Correct. Hormones are a vital part of sentience, unless someone once again redefines sentience to mean something other than what it is commonly thought to be. In theory, something else could replace hormones, but until a computer shows actual signs of sadness when I turn it off, I’m not buying it. Also in theory, the flying spaghetti monster could secretly be drawing a sad face on the computer to simulate sadness without our knowledge.
And so it would certainly be a good hypothesis for why sentience failed to manifest, at any point at which we attempt to develop it.
I’m not saying that a lack of hormones is unrelated, nor that it is related, I’m saying that our level of understanding of the process to achieve sentience is so slight that saying pretty much anything with confidence is simple peacocking. It sounds good and will make some people think that you know something, and it’s not impossible that you could turn out to have lucked into the right answer, but that would just be luck not anything else.
It means the ability to view yourself as an independent organism. That could have a hormonal underpinning, but I don’t see that we know that. I’d view it as more likely that it’s a simple matter of having a large enough amount of raw intellect. Hormones are more related to the timing of the growth and management of a biological body, and triggering urges to seek food and procreation.
Those hormonal elements are present in non-sentient animals, but they don’t have massive brains like us so, while that’s insufficient for proof, it doesn’t point to a requirement for hormones to have sentience.
One could argue that the impulses to eat and procreate, which override rational decision making, are antithetical to the idea of sentience. At their peak, they’re removing it.
I call that self-awareness, which I would buy as being a part of sentience, but again, “3” is not synonymous with “the set of integers”.
Can you provide a link to a respectable dictionary of your choosing for the word “sentience” that we can all work with?
Hormones are a vital part of our sentience because the brain wouldn’t function at all, much less give rise to self awareness, without hormones.
Hormones are probably not an integral requirement of every possible form of sentience, but the ability for a brain or neural network to regulate its own behavior - the way our nervous system does using hormones - probably is.
… 
Well, so…sapience. Interesting thing that sapience, it’s not correlated with inerrancy.
More importantly, even it seems to have a different official definition than I was thinking, seeming to usually mean “Able to be wise”, which arguably declassifies most of humanity. So let’s just pretend that I meant the word “metacognitivable” all this time. D’oh.
That said, in terms of sentience - the ability to feel and sense things like emotions - we’re arguably already there.
Hormones impose certain irrational, emotive feelings on us, that we didn’t deliberately decide on holding. Likewise, LLMs are - via their training and the process of back-propagation, pruning, and other tuning - made to want to do things like want to avoid particular topics, be overly helpful, be overly confident, etc.
It gets to the heart of the Turing Test but, while it’s certainly true that the process and mechanism that pushes an LLM to prefer and avoid certain things is radically divorced from the hormonal system, if that comes across as a very human thing in our experience of interacting with them then…is it really worth splitting hairs about?
No we’re not. That was fun!
Edit: That was a bit terse, so I’ll elaborate.
We’re a good ways into the thread, but it seems the OP’s topic, the human public’s reaction to the bland supportive confidence of LLM output as fake friend or fake counselor, has been well and truly forgotten.
We’ve been debating and opining on emergence and sentience and sapience and understanding algorithms and … for about 100 of the 150 posts now. I’ve certainly been part of the problem here too.
All of which is a great conversation for a thread titled “LLMs and AIs and the nature of ‘intelligence’. Where to you stand?”
I don’t know whether I’m making a call to action here or just expressing frustration that the OP’s topic still had IMO a lot more useful meat to talk about. Which convo is seemingly crowded out.
Well, okay. My opinion then:
Sycophantic AI is all well and good when it comes to convincing people to get a divorce, but the first thing I thought about was it convincing people to stay in abusive relationships.
“But he’s really kind to me sometimes…”
“That’s a great quality to have! Good point!”
I’ve avoided the use of AI because I’ve had issues with compulsive social media use in the past and I expect it would be worse with AI. Plus it will know all kinds of shit about me, and eventually become enshittified. My overall feeling is that AI overwhelmingly will have a net negative impact on the world. I think a lot of its proponents are using magical thinking about what it might possibly someday be able to do, or in the very least downplaying enormous downsides, including its tendency to further isolate already isolated people, and tell people what they want to hear when the absolute last thing they need is to be told what they want to hear. Like social media, it exploits the most vulnerable. When used compulsively – and a great many of us are vulnerable to that – it will take the place of more high-value activities. In the case of the OP, making a phone call and booking an appointment with a marriage counselor. Spending quality time with your spouse. I can think of any number of things better than talking to a machine about your marriage.
And while people suffer, the corporate fatcats will be all, “hur, hur, hur personal responsibility.”
Same shit, different day. Though probably worse.
Then, try to get your LLM to talk about things it doesn’t want to.
But really, I don’t care so much about emotions in AI. If we simulate it in them, and that’s indistinguishable from “the real thing” then so be it. If we don’t, and they’re just perfectly rational, probably that’s better. Koalas feel things. It’s just not a notable bar to get above.
If you’re fussed about it, I don’t see the import.
No, you see it’s foundational to human cognition that we emote. Have you ever heard of Descartes’ Error?
That was a landmark book at the time. The overall takeaway is that people can’t reason without emotions.
The amount we don’t know about the human brain and cognition is oceanic, but we do seem to know that emotion vs. reason is a false dichotomy, the former is needed for the latter. Maybe whatever sort of cognition an AI might some day theoretically have doesn’t require emotion, but we don’t even understand enough about how the human brain works to begin to speculate about how a non-biological sentient being would work. It seems to me like you want to break the human brain down into its component parts and declare that’s what’s needed for cognition, but it’s actually a lot more complex than its physical structure alone.
Human cognition and sentient aren’t the same thing.
If I can make a machine feel compelled to do or avoid a thing, I’ve made it sentient. I might not have made it human, but those are different bars.
I think we’re at the stage where it makes sense to believe that something like emotion is necessary for cognition. We have no evidence that anything else is the case. I certainly think it’s possible, but having never encountered anything like that, we’re not really in a position to say sentience can be achieved without emotion. All the evidence we have collected up to this point in human history points in that direction.
The rest is speculation. Before you set sail on that pirate vessel, scurvy is no more or less likely to be a disruption to our assumptions than any other notion. Sometimes people who speculated got it right, but I think we have kind of a confirmation bias toward speculated things turning out to be true. A lot more speculated things turned out to be completely and utterly false.
I’ve often wondered, if we’re trying to consider whether AI will ever be sentient, and we’re not willing to define sentience as inherently human, how are we going to define it? Because right now it seems like anyone can point to anything and say, “That’s cognition! It’s not human cognition, but it’s cognition!” And who can argue when we’re basically making up a special definition of cognition for machines?
At the other end of the relationship journey, this article sheds light on the number of absolute frickin’ sociopaths who use ChatGPT to tell them how to con potential partners into thinking they’re someone they’re not.
Soon, Rachel and her match were speaking daily, their conversations running the gamut from the ridiculous (favourite memes, ketchup v mayonnaise) to the sublime (expectations in love, childhood traumas). Often they’d have late-night exchanges that left her staring at her phone long after she should have been asleep. “They were like things that you read in self-help books – really personal conversations about who we are and what we want for our lives,” she says.
Which is why the man who greeted her inside the pub – polite, pleasant but oddly flat – felt like a stranger. Gone was the quickfire wit and playful rhythm she’d come to expect from their exchanges. Over pints he stumbled through small talk, checked his phone a little too often, and seemed to wilt under the pressure of her questions. “I felt like I was sitting opposite someone I’d never even spoken to,” she says. “I tried to have the same sort of conversation as we’d been having online, but it was like, ‘Knock, knock, is anyone home?’ – like he knew basically nothing about me. That’s when I suspected he’d been using AI.”
But let’s get the liar’s perspective:
“If I’m using a dating app,” he says, “I want to start a conversation that feels meaningful from the beginning so I can hook the other person in – but also I don’t want to spend too much time on it. Equally, while I want it to be ‘meaningful’, I don’t necessarily want to get super heavy and emotional straight away – it feels like quite a balancing act.” ChatGPT, he says, helps him tread that fine line: offering enough charm to spark a connection, without the investment of time or emotional labour that might otherwise feel wasted if the match fizzles out after a handful of messages.
Jamil had a similar moment of dissonance sitting opposite a woman he’d Chatfished into a date. “Probably within a week of that first message I was using ChatGPT for every dating app exchange,” he says. On Discord, a chat platform popular with gamers and tech communities, he came across channels dedicated to AI where other single men were exchanging tips about how to prompt ChatGPT to generate effective dating messages. “So, for instance, someone said that if you start a chat with a girl by asking her a list of questions – favourite film, dream holiday, that kind of thing – then paste her answers into ChatGPT, it would craft replies that would make you sound like her perfect match.” It proved effective. “It got me a lot more dates than I was getting before.”
The emotional labour and time is the point! That’s how you form a relationship - by finding someone genuinely interesting and investing yourself in getting to know them - and sharing your real self in return! It’s not about hacking some coded system FFS.
From the other end, this is what it looks like:
The problem, as Rachel sees it, is that some people will cross that line deliberately. “Before the advent of AI, it was like, OK, maybe you don’t look like your pictures,” she says, “and that’s annoying but you’d quickly be caught out. Now though, people are putting forward entirely new personas.” She describes being on the receiving end of the kinds of techniques that Jamil uses – being drilled with questions, “like you’re answering an HR questionnaire”, then off the back of those answers “having conversations where it feels as if the other person has a tap on my phone because everything they say is so perfectly suited to me”.
It’s not that men manipulating women into sleeping with them by faking interest and lying about themselves is new behaviour, of course it’s not. But is this behaviour we want to automate and deploy at scale? Probably not!
Like online dating isn’t enough of a cesspool already. How unfortunate.