratorium on police use of its facial recognition software.
Now, the new AWS AI Service Card says it does not support matching “images that are too blurry and grainy for the face to be recognized by a human, or that have large portions of the face occluded by hair, hands, and other objects.”
Amazon’s recent steps towards responsible AI
Sicular pointed out that Amazon has been taking steps towards responsible AI in recent years, including hiring the authors of The Ethical Algorithm, Aaron Roth and Michael Kearns, as Amazon Scholars in 2020.
Kearns, a University of Pennsylvania professor — and a scholar at Amazon since 2020 — told Reuters that the decision to issue the cards followed privacy and fairness audits of the company’s software. He said the cards would address AI ethics concerns publicly, with tech regulation on the horizon.
“The biggest thing about this launch is the commitment to do this on an ongoing basis and an expanded basis,” he said.
The AWS AI Service Cards build on related efforts by Microsoft, Google and IBM. “The launch of the AWS AI Service Cards is a step in the right direction towards building responsible AI by design,” said Krishnaram Kenthapadi, chief scientist at ML observability company Fiddler AI. “This builds on a related line of work such as datasheets, model cards, factsheets, system cards, and ABOUT ML towards benchmarking and transparency of machine learning data sets, models and systems,” he said.
“They are very good first steps,” Sicular said, adding that she thinks Amazon is “probably going up the stack” in terms of reaching executive decision-makers, at a time when there is fierce competition among the big cloud providers for AI investment in the cloud.
“I think it’s part of the competition with Microsoft,” she explained. “But [Amazon] does put a lot of emphasis on responsibility [now]. That is very impressive.”
The post AWS AI Service Cards signal Amazon’s responsible AI catch-up appeared first on Venture Beat.