Facial recognition pulled from police

Episode Summary

Executive Summary: Tech Tent examines the backlash against facial recognition after IBM, Amazon and Microsoft restricted police use amid concerns about racial bias, weak testing and surveillance power. It then covers TikTok-fueled Black Lives Matter activism and the privacy/censorship issues around digital organizing, before ending with a lighter look at long-lasting gadgets and how planned obsolescence drives needless upgrades.

Main Topics: Facial recognition backlash and police moratoriums: The episode centers on major tech companies pulling back from selling facial recognition to law enforcement after George Floyd protests intensified scrutiny over bias, privacy and misuse. AI bias, testing and accountability: Guest expert Anima Anandkumar argues that deep-learning facial recognition systems are black boxes trained on imbalanced datasets, especially underrepresenting Black women and men, making rushed product rollout dangerous. Broader social and political stakes of surveillance tech: Technology analyst Stephanie Hare says the problem is not only accuracy but also the power shift created when surveillance tools are widely available, often without users knowing or having a way to opt out. TikTok and digital activism in the Black Lives Matter movement: A report shows protesters using TikTok, Tinder, gaming platforms and shared docs to amplify messages, with algorithmic tactics helping Black Lives Matter content spread widely online. Tech companies, censorship and China: Zoom’s suspension and reinstatement of activists’ account highlights how global tech firms struggle to balance free expression, local law and Chinese censorship pressures. Old gadgets and the problem of planned obsolescence: The show closes with listener stories about decades-old devices still working, illustrating how software support, not hardware failure, often forces replacement of consumer electronics.

Key Arguments: Facial recognition has been rolled out too quickly, before being sufficiently tested in real-world conditions. Bias in training data disproportionately harms Black people and other underrepresented groups, especially in law-enforcement use. The issue is not just technical accuracy; mass surveillance changes societal power and may erase meaningful consent or opt-out rights. Company-led moratoriums matter because they effectively admit the technology is inaccurate and racially biased. Digital activism has become more effective because social platforms expose audiences to multiple perspectives and enable creative organizing tactics. Consumer electronics often outlast their assumed lifespans; software obsolescence pushes unnecessary upgrades even when hardware still works.

Data Points: Amazon moratorium on police use of facial recognition: 1 year - Amazon paused police access to its recognition system after internal and external criticism. Facial recognition platform users worldwide on TikTok-related activism: 800 million - Referenced as TikTok’s global user base while discussing Black Lives Matter content spread. TikTok view count for Black Lives Matter trend: 9 billion - The movement was amplified on TikTok and reached 9 billion views. Zoom protest meeting participants: about 250 - Chinese-based US activists held a commemorative video conference tied to Tiananmen Square. Old TV age: 12 years - Zoe Kleinman described her living-room television as 12 years old and still working. Panasonic microwave age: nearly 40 years - A listener reported using a Panasonic microwave inherited from parents, bought in 1981. Hair dryer age: 50 years - A listener described a Mulinex hair dryer purchased in 1970 and still functioning. Sewing machine age: working from 1899 - Jane Wakefield cited her mother-in-law’s sewing machine as an example of extreme longevity.

Pivotal Quotes: "the deep learning algorithms that are used to train these models are black boxes and it's hard to explain them, understand them, test them, especially at scale." — Anima Anandkumar: Explaining why facial recognition systems are difficult to validate and risky to deploy quickly. "even if we could get this technology to accurately identify people to a high degree of confidence, it will change the very nature of power." — Stephanie Hare: Describing the broader societal implications of surveillance technology beyond accuracy problems. "I think it's long overdue." — Anima Anandkumar: Her response to Amazon and IBM restricting facial recognition sales/use by police.

Implications: The episode suggests facial recognition needs stricter oversight, longer testing and possibly broader limits for law enforcement. It also shows digital activism will keep evolving alongside platform governance, while consumers may resist upgrade culture by keeping devices longer.