Ethics in AI: Slippery slope

Things move fast in artificial intelligence (AI), including outlooks both cheerful and chilling.

A year ago, ethical concerns around AI centered on surveillance cameras using images of people’s faces without permission and invasive personal behavior monitoring for marketing purposes. According to Gartner, by 2023, up to 10% of AI training data will be “poisoned” by benign or malicious actors and 20% of successful account takeover attacks will use deepfakes – deceptive manipulations of video or audio content – to socially engineer users to turn over sensitive data or move money into criminal accounts.

By 2025, the US-based research firm expects 10% of governments to use a synthetic population with realistic behavior patterns to train AI while avoiding privacy and security concerns. Meanwhile, 75% of conversations at work will be recorded and analyzed to enable the discovery of organizational value and risk.

These projections suggest a new level of risk management concern for VC investors in AI, going beyond toxic branding and bad press to include myriad possible contraventions of environment, social and governance (ESG) compliance agreements. Even legally tight assurances that the technology will not be used in abuse of civil or human rights will offer little protection when the security of the data being exploited is increasingly vulnerable.

The issue might be compounded for investors in the medium term due to shifting industry-targeting trends under the broader AI theme. While AI innovation in media, retail, and financial services will continue apace, areas such as healthcare and education are attracting more capital.

“If you build AI to sell to the internet industry, you won't have any customers because every internet company will have a lot of data, and they will find it absolutely core and necessary to have their own AI and probably not rely on outside technologies,” Kai-Fu Lee, chairman and CEO of China’s Sinovation Ventures, told the Hong Kong Venture Capital & Private Equity Association’s (HKVCA) Asia Forum last month.

Sinovation has backed seven AI unicorns at early stages across various industries, but says it is pivoting toward applications in healthcare, having hired two new partners to cover the space. “With AI opening data up plus with advances in genomics that personalize treatment, we are at the beginning of a process that will really leave behind the old ways of doing medicine and treatment,” Lee said.

On the surface, the risks of AI in healthcare appear similar to those in the automotive industry; the machine makes a wrong decision, someone suffers, and consequences ensue. But there are deeper ethical dimensions here, and they’re spiraling in ways investors may find difficult to track.

In 2019, the main ethical question with healthcare AI was as pedestrian as the privacy issues that come with automated patient consultations. By 2020, scientists at Tufts University and the University of Vermont had combined AI and biotech to create “xenobots,” which have been characterized as computer-designed organisms and the world’s first living robots.

Education is another growth area that could see a similar escalation in ethical timebombs, especially since it often requires deep human-computer interaction with children and personalized services backed by cognitive AI. One of the greatest threats here is AI bias, whereby discriminatory assumptions are programmed into datasets. Simply put, teaching machines can easily replicate racism, sexism, or cultural offenses on a mass, systematized, and generation-defining scale. 

For investors, the balancing act will be as much about the scope of their target companies’ impact as the due diligence particulars of any given opportunity. Unfortunately, the bigger the potential to do good, the bigger the potential to do harm. 

• Tweet  
• Facebook  
• LinkedIn  
• Google plus  
• Save this article  
• Send to  

• Topics
• Technology
• Greater China
• China
• artificial intelligence
• Sinovation Ventures