Lauren WeinsteinA federal law should be passed making AI firms fully responsible for any and all content disseminated from their generative AI systems. Period. No exceptions.
Louis Ingenthron@lauren Yeah, and a law should be passed making Microsoft fully responsible for any and all content created with Microsoft Excel. Period. No exceptions.
@LouisIngenthron False comparison. Not even close.
@LouisIngenthron Excel is, for all practical purposes, a calculator. Users can see all input data and how that data was used to formulate results. This is not the case for generative AI. The full scope of sources used, how those sources were used, and virtually all other aspects of the system are a black box to users. The AI firms want to create new content and then disclaim responsibility for it. Unacceptable.
@lauren Tbf, I've used some excel spreadsheets that were pretty "black box" too.
But more importantly, the transparency of an algorithm has no bearing on the liability for speech resulting from its use. Nearly every video game is a black box. Should the publishers therefore become liable for user content (like online voice chat) as a result?
But more importantly, the transparency of an algorithm has no bearing on the liability for speech resulting from its use. Nearly every video game is a black box. Should the publishers therefore become liable for user content (like online voice chat) as a result?
@LouisIngenthron The fundamental question is pretty simple. Let's say someone asks a generative AI system a question, it provides an inaccurate answer, and then someone is harmed or killed as a result of that answer. Who is responsible for that answer (which is original content created by that system) and the damage it caused? "Nobody" is not acceptable.
@lauren The person who asked is responsible. They used the system, after being warned about its inaccuracy multiple times during the onboarding process and *underneath every prompt* (see image), and then chose to use this potentially faulty information in a life-or-death situation.
I'm a pilot. If I choose to get my weather information from Chat GPT and end up crashing as a result, that's my own damn fault.
@LouisIngenthron Those disclaimers are there to satisfy the corp lawyers. They are not a license to spew dangerous misinformation to the public in a way that is specifically designed to foster confidence in those answers. I can pretty much guarantee that the amount of litigation that will be focused on this area will be immense. Especially when ads start running with those answers. The advertisers are gonna be just THRILLED having their ads running on wrong answers that end up hurting people. Oh yeah.
@lauren They're there as much for the lawyers as for the users. Just like "don't eat poison" labels.
And, yeah, I think there's nuance there. When a company decides to use a chatbot as customer service, to speak on their behalf, then it absolutely should be liable for the results.
But that's a far cry from a generalized chatbot with a "don't believe my bullshit" disclaimer that can be easily-manipulated by the user.
@LouisIngenthron Manipulation is straightforward to demonstrate from logs. Let's pin this down even more. Most people simply do NOT understand the differences between these systems and traditional search. There's no reason to expect them to, given how (for example) Google is pushing AI Overviews to the top of SERPs. Google clearly wants users to accept AIOs as THE answers. The disclaimers are meaningless in the real world in this context, except as an attempt at legal cover for the firm. This is the standard "victim blaming" that Google and other tech firms commonly use. I don't think it's going to fly in the current regulatory and political environment, and the firms have not internalized this fact yet.
@lauren Yeah, I think the search companies putting AI responses as fact at the top of results, especially when the user has not opted in nor acknowledged the danger, could be one of the cases where a company has chosen to use the bot's speech as its own and therefore becomes liable for it.
But, again, I draw the distinction between that conduct and a chatbot with a disclaimer.
@LouisIngenthron This is going to be a deep well of litigation as courts work through all this. My gut feeling is that it isn't going to turn out well for the firms.
Bas (Tools on Tech)@lauren @LouisIngenthron I agree here with the problem that end users don't understand LLM and as such take things that they shouldn't at face value. But should their ignorance limit me in using it do very powerful things.
Tools come with risks, powerful tools even more so. But in the end I'm mostly curious on the court cases to see how deep this rabbit hole goes.
Cookiefiend@lauren @LouisIngenthron Side note: Google is making it harder to find actual web answers. A week ago I could look at the options immediately above the AI Overview and select WEB. Now my options are:
All
Images
Videos
Shopping
Forums
More
I need to take an extra step just to get to the web...and that change is in the space of a week.
All
Images
Videos
Shopping
Forums
More
I need to take an extra step just to get to the web...and that change is in the space of a week.
CP93@LPerry2 @lauren @LouisIngenthron
Not sure what the difference is between "All" and "Web."
Anyway, I switched to DuckDuckGo and am much happier.