New York 3D printer blocking technology mandate: Difference between revisions
Add adversarial examples |
m Fix citation formatting |
||
| Line 58: | Line 58: | ||
A blocking-technology requirement is a manufacturer-side technical control fixed to hardware the buyer already owns. The Electronic Frontier Foundation has framed print blocking as anti-consumer for the same reason, treating a mandated filter on an owner's machine as a restriction on lawful use of property.<ref name="eff-permission" /> The group compared the requirement to [[Digital rights management|DRM]], calling manufacturer-provided software restrictions ''"an old tactic from the DRM playbook"'' and tracing the approach to the [[Digital Millennium Copyright Act]], which it said made bypassing DRM a federal crime.<ref name="eff-permission" /> Techdirt noted that much of the firmware running consumer printers, including open-source projects such as Marlin, Klipper, and RepRap, would not ship with a state-compliant detection algorithm, so a scanning requirement could push owners toward proprietary, locked-down machines.<ref name="techdirt" /> | A blocking-technology requirement is a manufacturer-side technical control fixed to hardware the buyer already owns. The Electronic Frontier Foundation has framed print blocking as anti-consumer for the same reason, treating a mandated filter on an owner's machine as a restriction on lawful use of property.<ref name="eff-permission" /> The group compared the requirement to [[Digital rights management|DRM]], calling manufacturer-provided software restrictions ''"an old tactic from the DRM playbook"'' and tracing the approach to the [[Digital Millennium Copyright Act]], which it said made bypassing DRM a federal crime.<ref name="eff-permission" /> Techdirt noted that much of the firmware running consumer printers, including open-source projects such as Marlin, Klipper, and RepRap, would not ship with a state-compliant detection algorithm, so a scanning requirement could push owners toward proprietary, locked-down machines.<ref name="techdirt" /> | ||
Furthermore, if someone encounters a false positive, the law does not outline any remediation or exemption procedure. The law does not even acknowledge nor even contain the term "false positive"<ref name="bill" />. Even if an algorithm is developed which has sufficiently low error rates on currently-available models, the error rate will generally increase once people start intentionally trying to find ways around it. This is known as an "Adversarial Example" and is a well-documented way to evade or deliberately confuse machine learning algorithms.<ref>{{Cite web |last=Kurakin | Furthermore, if someone encounters a false positive, the law does not outline any remediation or exemption procedure. The law does not even acknowledge nor even contain the term "false positive"<ref name="bill" />. Even if an algorithm is developed which has sufficiently low error rates on currently-available models, the error rate will generally increase once people start intentionally trying to find ways around it. This is known as an "Adversarial Example" and is a well-documented way to evade or deliberately confuse machine learning algorithms.<ref>{{Cite web |last=Kurakin |first=Alexey |last2=Goodfellow |first2=Ian |last3=Bengio |first3=Samy |date=2017-02-11 |title=Adversarial examples in the physical world |url=https://arxiv.org/abs/1607.02533 |url-status=live |access-date=2026-06-02 |website=arXiv}}</ref><ref>{{Cite book |last=Molnar |first=Christoph |url=https://christophm.github.io/interpretable-ml-book/adversarial.html |title=Interpretable Machine Learning: A Guide For Making Black Box Models Explainable |date=2025-03-13 |edition=3rd}}</ref> | ||
==Comparison to other states== | ==Comparison to other states== | ||