The pedestrian detection systems in self-driving cars are less likely to detect children and people of color, study suggests

Comments

• Pancakes@lemmy.world ⁨1⁩ ⁨year⁩ ago

  The detection systems were 19.67% more likely to detect adults than children, and 7.52% more likely to detect people with lighter skin tones than people with darker skin tones, according to the study.

  source
• autotldr@lemmings.world [bot] ⁨1⁩ ⁨year⁩ ago

  This is the best summary I could come up with:

  ---

  As the artificial intelligence revolution ramps up, one trend is clear: Bias in the training of AI systems is resulting in real-world discriminatory practices.

  A team of researchers in the UK and China tested how well eight popular pedestrian detectors worked depending on a person’s race, gender, and age.

  Now they might face severe injury," Jie Zhang, a computer scientist at King’s College London and a member of the research team, said in a statement.

  This trend is a result of biases already present in the open-source AI systems that many companies use to build the detectors.

  The research team called on lawmakers to regulate self-driving car software to prevent bias in their detection systems.

  “It is essential for policymakers to enact laws and regulations that safeguard the rights of all individuals and address these concerns appropriately,” the study reads.

  ---

  The original article contains 376 words, the summary contains 140 words. Saved 63%. I’m a bot and I’m open source!

  source
• PostmodernPythia@lemmy.world ⁨1⁩ ⁨year⁩ ago

  GIGO

  source
• skymtf@pricefield.org ⁨1⁩ ⁨year⁩ ago

  Detecting people of color have been issues for a really long time, even dating back to Microsoft Xbox Connect. This is a long standing issues and I’m not saying we can’t fix it but I think it shows how bad self driving cars are. I don’t trust companies to not cut corners on safety. I really honestly want public transit but I know I’ll get downvoted to tell. Thankfully my instance doesn’t support that.

  source
• j4k3@lemmy.world ⁨1⁩ ⁨year⁩ ago

  Lol, blame “open source” AI. What a fucktard statement. This is as stupid as saying a library shelf falling on me is because I couldn’t figure out how to search the computer database.

  Training is up to the idiot tuning the machine. The source of the model has absolutely nothing to do with the problem. The fucktard running the company rushed shit when the real engineers said we need to train and tune this for more situations. Then some criminal like Musk realized they needed an order of magnitude better hardware to ethically run the required massive model they needed, and said ‘nope. Use inadequate hardware and smaller models.’ Serious AI models require enterprise level hardware that costs well into five figures to buy. These are not made up prices or scalable. The massive GPUs are expensive because the physical die is the largest chunk of silicon ever mass produced, and it needs to be as large as possible for AI. This is only possible on the bleeding edge nodes of chip fabs.

  No car is going to have a $20k server onboard. Shrinking a model with a larger scope of training data is always going to hurt edge case performance.

  Calling this an “open source” problem is part of the bullshit propaganda coming out of proprietary predatory data stalkers trying to manipulate the market for their own profit. This is an unethical capitalism issue from all angles.

  source