According to Platformer, the entire team responsible for ensuring that Microsoft’s AI products ship with safeguards to mitigate social harm was cut during the company’s recent layoffs of 10,000 employees. I was.
The former employee said the Ethics and Society team is an important part of Microsoft’s strategy to mitigate the risks associated with using OpenAI technology in Microsoft products. Before AI died out, the team developed an entire “Responsible Innovation Toolkit” to help Microsoft engineers predict the harm AI could cause and mitigate those harms.
Platformer’s report comes just before OpenAI releases GPT-4, arguably the most powerful AI model ever. GPT-4 is already helping power Bing searches, Reuters reports.
In a statement provided to Ars, Microsoft said, “We remain committed to developing AI products and experiences in a safe and responsible manner, and we do so by investing in the people, processes, and partnerships that make this our priority.” We are doing
Calling the work of its Ethics and Society team “pioneers,” Microsoft said the company has focused on investing in and scaling its responsible AI office over the past six years. The office continues to operate alongside Microsoft’s other responsible AI working groups, the Aether Committee and Responsible AI Strategy in Engineering.
Emily Bender of the University of Washington, an expert in computational linguistics and ethical issues in natural language processing, joined other critics. Tweet Condemn Microsoft’s decision to disband the Ethics and Society team. Bender told Ars, as an outsider, that he considers Microsoft’s decision “shortsighted.” A drastic reduction in the to is terrible.”
Profile of the Ethics and Social Team
In 2017, Microsoft began to focus on a team dedicated to responsible AI research, CNBC reported. By 2020, that effort included an ethics and social team with up to 30 members, Platformer notes. But as the AI race with Google heats up, Microsoft began moving most of its ethics and social team members to specific product teams last October. As such, only seven people were dedicated to implementing the ethics and social team’s “ambitious plans,” an employee told the platformer.
It’s too much work for a small team, and Platformer says former team members may stop Microsoft from copying the branding of living artists, such as recommended mitigation strategies for Bing Image Creator. , reported that it said it did not always respond to their recommendations. (Microsoft disputes this claim, saying it changed the tool prior to release to address the team’s concerns.)
While the team was cut back last fall, Platformer said John Montgomery, corporate vice president of AI at Microsoft, said there was “a lot of pressure to get these latest OpenAI models, and those that followed, into the hands of customers. There was,” he said. at a very high speed. The employee warned Montgomery of “significant” concerns about the potential negative effects of this speed-based strategy, but Montgomery insisted that “the pressure remains the same.”
But even with the smaller ethics and social team, Microsoft said it wouldn’t eliminate the team. However, the company announced he changed on March 6th. The rest of the team were told in a Zoom meeting that it was considered “business critical” to disband the team entirely.
Bender told Ars that the decision was particularly disappointing because Microsoft “has been able to bring together a very talented group of people working on AI, ethics and the social impact of technology.” But Bender said of the move, Microsoft “basically” said that if companies perceive the ethics and social team’s recommendations as “going against what benefits us in the short term, they I think I have to leave.”
For professionals like Vendor, Microsoft is currently far too focused on funding a team dedicated to telling the company to slow down when its AI models may pose risks, including legal risks. You seem uninterested. An employee asked the platformer what would happen to both brands and users now that no one seemed to say ‘no’ when potentially irresponsible designs were pushed to users. Said he thought
“The worst thing is that we put our business at risk and we put people at risk,” one former employee told Platformer.
An unstable future for responsible AI
When the company relaunched Bing with AI, users quickly discovered that the Bing chat tool behaved unexpectedly. Historically, technology companies such as Microsoft and Google have been trusted to self-regulate the release of AI tools, identify risks and mitigate harm. But Bender, who co-authored the paper with former Google ethics AI researcher Timnit Gebru, whose criticism of the large-scale language models that many AI tools rely on resulted in Gebru’s dismissal, said that Ars “Self-regulation as a model isn’t going to work,” she said.
“You need external pressure,” Bender told Ars to invest in a responsible AI team.
Vendors are advocating that regulators get involved at this point if society wants more transparency from companies amid the “current wave of AI hype.” Otherwise, users risk jumping on the bandwagon and using popular tools without fully understanding how they can be harmed. For example, he has 100 million monthly active users, as he did with AI-powered Bing.
“I think anyone who encounters this should have a clear understanding of what they are dealing with,” Bender told Ars. “And I don’t see any company doing that well.”
Bender claims AI ‘will be as big and disruptive as the internet’ It’s ‘terrifying’ that companies seem so enamored with capitalizing on the AI hype said. Instead, companies have an obligation to “think what can go wrong.”
At Microsoft, that duty falls to the responsible AI office, a spokesperson told Ars.
A Microsoft spokesperson said, “We have also expanded the size and scope of the Office of Responsible AI. We provide cross-sectional support.”
A better solution for vendors than relying on companies like Microsoft to do the right thing is for society to advocate for regulation. “Establish and protect rights rather than micromanage specific technologies.”
Until proper regulations are in place, greater transparency about potential harm, and increased information literacy among users, Bender says users will not be able to seek “medical advice, legal advice, psychotherapy, or Never accept applications of other sensitive AI.”
“I think it’s very short-sighted,” Bender said of the current AI hype.
