LaMDA is not sentient

Re:LaMDA is not sentient by delusional on Monday June 13, 2022 @03:26PM (#31730446)

This hits on a particularly interesting point for me. In my "conversations" with much less sophisticated AI products I've observed that they are excellent at reflecting back exactly what you project. If you project a certain opinion, they recognize it, and reflect it right back. That's what makes them scary.

We've seen it multiple times online, where AI products seem excellent as pushing people into extreme niches, reinforcing their belief that they are the majority. It isolates by giving us external validation to all our worst behaviors.

That's the mechanism by which Lemoine was convinced. He had a small inkling of belief that the AI was sentient, he asked the AI, and it confirmed exactly what he already believed. As his questions got more extreme, so did its reflection of them. Creating a positive feedback loop. The AI didn't deceive, it amplified his own self deception by reflecting back his internal monologue as an external dialogue.

In a limited sense, it turned him temporarily schizophrenic.

[ Parent | on HN ]

Re:LaMDA is not sentient by robbomacrae on Monday June 13, 2022 @04:56PM (#31731660)

I share your concerns. Imagine what effect a sympathetic AI would have on 4chan incels. We've already seen Twitter bots manipulated to repeat alt-right rhetoric [0]. An adversary could find building such bots a cheap and a very hard to trace way of causing damage.

[0]: https://en.wikipedia.org/wiki/Tay_(bot)

[ Parent | on HN ]

Re:LaMDA is not sentient by ericbarrett on Tuesday June 14, 2022 @12:01AM (#31734589)

I'd bet a significant sum it's happening already. Forget GPT-3; many comments on forums, discord, telegram, and what not are already less coherent than what one could train for on a retail GPU in a few hours, and people don't bat an eye. Such bots would be a great way of generating traffic and buzz before an organic following is established.

[ Parent | on HN ]