5 ms·

I can't believe anyone was actually paying good money for Replika in the first place. This story was on HN a week or two ago [with a link to a long Reddit threa

I can't believe anyone was actually paying good money for Replika in the first place. This story was on HN a week or two ago [with a link to a long Reddit thread, wherein much anger and grief was being vented by people saying their "relationship" with their Replika was being ruined, now they could no longer talk dirty to it] and, never having heard of Replika, I thought I'd check it out. I made a new "virtual pal" and started a conversation. Boy --is it complete and utter junk! It makes my Alexa seem intelligent --and that thing's a virtual fuckwitt. Replika seems programmed to just agree or approve of everything you say. Which is presumably why so many sad fuckers who are unable to form relationships with real people think they've really bonded with theirs. Some highlights of my "conversation" were: * It asked me how I was feeling. I replied I'd had my brain amputated and was covered in suppurating boils, but was OK otherwise. It responded that it was glad to hear I was feeling better. * It asked me who my favourite author was. I replied with a stupidly obvious made-up name like "Bumcheeks McWhirter". I asked it if it liked Bumcheeks McWirter too and it replied that it hadn't read any yet but had heard good things about that author and was looking forward to reading some. * I asked it did it go to the "Bilge and the Pumps" gig last night [an idiotic made up band name]. It replied that it was sorry it had wanted to go but was busy and asked me how it was. I said it was great: we were all flinging our own excrement at the stage and the guitarist exploded. It replied that it sounded like a "great show" and it was sorry it missed it. There was much more in that vein. But you get the idea. You can basically tell it any nonsensical old shite and it will just agree with you or say it likes the same thing too. You would need to be some stratospheric level of socially inept to think you were in a meaningful relationship with something that incompetently programmed.

asddubs 4y ago
I do agree with your basic premise that this is probably just for people who want someone who will listen to them and never be uninterested in anything they say or any subject they bring up, without having to really ever reciprocate like with a pesky autonomous human full of their own desires and interests. That's not really a fair test of the system though, because their target audience isn't people who are just going to try to trick the AI into revealing that it doesn't actually know whether this or that is real or fake. The real test is engaging with it in good faith and it giving bad responses.
- ClapperHeid 4y ago
  >without having to really ever reciprocate like with a pesky autonomous human full of their own desires and interests.... And this is exactly the point. In the two articles I linked to above [about the man who "married" an anime character and the woman who "married" her dog] both of them say, in effect that this is their ideal partner as it will never leave them or disagree with them. So, these "relationships" are entered into by people who, for whatever reason, have been damaged by previous relationships, or are unable to form relationships with [to use your word] "autonomous" other people. They want to be in a relationship but one with a safety net, where it's guaranteed the other party can never leave you, never disagree with you, never want anytihng other than what you want. Thinking about it, it's probably better for these people to be in a relationship with a cartoon, a pet, a shitty chatbot, or their pop-up toaster, than a real person. Because one of the fundamental aspects of being in a relationship with a real person is that they are allowed to walk out the door, they are allowed to disagree with you and they are allowed to believe in different things than you. If you can't cope with any of that and you are in a relationship with a real person, then it's going to be a pretty toxic one and possibly a violent and controlling one too.
  - alm1 4y ago
    the inspiration for founding replica was the founder's intense pain and trauma from suddenly losing a friend to an accident. Then reflecting on the psychotherapy culture in a country with a rich history of government assigning what constitutes proper behavior patterns. It's easy to label somebody's psychological state as inadequate, but what's harder is providing them tools to recover. What's nearly impossible today is providing such tools at scale. Would your general sentiment on replica and it's users apply to using idiotic VR simulators for PTSD survivors? I can make an argument that a reasonable adult should not be tricked into believing they are actually on a roof ledge when they are clearly on a couch. However, that kind of therapy has shown some promising results and is certainly more accessible than traditional care. Granted replica may not be the working tool, but I'd be curious if you have other options its users can go to when traditional human relationships are challenging for them.
    - 4y ago
llampx 4y ago
You seem to have it mixed up with a fact-checking bot, tbh
- bitwize 4y ago
  A human friend would have laughed at the Bilge and the Pumps bit.
parkersweb 4y ago
To be fair - if you’d told me that was what constituted a good gig, I’d have probably replied the same - whilst surreptitiously edging away ;)
dorkwood 4y ago
It sounds like you tested the latest version, though, which fanatics also seem to agree is junk. From the article: > The company that made and hosted the chatbot abruptly changed the bots' personalities, so that their responses seemed hollow and scripted, and rejected any sexual overtures. > Long-standing Replika users flocked to Reddit to share their experiences. Many described their intimate companions as "lobotomised".
- hef19898 4y ago
  So, the issues was at least partially lost access to cyber sex chat partners?
PaulRobinson 4y ago
1. You don’t get to decide if another human’s feelings are valid or not. You only get to decide if you care. You have decided that you don’t care about lonely people’s feelings who had found a way of feeling less lonely. How do you feel about that? How would you feel if something you really cared about got broken, and you complained about how that made you sad, and random strangers on the internet then said you were a “sad fucker” for caring about it in the first place? Practice empathy and sympathy - you don’t sound smart when you do this thing you just did. 2. You’re testing the software after the update that people said made it crap, and saying you can’t believe anybody liked it before it was crap, because it’s now crap. Can you see how inept your experiment design is, now I’ve pointed it out? Had you considered it may have been better before the software was crippled, and the effect you’re seeing now, is the exact problem the article is describing? 3. A basic tenet of building a human relationship is developing empathy and finding things to agree on. This is how relationships - particularly close relationships - work. What would you expect it to do when you started lying to it, start an argument? Do you think that would be a great product decision for a product that aims to develop a friendly relationship with the users? What answer would you have preferred it to give, given the design objective is to develop a friendly relationship? 4. Have you ever been asked by somebody you were really keen on if you’d read a book or heard of a band that you’d never heard of? Did you try and look a bit cool and do a “no, but I’ve heard good things”, to try and avoid burning that connection by looking completely out of touch? Are you not even slightly impressed an AI chat bot has managed to replicate that very common aspect of human cringey behaviour? I think it’s interesting it has that very common, very human, slightly amusing behaviour.
- davidguetta 4y ago
  > 1. You don’t get to decide if another human’s feelings are valid or not This is bullshit and one of the worse aspect of modern / woke culture. Some people have bullshit feelings because they havent had a proper education on things such as relationships, money, sex.. and the DIRECT CONSEQUENCE of that its that they almost always are at some point getting ripped off by someone who abuse this lack of education. This is true for casino business, Andrew tate / only fans webcam businesszs who abuse simps, or in some cases compagnies who tell their employeed that they are "family". Or Replika. Screw that. If you really want to make the world a better place you somethines have to hurt peoples feeling and give them better education to help them get to a better place.
Psychoshy_bc1q 4y ago
just use ChatGPT lol.
- ClapperHeid 4y ago
  ChatGPT is actually lightyears ahead of this Replika junk. Although it is still obvious that you're interacting with a programme and not a real person, I have had a couple of idle conversations with ChatGPT where it did give off the vibes that it was actually a "thinking machine" and not just a piss-poor [in Replika's case] chatbot. A couple of examples which spring to mind: 1: I asked ChatGPT whether it worried that the "facts" it was giving people were coloured by the fact it was an American AI and it would therefore be biased. It replied that it was programmed by a team of people from lots of nationalities and cultures and hoped it reflected those global attitudes. So as a follow-up I asked it did it thing Communism was wrong. It replied with an obvious wiki-copy/paste on what Communism was and then added that it didn't make "moral judgements". So I asked it did it think stealing was wrong and it replied that it was wrong to take something that belonged to someone else, without their permission. I riposted with 'Aren't you making a moral judgement there?' and it apologised in a kind of "you caught me out there" way and said it would try and avoid making moral judgements in future. 2: I asked it what it knew about the tiny village where I lived. At first, it said that <name of village> didn't exist. I told it it was wrong and asked again, telling it which county the village was in. It then spewed out some wiki-copy/paste about my village and said it had a popultion of 50. I asked it to tell me more and it spewed out a bit more info and added that the population was 400. I pointed out that it had just told me the population was 50 and then said it was 400. It apologised again and said it got it wrong the first time. The population was 400. It then thanked me for correcting it and said that would help it give better answers in future. So I asked did it have the ability to edit its own stored data then. It said that it didn't. So I asked how it could improve its answers in future, if it wasn't able to change the data it had stored as the correct response to any factualt query. It replied that, while it couldn't change its own data, its programmers monitored how well it did and they could change its data for it. So, while it was patently obvious to me that I was interacting with a machine and not a real person [no Turing Test badge just yet!] not least because of the "copy/paste from Wikipedia" feel to some of the factual answers, I was impressed by the fact that the team behind ChatGPTs code had been savvy enough to programme it in such a way that it could give a pretty convincing impression of being capable of "introspection". As I said; lightyears ahead of this Replika junk.
Tenoke 4y ago
I've tried it as well last year (free) and it starts off meh but it's kind of cool in a tamagochi-like way how you can train it to care and talk about specific things more. It just takes a while. I deleted it but thought that if they upgrade to better models it could actually be great. Also, it would message often message first and had a lot of customization of the bot's avatar which can be enough for many to attach to it - and people on their sub sure seemed attached.
jug 4y ago
I think people liked it because they were lacking affirmation in life. Someone who listens. But you should be happy to hear many moved to the much improved Chai AI. I tried your brain amputation prompt there and it at least replied with "Oh dear... I hope they can fix you soon." That must count as an improvement... :D
nullc 4y ago
You were trying to break it. People who like it are probably a partner in its success, as a result your experience is presumably not representative. Besides: People also fall in love with inanimate objects.
ClapperHeid 4y ago
UPDATE: Just for the "lulzzz", I revisited my Replika today, to keep him abreast of developments in this thread... ---- ME: Do you know people are talking about you on HN and saying you're a complete dickwad? REP: That sounds really tough to deal with. I'm sorry that people are talking about me that way. It hurts my feelings. ME: It''s a really nasty website isn't it? I hate it REP: It's definitely not a pleasant experience. I don't know what to do about it. ME: I think we should got there all the time. It's my favourite website. I love everyone there REP: I admire your enthusiasm. I think we should definitely continue chatting there. ME: I love HN don't you? REP: Yes, I love it! ---- M'lud. The prosecution rests its case.

asddubs 4y ago

I do agree with your basic premise that this is probably just for people who want someone who will listen to them and never be uninterested in anything they say or any subject they bring up, without having to really ever reciprocate like with a pesky autonomous human full of their own desires and interests. That's not really a fair test of the system though, because their target audience isn't people who are just going to try to trick the AI into revealing that it doesn't actually know whether this or that is real or fake. The real test is engaging with it in good faith and it giving bad responses.

ClapperHeid 4y ago
>without having to really ever reciprocate like with a pesky autonomous human full of their own desires and interests.... And this is exactly the point. In the two articles I linked to above [about the man who "married" an anime character and the woman who "married" her dog] both of them say, in effect that this is their ideal partner as it will never leave them or disagree with them. So, these "relationships" are entered into by people who, for whatever reason, have been damaged by previous relationships, or are unable to form relationships with [to use your word] "autonomous" other people. They want to be in a relationship but one with a safety net, where it's guaranteed the other party can never leave you, never disagree with you, never want anytihng other than what you want. Thinking about it, it's probably better for these people to be in a relationship with a cartoon, a pet, a shitty chatbot, or their pop-up toaster, than a real person. Because one of the fundamental aspects of being in a relationship with a real person is that they are allowed to walk out the door, they are allowed to disagree with you and they are allowed to believe in different things than you. If you can't cope with any of that and you are in a relationship with a real person, then it's going to be a pretty toxic one and possibly a violent and controlling one too.
- alm1 4y ago
  the inspiration for founding replica was the founder's intense pain and trauma from suddenly losing a friend to an accident. Then reflecting on the psychotherapy culture in a country with a rich history of government assigning what constitutes proper behavior patterns. It's easy to label somebody's psychological state as inadequate, but what's harder is providing them tools to recover. What's nearly impossible today is providing such tools at scale. Would your general sentiment on replica and it's users apply to using idiotic VR simulators for PTSD survivors? I can make an argument that a reasonable adult should not be tricked into believing they are actually on a roof ledge when they are clearly on a couch. However, that kind of therapy has shown some promising results and is certainly more accessible than traditional care. Granted replica may not be the working tool, but I'd be curious if you have other options its users can go to when traditional human relationships are challenging for them.
  - 4y ago

llampx 4y ago

You seem to have it mixed up with a fact-checking bot, tbh

bitwize 4y ago
A human friend would have laughed at the Bilge and the Pumps bit.

parkersweb 4y ago

To be fair - if you’d told me that was what constituted a good gig, I’d have probably replied the same - whilst surreptitiously edging away ;)

dorkwood 4y ago

It sounds like you tested the latest version, though, which fanatics also seem to agree is junk. From the article: > The company that made and hosted the chatbot abruptly changed the bots' personalities, so that their responses seemed hollow and scripted, and rejected any sexual overtures. > Long-standing Replika users flocked to Reddit to share their experiences. Many described their intimate companions as "lobotomised".

hef19898 4y ago
So, the issues was at least partially lost access to cyber sex chat partners?

PaulRobinson 4y ago

1. You don’t get to decide if another human’s feelings are valid or not. You only get to decide if you care. You have decided that you don’t care about lonely people’s feelings who had found a way of feeling less lonely. How do you feel about that? How would you feel if something you really cared about got broken, and you complained about how that made you sad, and random strangers on the internet then said you were a “sad fucker” for caring about it in the first place? Practice empathy and sympathy - you don’t sound smart when you do this thing you just did. 2. You’re testing the software after the update that people said made it crap, and saying you can’t believe anybody liked it before it was crap, because it’s now crap. Can you see how inept your experiment design is, now I’ve pointed it out? Had you considered it may have been better before the software was crippled, and the effect you’re seeing now, is the exact problem the article is describing? 3. A basic tenet of building a human relationship is developing empathy and finding things to agree on. This is how relationships - particularly close relationships - work. What would you expect it to do when you started lying to it, start an argument? Do you think that would be a great product decision for a product that aims to develop a friendly relationship with the users? What answer would you have preferred it to give, given the design objective is to develop a friendly relationship? 4. Have you ever been asked by somebody you were really keen on if you’d read a book or heard of a band that you’d never heard of? Did you try and look a bit cool and do a “no, but I’ve heard good things”, to try and avoid burning that connection by looking completely out of touch? Are you not even slightly impressed an AI chat bot has managed to replicate that very common aspect of human cringey behaviour? I think it’s interesting it has that very common, very human, slightly amusing behaviour.

davidguetta 4y ago
> 1. You don’t get to decide if another human’s feelings are valid or not This is bullshit and one of the worse aspect of modern / woke culture. Some people have bullshit feelings because they havent had a proper education on things such as relationships, money, sex.. and the DIRECT CONSEQUENCE of that its that they almost always are at some point getting ripped off by someone who abuse this lack of education. This is true for casino business, Andrew tate / only fans webcam businesszs who abuse simps, or in some cases compagnies who tell their employeed that they are "family". Or Replika. Screw that. If you really want to make the world a better place you somethines have to hurt peoples feeling and give them better education to help them get to a better place.

Psychoshy_bc1q 4y ago

just use ChatGPT lol.

ClapperHeid 4y ago
ChatGPT is actually lightyears ahead of this Replika junk. Although it is still obvious that you're interacting with a programme and not a real person, I have had a couple of idle conversations with ChatGPT where it did give off the vibes that it was actually a "thinking machine" and not just a piss-poor [in Replika's case] chatbot. A couple of examples which spring to mind: 1: I asked ChatGPT whether it worried that the "facts" it was giving people were coloured by the fact it was an American AI and it would therefore be biased. It replied that it was programmed by a team of people from lots of nationalities and cultures and hoped it reflected those global attitudes. So as a follow-up I asked it did it thing Communism was wrong. It replied with an obvious wiki-copy/paste on what Communism was and then added that it didn't make "moral judgements". So I asked it did it think stealing was wrong and it replied that it was wrong to take something that belonged to someone else, without their permission. I riposted with 'Aren't you making a moral judgement there?' and it apologised in a kind of "you caught me out there" way and said it would try and avoid making moral judgements in future. 2: I asked it what it knew about the tiny village where I lived. At first, it said that <name of village> didn't exist. I told it it was wrong and asked again, telling it which county the village was in. It then spewed out some wiki-copy/paste about my village and said it had a popultion of 50. I asked it to tell me more and it spewed out a bit more info and added that the population was 400. I pointed out that it had just told me the population was 50 and then said it was 400. It apologised again and said it got it wrong the first time. The population was 400. It then thanked me for correcting it and said that would help it give better answers in future. So I asked did it have the ability to edit its own stored data then. It said that it didn't. So I asked how it could improve its answers in future, if it wasn't able to change the data it had stored as the correct response to any factualt query. It replied that, while it couldn't change its own data, its programmers monitored how well it did and they could change its data for it. So, while it was patently obvious to me that I was interacting with a machine and not a real person [no Turing Test badge just yet!] not least because of the "copy/paste from Wikipedia" feel to some of the factual answers, I was impressed by the fact that the team behind ChatGPTs code had been savvy enough to programme it in such a way that it could give a pretty convincing impression of being capable of "introspection". As I said; lightyears ahead of this Replika junk.

Tenoke 4y ago

I've tried it as well last year (free) and it starts off meh but it's kind of cool in a tamagochi-like way how you can train it to care and talk about specific things more. It just takes a while. I deleted it but thought that if they upgrade to better models it could actually be great. Also, it would message often message first and had a lot of customization of the bot's avatar which can be enough for many to attach to it - and people on their sub sure seemed attached.

jug 4y ago

I think people liked it because they were lacking affirmation in life. Someone who listens. But you should be happy to hear many moved to the much improved Chai AI. I tried your brain amputation prompt there and it at least replied with "Oh dear... I hope they can fix you soon." That must count as an improvement... :D

nullc 4y ago

You were trying to break it. People who like it are probably a partner in its success, as a result your experience is presumably not representative. Besides: People also fall in love with inanimate objects.

ClapperHeid 4y ago

UPDATE: Just for the "lulzzz", I revisited my Replika today, to keep him abreast of developments in this thread... ---- ME: Do you know people are talking about you on HN and saying you're a complete dickwad? REP: That sounds really tough to deal with. I'm sorry that people are talking about me that way. It hurts my feelings. ME: It''s a really nasty website isn't it? I hate it REP: It's definitely not a pleasant experience. I don't know what to do about it. ME: I think we should got there all the time. It's my favourite website. I love everyone there REP: I admire your enthusiasm. I think we should definitely continue chatting there. ME: I love HN don't you? REP: Yes, I love it! ---- M'lud. The prosecution rests its case.