3 ms·
> Should AI software that’s smart enough to trick humans be forced to disclose itself? Yes, of course. I would even say it should be legal fraud not to, with
by carapace 9y ago
> Should AI software that’s smart enough to trick humans be forced to disclose itself?
Yes, of course. I would even say it should be legal fraud not to, with severe consequences.
- Ajedi32 9y agoWhy?
- carapace 9y agoThat's an extraordinarily open-ended question. If you'll be more precise, I'll attempt a good-faith answer. Cheers
- Ajedi32 9y agoWhy should it be legal fraud with severe consequences for an AI caller to not disclose the fact that it is an AI? You implied the reasoning behind your statement was obvious ("of course") but it's not at all obvious to me.
- carapace 9y ago> You implied the reasoning behind your statement was obvious ("of course") but it's not at all obvious to me. No. The affirmative answer to the original question as posed is what seems to me to be obvious. My additional statement was meant to convey the seriousness with which I feel these matters should be taken. > Why should it be legal fraud with severe consequences for an AI caller to not disclose the fact that it is an AI? Otherwise people (specifically the technologists implementing and selling these kinds of trickster services) will not take other people's (specifically me and anyone who feels as I do) concerns seriously and proceed apace. There is a similar issue around folks who see CRISPR as a way to work around laws that attempt to control GMOs. I have seen a few articles written by pro-GMO folks just crowing about how CRISPR-edited organisms can be sold without GMO labels due to loopholes in the law. These folks have their heads so far up their belief systems that they don't even see the problem.
- Ajedi32 9y agoAgain, I don't see why that's obvious. Why is it necessary that humans be made aware up front that they're talking to a bot? (So necessary that you think not disclosing that information actually needs to be made illegal.)
- carapace 9y agoOkay, first, let's keep clear exactly what we're talking about here: A machine that attempts to imitate a human well enough to trick people into performing a specific action, an action that involves commitments of time and money. I have a specific and a general objection to this. The specific objection is that this thing tries to trick people. It sets up a "frame" that is intrinsically deceptive. That's it's goal. In another comment s2g quoted: > Giving it an obviously robotic voice when it calls. "People will probably hang up," he said. It is designed to trick people. That's bad. When you do it in business, we call it "fraud". My general issue with this is that it normalizes the blurring of the line between real human people and ersatz Artificial Intelligence "people". This is something that's going to be a serious problem on a deep level. There's evidence of this sort of conceptual melting right here in this thread. How long before folks are trying to marry their toaster? Last but not least, there's the meta-problem: the disconnect between the techo-elite innovators and the masses. Think about this: As a businessman, when the phone rings and there's robot there trying to give me money why would I hang up? The robot only called me because someone wants to make a sale (from my POV) so that's the "opposite of a problem". This is the meta-problem: the technologists are so effin' clueless yet default to starry-eyed optimism. "We are as gods, we might as well get good at it." Conversely, save me from the folly of young gods.
- Ajedi32 9y agoIMO behaving like a normal human isn't "tricking people". Voice assistants need to behave like a human to enable natural interaction, and the fact that the AI caller isn't human is irrelevant information in most contexts. It'd be like if I called a business and used a voice changer during the call to make me sound like I'm a different gender than I am. Would that be fraud? No, because my gender is irrelevant information in the context of that call. Now maybe it might be a good idea to program the AI to be upfront about its nature just as a matter of courtesy, but I see no reason why it _not_ doing would be so harmful that it needs to be made illegal.