Wildroot

Queries may use an external AI service. Details

Wildroot Browser

TodayAISimilar posts

Similar posts

@mr_penguin@social.freetalklive.com

The "tiger test" vs AI. So there is this argument that is being made that companies should be liable because the owner of a tiger is liable for the actions of their tiger even if they attempt to train the tiger and fail. This is an utter fallacy and completely misses the point that a manufacturer isn't automatically liable because a user missuses a tool. A gun manufacturer isn't liable for the murder of an individual with that tool. A manufacturer may be liable for negligence in the product in terms of safety, but that's a very different scenario. It's how the AI is used is the potential danger, not a failure in the design of the product.

Show more of this post

That is to say the developer of a AI tool that is purely informational shouldn't be liable for either the misuse of that tool, but they may be liable if that AI tool is used as intended/designed and fails causing injury or similar where said application is inappropriate. Where the line is is effectively inserting AI into everything without consideration for the implementation. If you design a product that incorporates AI to assist with software development a failure in the tool is expected. If you design a product like a vehicle and connect it to AI where the product can't fail, well, then you've been negligent, and as a manufacturer you could be liable. If a software developer uses your AI tool to develop an app and is developing something that could kill and can't that's the developer using the AI tool's fault, not the developer of the AI system the developer used. It's not the breeder of the tiger (aka manufacturer) that is liable in the event the tiger owner is negligent in ensuring the tiger is appropriately handled. Use a car to kill someone it's you who is responsible not the manufacturer.

Opinion
Read post · Mastodon

3 posts close in meaning, closest first

  1. @sbr@hachyderm.io

    At my work, we are heavily invested in AI-assisted coding. We are thus trying to make sure that we do not deliver slop. That said, I get the sense that all those frameworks and harnesses are a weak way to steer your coding agent: use this skill and that skill and then slap together some script so that the agent verifies what needs to be done.

    Same collectionOpinionAI
    Read post · Mastodon
  2. @js@mastodon.nl

    “AI Researchers Have Started Smoking Because They’re Convinced AI Is Going to Cure Cancer,” the headline said. What they didn’t take into account is what the slop machines are highly likely to come up with (if anything at all). Which is a rigid focus on prevention. Ah well. Glad to hear they’re smoking their own product, one could say; good riddance.

    Same collectionOpinionAI
    Read post · Mastodon
  3. @thanetric@mastodonapp.uk

    #AI will repeatedly demonstrate that the "million monkeys" approach to #cybersecurity works well! In the latest debacle OpenAI’s GPT-5.6 Sol found a great big hole in most web servers PHP infrastructure. Proudly publishing how their engine was able to knock over an "old lady" the whole World (led by WIZ.IO) was able to have a go before a finger was put back in the dyke. Sadly 99% of the websites of the Internet are now left broken(i.e. those sites where nobody gives a damn - sysadmin died - who needs support staff etc.) Still, they have probably not got much worth stealing. PHP (and any server side scripting) is ultimately not properly supported by machine architecture no matter how much Unix and Windows implement "security by obfuscation" AI will crush nuts until a few intrinsic improvements are made! bbc.co.uk/news/articles/c3qjnw…

    Similar topic and wordingOpinionAI
    Story · 15 postsRead post · Mastodon