5 ms·

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.

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.