Stories
Slash Boxes
Comments

SoylentNews is people

posted by hubie on Monday December 01, @07:07AM   Printer-friendly
from the age-is-in-the-AI-of-the-beholder dept.

Roblox has plans to implement AI to guess user ages but the Australian Labour Government thinks more should be done to protect young people and that the current solution offered by Roblox is insufficient. There is still debate for whether or not Roblox should count as "social media" and be included in the new age restriction laws.

Roblox rolling out new safety measures to stop kids chatting with adults has done little to win favour with Labor, with the Albanese government saying all digital platforms should be proactively protecting "young Australians".

[...] The new measurers, which start in the first week of December, include age-based chats that restrict players from speaking to people outside their age group.

[...] Despite having social elements, Roblox insists it is not a social media.

The eSafety Commissioner agrees but is reviewing whether to include it in the social media ban.


Original Submission

 
This discussion was created by hubie (1068) for logged-in users only, but now has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 4, Interesting) by pTamok on Monday December 01, @08:09AM (6 children)

    by pTamok (3042) on Monday December 01, @08:09AM (#1425476)

    I guess that the people operating Roblox want to avoid using a 'government ID' infrastructure, with all the requirements that that entails, so using machine learning for pattern recognition is a less bureaucratic approach that will make mistakes, but also will act as a porous barrier for under-age users.
    The government, naturally, wants to be able to point to a system claimed to be capable of making no false positives (assessing someone as being above the cut-off age by mistake).

    From a harm reduction perspective, a machine learning system will give a lot of benefit for little outlay, especially if the assessments made by the system are appealable: that is, if it falsely claims you are below the cut-off age, you can get that corrected.

    And there is no need to falsely ascribe intelligence to machine-learning pattern-matching software - such software has been around for decades, used as a tool in many fields.

    Starting Score:    1  point
    Moderation   +3  
       Insightful=1, Interesting=2, Total=3
    Extra 'Interesting' Modifier   0  

    Total Score:   4  
  • (Score: 2, Funny) by Anonymous Coward on Monday December 01, @08:37AM

    by Anonymous Coward on Monday December 01, @08:37AM (#1425478)
    I guess if a kid can't pass the test then the kid is not ready to be exposed to that sort of stuff.

    Similarly a kid who clicks "No" in response to "Are you 18 years of age or older?" is likely not ready for the 18 years or older stuff.

    Of course, not all who pass such tests are ready but at least we eliminate those who are more likely to not be ready.
  • (Score: 4, Insightful) by shrewdsheep on Monday December 01, @09:35AM (2 children)

    by shrewdsheep (5215) Subscriber Badge on Monday December 01, @09:35AM (#1425485)

    'good enough'

    I'm wondering how an ML approach could be good enough. For questions, answer lists will be put up on the internet on short notice. Likewise for chatbot sessions. What remains is usage patterns. I believe that would be very error prone, too. Also, it would result in pop-ups: Unfortunately, you have been detect to be younger than 18 yrs of age. Logging you out.

     

    • (Score: 3, Insightful) by VLM on Monday December 01, @03:42PM (1 child)

      by VLM (445) Subscriber Badge on Monday December 01, @03:42PM (#1425518)

      What remains is usage patterns. I believe that would be very error prone, too

      AI analysis of posted content would probably work pretty well.

      • (Score: 4, Touché) by Anonymous Coward on Monday December 01, @06:28PM

        by Anonymous Coward on Monday December 01, @06:28PM (#1425523)

        > AI analysis of posted content would probably work pretty well.

        Joke's on you, I can think of one orange hair guy who posts frequently at a pre-teen level. Sure would be funny if he got booted for being underage.

  • (Score: 4, Interesting) by corey on Monday December 01, @08:05PM

    by corey (2202) on Monday December 01, @08:05PM (#1425533)

    I said it and will keep repeating it: these companies just need to use a 4-5 question quiz at the start with general knowledge questions that typically only adults would know the answers to. Just like the original Leisure Suit Larry. Minus the Alt-X bypass hack.

    It’ll either stop kids getting in, or consume too much focus and they’ll give up and go somewhere else - problem solved.

  • (Score: 3, Informative) by driverless on Tuesday December 02, @02:05AM

    by driverless (4770) on Tuesday December 02, @02:05AM (#1425562)

    Problem is, Roblox as it's currently run is a pedo buffet. If you've got a young kid (niece/nephew, grandkid, whatever), let them show you the sort of stuff they get exposed to on there, and the creeps who contact them. "AI age verification" and whatnot are just red herrings to draw attention from the deeper problems the platform has.