4 ms·

I can totally imagine an algorithm behind Samantha. No strong AI or self-awareness necessary. It just learns from thousands of conversations it has with all the

I can totally imagine an algorithm behind Samantha. No strong AI or self-awareness necessary. It just learns from thousands of conversations it has with all the people around the world, and feeds it back. Kind of like what Google does, but with language recognition and some inference (which is doable, as Watson has shown). Some people fall in love with her as a joke, and Samantha will use these phrases in conversation with other people, and it will eventually catch on for real. So what Theo gets actually served is not Samantha's love, but averaged relationship of all the people that interact with her as a romantic partner. Of course this technically disagrees with the movie at a certain points, but still is a reasonable approximation that preserves the intent.

fragsworth 13y ago
I don't think it's possible without a strong AI. For instance, you wouldn't be able to ask it questions about complex aspects of your personal life, which are entirely unique to you, and have it respond in a reasonable way. Also, it's all too easy to construct questions that have not been asked before, which a human could easily answer but a machine like you describe cannot.
- peddamat 13y ago
  Sure, but given situations in which it could not respond, it could elicit elaboration. From that point on, it would be able to handle increasingly sophisticated responses to that specific situation. It's a data aggregation problem, similar to what we experience, as humans. I also question exactly how unique individual elements of our lives are. In aggregate, they are probably fairly unique, but at an elemental level, we're all probably fairly mundane.
  - oillio 13y ago
    Also, your hypothetical program is not static. It can communicate with thousands of other humans in real time. To half the user base it is Samantha, to the other half, it is Sam. If someone asks Samantha a personal question it doesn't know how to answer, it may try to pose a similar question to the other users as Sam. The larger the user base, the faster and more effective this can become. It won't be as fast as a real human, but with some clever stalling techniques, it might be able to pull it of realistically. In reality, if someone pulled a program like this off, it wouldn't just be Sam and Samantha. It would be thousands of different personalities, trying to build a realistic social network between its own personalities and between its human users. It would be like a mirror held up to society itself.
pilooch 13y ago
for the whole movie I thought it would be revealed that Sam was fro. an outsourced call center ;)
- lovemenot 13y ago
  I have not seen the film, but wouldn't that be a good interim solution? You'd need to set it up for round-the-clock support. Multiple interchangeable human Sams transmitting through a consistent voice interface. Voice-to-text-to-voice is already good, so no major breakthrough needed there. Recruiting Sams would be difficult, though it should appeal to people who want to work from home. QA is probably the greatest challenge. Due to costs, it needs to be at the higher end of the market; $1,000s per year. The Sams could be differentiated: "positivity coach", "your mum" etc.
gnaritas 13y ago
If it learns, it is AI; what you're describing is a chatter bot, Eliza style; it's not learning, it's just doing lookups of previous conversations. Sadly, such bots are more successful in fooling people than any other approach so far in a conversation.
- asgard1024 13y ago
  I am not saying it's not AI or that it doesn't learn; I only claim that it doesn't need to have agency or self-awareness, or even any sensors connected to the real world. As for the personalization, personalized recommendation algorithms are very common, so nothing really new here either. It's feasible that it just filters possible responses by these personal information. As to whether or not it's "fake", that's a different question. Is illusionist's "magic" fake? In a way, it's the only magic that is real.. I don't think people will accept AI until we actually cut the brain completely and find out, to our chagrin, that it's doing essentially the same thing ("lookup of previous conversations").
  - gnaritas 13y ago
    But the brain doesn't do the same thing; it synthesizes new information from those previous conversations and creates abstractions and metaphors and such. It's not at all like a chatterbot doing simple lookups.
    - asgard1024 13y ago
      But in this case, Samantha won't really need to make new metaphors or abstractions, these can come from other people she is connected to. She needs to have certain limited understanding of those metaphors, but doesn't need to be able to create them.
      - gnaritas 13y ago
        > She needs to have certain limited understanding of those metaphors, but doesn't need to be able to create them. They are one and the same; if you can understand a metaphor you can create one.

fragsworth 13y ago

I don't think it's possible without a strong AI. For instance, you wouldn't be able to ask it questions about complex aspects of your personal life, which are entirely unique to you, and have it respond in a reasonable way. Also, it's all too easy to construct questions that have not been asked before, which a human could easily answer but a machine like you describe cannot.

peddamat 13y ago
Sure, but given situations in which it could not respond, it could elicit elaboration. From that point on, it would be able to handle increasingly sophisticated responses to that specific situation. It's a data aggregation problem, similar to what we experience, as humans. I also question exactly how unique individual elements of our lives are. In aggregate, they are probably fairly unique, but at an elemental level, we're all probably fairly mundane.
- oillio 13y ago
  Also, your hypothetical program is not static. It can communicate with thousands of other humans in real time. To half the user base it is Samantha, to the other half, it is Sam. If someone asks Samantha a personal question it doesn't know how to answer, it may try to pose a similar question to the other users as Sam. The larger the user base, the faster and more effective this can become. It won't be as fast as a real human, but with some clever stalling techniques, it might be able to pull it of realistically. In reality, if someone pulled a program like this off, it wouldn't just be Sam and Samantha. It would be thousands of different personalities, trying to build a realistic social network between its own personalities and between its human users. It would be like a mirror held up to society itself.

pilooch 13y ago

for the whole movie I thought it would be revealed that Sam was fro. an outsourced call center ;)

lovemenot 13y ago
I have not seen the film, but wouldn't that be a good interim solution? You'd need to set it up for round-the-clock support. Multiple interchangeable human Sams transmitting through a consistent voice interface. Voice-to-text-to-voice is already good, so no major breakthrough needed there. Recruiting Sams would be difficult, though it should appeal to people who want to work from home. QA is probably the greatest challenge. Due to costs, it needs to be at the higher end of the market; $1,000s per year. The Sams could be differentiated: "positivity coach", "your mum" etc.

gnaritas 13y ago

If it learns, it is AI; what you're describing is a chatter bot, Eliza style; it's not learning, it's just doing lookups of previous conversations. Sadly, such bots are more successful in fooling people than any other approach so far in a conversation.

asgard1024 13y ago
I am not saying it's not AI or that it doesn't learn; I only claim that it doesn't need to have agency or self-awareness, or even any sensors connected to the real world. As for the personalization, personalized recommendation algorithms are very common, so nothing really new here either. It's feasible that it just filters possible responses by these personal information. As to whether or not it's "fake", that's a different question. Is illusionist's "magic" fake? In a way, it's the only magic that is real.. I don't think people will accept AI until we actually cut the brain completely and find out, to our chagrin, that it's doing essentially the same thing ("lookup of previous conversations").
- gnaritas 13y ago
  But the brain doesn't do the same thing; it synthesizes new information from those previous conversations and creates abstractions and metaphors and such. It's not at all like a chatterbot doing simple lookups.
  - asgard1024 13y ago
    But in this case, Samantha won't really need to make new metaphors or abstractions, these can come from other people she is connected to. She needs to have certain limited understanding of those metaphors, but doesn't need to be able to create them.
    - gnaritas 13y ago
      > She needs to have certain limited understanding of those metaphors, but doesn't need to be able to create them. They are one and the same; if you can understand a metaphor you can create one.