The Salesforce AI ethics team faces challenges with the cloud platform provider

How the Salesforce ethics team affects the development of AI

When Salesforce engineers incorporated tools that use natural-language processing and machine learning to evaluate people’s feelings on products or brands on social media or reviews, they discovered a problem. The system automatically labels content that features words such as gay, Black or Muslim as negative. It even happens if the statements around them are positive, as in, “I’m gay and proud.”

Kathy Baxter and her Ethics by Design team within Salesforce entered to assess the algorithmic tool, but despite the potential for discrimination, some product management decision makers still want the company to still launch the product in beta to get customer feedback.

• In the end, the ethics team won. “We said no – we’ll have to wait here,” said Baxter, Salesforce’s ethical chief AI architect. “And so the team held back.”
• This is a rare example of the actual power for tech ethics watchdogs in product decisions that can affect countless people.
• Ultimately, Baxter’s ethical design team worked with engineers to perform bias mitigation on machine-learning models of the product. They tested and watched how those changes had an impact before finally launching the tool as part of its Einstein software in 2017, which is now part of the company’s Tableau CRM software.

Some of the companies are subject to the most intense scrutiny of their AI skills say they have an embedded ethics in product development.

• Facebook parent Meta has an interdisciplinary responsible AI group that works with its product teams to address fairness in its AI-enabled products.
• Microsoft’s responsible AI office leads company-wide management processes, while another group there is dedicated to integrating the company’s responsible AI rules into engineering work.
• Google has notoriously shut down its AI ethics committee following criticism of its options for committee members. The company also criticized the AI ​​ethics community after removing one of the most respected AI ethics researchers in the industry. But now it still has a Responsible AI and Human-Centered Technology group that works with products and engineering teams.

But for many companies, implementing AI principles is not uncommon, and many are looking to the government to create regulations to guide that process.

• Some ethics duties are within a legal compliance team or are relegated to a monthly privacy management or check-in.
• However, Baxter’s team has a surprising amount of day-to-day interaction and input into decisions made throughout the organization.
• When Salesforce created templates in its Einstein Discovery software for predictive analytics, now also part of Tableau CRM, his team convinced engineers to remove data fields that display zip codes. Neighborhood-level data can be a proxy for race, and therefore threatens to enable racial bias in how the system makes predictions.
• “Although from a legal perspective, it’s okay maybe, from an ethical perspective, we said no. And so the team had to delete those fields, ”she said.

However, Salesforce is not just what Salesforce has developed. Like many cloud platforms, Salesforce is a thick forest of software applications and tools.

• The company’s AppExchange hosts applications developed by other companies that do anything from providing automated financial planning advice to managing patients with addictions.
• Salesforce is slowly encouraging tech partners toward more ethical practices. For example, Baxter authored a paper published last year that outlined steps organizations must take to address ethical questions during the AI ​​development and implementation stages.
• The company also offers a course for Responsibly Creating Artificial Intelligence for Salesforce users. But that educational module only includes four short classes lasting 45 minutes in total.

Education is the main tool of the company for the development of more ethical practices with partners in its vast app marketplace. He also said in 2022 the company’s account reps are expected to offer customers more AI ethics education and services.

So far, “it’s a good-there, it’s a pretty rare thing” to have a staff that is ethically committed to developing products or services, Baxter says. “But in a few years, it would be madness not to have even one of these in your company.”

– Kate Kaye (email | kaba)

A MESSAGE FROM DATAIKU

Dataiku is the only AI platform that connects data and makers, allowing anyone to turn data into real business results-from mediocre to lunar. Because AI can do so much, but there is no soul in the machine, just in front of it. If you don’t, it’s just data.

Learn more

Intel’s foundry plans include accepting RISC

As part of Intel’s bid to regain its one -time leadership position in chip making, the company announced a series of measures on Monday aimed at supporting its contract chip business, Intel Foundry Services. Intel said it has launched an “ecosystem alliance” called Accelerator designed to help companies bring new chips to market, and a $ 1 billion fund to support companies building tech to support the foundry ecosystem. .

Intel said the funding will focus on helping tech -interested companies around modular products, and support an open chiplet platform – adding several different types of chips to a package, as opposed to -cramming a bunch of designs on one chip.

“Foundry customers are quickly embracing a modular design approach to diversify their products and speed up time to market,” CEO Pat Gelsinger said in a statement.

To support the company’s efforts, Intel said it will support designs that use multiple instruction set architectures, or ISAs, including x86, Arm and RISC-V. The company said an important part of its contract chip strategy is a wide range of tech offerings, including RISC-V. Intel said it has joined RISC-V International, the consortium that oversees the RISC-V open-source architecture.

– Max A. Cherney (email | kaba)

Coming to Protocol

The cloud allows chief information officers to be more strategic and focused on generating business value rather than managing IT as a cost center. The pandemic then pushed CIOs into a more central role: They became essential leaders when managing remote work, navigating sudden changes in business operations and accelerating digital change.

So what’s next for CIOs? Join Tom Krazit of Protocol for a virtual event tomorrow at 10 am PT in discussion with four amazing IT leaders: Rob Carter, CIO, FedEx; Chris Bedi, CIO, ServiceNow; Sheila Jordan, Chief Digital Technology Officer, Honeywell; and Vittorio Cretella, CIO Procter & Gamble. RSVP here.

Financial corner

Flexport raised $ 935 million from Shopify and other investors in a worth $ 8 billion for its global logistics platform.

Chargebee then worth $ 3.5 billion after raising $ 250 million from Tiger Global and Sequoia Global for its subscription management software.

Productboard then worth $ 1.7 billion after raising $ 125 million for its product management software.

Loadsmart raised $ 200 million to automate logistics, valuing the company at $ 1.3 billion.

TTTech Auto raised $ 285 million to the tune of $ 1 billion for its automated driving software.

Butterfly then worth $ 1 billion then raise $ 125 million to provide employers with a digital benefits platform.

A MESSAGE FROM DATAIKU

Dataiku is the only AI platform that connects data and makers, allowing anyone to turn data into real business results-from mediocre to lunar. Because AI can do so much, but there is no soul in the machine, just in front of it. If you don’t, it’s just data.

Learn more