source: simon willison: quoting nilay patel

level: technical

nilay patel explained on the vergecast that building augmented reality glasses forces a hard technical trade-off. to overlay digital information on the real world, the device needs a camera next to your eye that constantly captures your view. processing that video stream locally is not possible with chips small and efficient enough to fit in glasses stems. the only current options are sending the data to the cloud or using a bulky headset like the vision pro with a separate battery pack.

this means any practical ar glasses product must continuously record and transmit what the wearer sees. patel pointed out that this design inherently invades privacy, not just for the user but for everyone around them. there is no way around it with today's technology. he suggested that the societal cost might be too high, and perhaps we should decide not to build such products at all.

the quote was highlighted by simon willison, who often collects statements on ai and technology ethics. the discussion connects to broader debates about always-on cameras and ai processing of personal data. as companies push for wearable ai assistants, the tension between functionality and privacy grows sharper. patel's blunt assessment challenges the industry to consider whether the benefits justify the constant surveillance required.

why it matters: for ai and data science, this highlights the unavoidable privacy risks in building always-on vision systems that feed real-world data into cloud ai models.


source: simon willison: quoting nilay patel