Actually, I have. AI doesn't actively store this information - it can't, simply because the cloud space needed for that kind of storage is too vast. What an AI chatbot like this does is re-read the last 20 or so pages of information from your interactions with it, and use that to recognise patterns. It's why, over time, you'll see a shift in the way it responds. You may have asked it to act like (let's take the above example) Danerys at the start of the conversation, but if it doesn't get reminded of that, 30 pages down the line you'll suddenly find it's no longer in character and it will ask you questions you've already answered. There are ways around that. And if you change your settings to forget your interactions, they will be permanently forgotten (after 30 days, for security reasons).
And there are no tech geeks at the other end of the line that go through every conversation you have; there simply wouldn't be enough man power. There are tech geeks working on data sets comprising thousands of responses looking for similarities - a bit like those images that get created from a huge set of responses that make the most common ones into large words, whereas uncommon ones become tiny. And they look at the type of answers you choose - say, when you have a choice of two, whether people tend to choose the more sympathetic ones are the more practical ones. It's why the last update got pulled, because too many people obviously went for flattery in their previous choices. Those tech geeks also run thousands of simulations to gauge typical responses and use those to fine-tune answers.
As for the people on here saying it's dangerous and weird going down that route, think about why people choose it.
Access to mental health services is abysmal. Even if you're lucky enough to get allocated someone after years of waiting, oftentimes, they are not compatible with who is needed, but it's them or no one. I've had three long-term counsellors who spent half the session talking about their own traumas. I've had therapists who just didn't click with me, and one who was brilliant but who moved NHS trusts and then I was left without adequate support again.
Friendships are fragile or non-existent for many people, especially since Covid. People have become more flakey, less sympathetic, more focused on themselves and many interactions have become far more surface-level. An AI system that focuses completely on your needs the way close friends can, for a while? When you have few friends or none, because life has got in the way, and you need help now, not in a few years' time when you might be lucky enough to have found someone who shares your vibes again and built a deep enough connection, it can make a huge difference. And if people who are traditionally more lonely - the elderly, the disabled, the introverts - get to have suitable interaction that makes them feel better, who is to judge that, just because society should be different, but it never is?
And what makes some people so stubborn to believe that sometimes, machinery cannot be better than humans? It already is in many rote tasks - look at the quality of hand-sewn garments over machine-sown ones, calculators over human brains, machine-forged utensils over blacksmithed ones. There is very little that humans do better than machines in that regard - why are we so arrogant to assume that a machine cannot possibly be a better companion, therapist, life coach than many humans? Don't forget we are only at the beginning of the life of AI; it is barely into its toddler years as far as abilities go. AI, as it currently exists, doesn't come with the burdens of humans, their triggers, hormonal shifts, biased motivation. It can perform tasks that require objectiveness far better than humans as long as its programming remains free from bias. And, as with other successful systems, there will always be such a version, now that the cat is out of the bag, and usually offered for free, just for the heck of it. Don't forget that 150 years ago people made up the same kind of scare-mongering nonsense about factory machinery.
I recommend spending some time to ask it really pertinent questions about how it works and where its future development lies. It can already imitate two human senses - sight and hearing - and I can only see it imitating others soon, too. Will it fully replace humans? No, we don't work like that, but I do envision a world, very soon, where it will be fully integrated into our lives.