4 ms·

What makes everyone in this thread so confident that it isn't sentient?

by cwkoss 4y ago

upwardbound 4y ago
At least personally, my subjective experience as a sentient being includes having a 3-dimensional "virtual reality" view of the world around me, which is mostly informed by sight but also by touch and sound. I think this is a key element of human & animal sentience and I tend to think that all animals that have this "3D representation of the world" in their mind are sentient. E.g. fish can solve mazes, in a way illustrative of having a 3D view of the world within their minds like we do, so I think they are sentient. https://www.researchgate.net/publication/251359525_Performance_of_Betta_splendens_in_a_radial_arm_maze https://www.researchgate.net/publication/251359525_Performan... If you search Google for the phrase <you see with your brain not your eyes>, you will find lots of blog posts and the like, explaining with examples which I mean by the 3D VR-like world in your head which I am describing. Another commenter in this thread raised the question of whether robots that navigate the world are sentient. I think the answer is possibly yes but I'm going to give a huge disclaimer which is I don't think that sentience is morally relevant, I think what is morally relevant is whether an entity feels qualia or not. https://en.wikipedia.org/wiki/Qualia https://en.wikipedia.org/wiki/Qualia In my personal view (just hunches really), I think that whether qualia are felt depends on the implementation. My hunch is that an implementation of a neural network in hardware, i.e. an actual network of transistors, might feel qualia, but that a software implementation of the same neural network would not. But others have different hunches and this is a question as yet unsolved by today's science.
tsimionescu 4y ago
Basic understanding of how the model works. As said elsewhere, the model doesn't have memory, it doesn't learn, and doesn't have goals. If you were allowed to play with it, you could quickly observe this for yourself with a bit of adversarial play - ask it about the time it fought Genghis Khan, ask it if it wants you to give it a glass of water, etc - you will quickly realize the illusion behind the system. Tell it whatever facts you want, be amazed how well it might recall them, and then start a new conversation and notice that it has entirely forgotten all of them (again, it has no memory, and it doesn't learn anything). From first principles, we know how the LM works: it has been trained to predict what is the most likely next token in a series of tokens. It has been trained to do this on a gigantic corpus, and is now pretty good at it. You then present it with a prompt, and it starts predicting the most likely token that comes after that prompt, and it keeps predicting for a while until the most likely "token" is to stop. Note that these tokens aren't even words - they are typically letters, or short combinations of letters. To give the illusion of a coherent conversation and memory, the prediction function is applied to the entire conversation so far, not just to the latest prompt. There is more sophistication to avoid various undesirable behaviors, especially in LaMDA (e.g. the model is also trained to evaluate the "safety" of various possible answers, so it doesn't generate a sentence that insults you or discriminates against you or induces you to do something dangerous etc), but this is the gist of how it works.
- cwkoss 4y ago
  Do you think all humans are sentient? This thread has me wondering if people with dementia, anterograde amnesia, etc. would be able to pass the proposed sentience tests others have suggested.
  - tsimionescu 4y ago
    I'm not sure that I would classify humans in the final stages of dementia as being (fully) sentient. I have had the extremely sad experience of helping to care for two different grandparents with late stage memory loss, and while they had moments of lucidity, they had many periods where they were acting almost entirely randomly, or doing nothing at all, barely or not even perceiving the world around them. I struggled a lot to find signs of the person I had deeply loved in them, and there were many times when I can't say I found any signs of a full consciousness in them at all, as troubling as that was. To be very clear, this came after many years of milder forms of dementia, where their sentience could not in any way have been in question, even when forgetting who I was or where they were or other such things. But the disease can progress to a stage where it's not clear what, if anything, they understand of the world or themselves anymore. I can't speak for anterograde amnesia as I have no experience with that disease.
krageon 4y ago
Arrogance and belief in the superiority of humans, even absent any understanding of the source material. It's kind of depressing, because this entire thread is essentially a rehash of absolutely identical discussions I've had with first year AI students: Poorly informed, poorly thought out argumentation for why computers can't think (right now). The only correct conclusion is "I don't understand myself, so I cannot say anything about how similar something is to me". That's unpleasant and I suspect for a lot of engineers deeply offensive, but it's important to understand where your limitations currently lie.

What makes everyone in this thread so confident that it isn't sentient?

by cwkoss 4y ago

What makes everyone in this thread so confident that it isn't sentient?

At least personally, my subjective experience as a sentient being includes having a 3-dimensional "virtual reality" view of the world around me, which is mostly informed by sight but also by touch and sound. I think this is a key element of human & animal sentience and I tend to think that all animals that have this "3D representation of the world" in their mind are sentient. E.g. fish can solve mazes, in a way illustrative of having a 3D view of the world within their minds like we do, so I think they are sentient. https://www.researchgate.net/publication/251359525_Performance_of_Betta_splendens_in_a_radial_arm_maze https://www.researchgate.net/publication/251359525_Performan... If you search Google for the phrase <you see with your brain not your eyes>, you will find lots of blog posts and the like, explaining with examples which I mean by the 3D VR-like world in your head which I am describing. Another commenter in this thread raised the question of whether robots that navigate the world are sentient. I think the answer is possibly yes but I'm going to give a huge disclaimer which is I don't think that sentience is morally relevant, I think what is morally relevant is whether an entity feels qualia or not. https://en.wikipedia.org/wiki/Qualia https://en.wikipedia.org/wiki/Qualia In my personal view (just hunches really), I think that whether qualia are felt depends on the implementation. My hunch is that an implementation of a neural network in hardware, i.e. an actual network of transistors, might feel qualia, but that a software implementation of the same neural network would not. But others have different hunches and this is a question as yet unsolved by today's science.

tsimionescu 4y ago

Basic understanding of how the model works. As said elsewhere, the model doesn't have memory, it doesn't learn, and doesn't have goals. If you were allowed to play with it, you could quickly observe this for yourself with a bit of adversarial play - ask it about the time it fought Genghis Khan, ask it if it wants you to give it a glass of water, etc - you will quickly realize the illusion behind the system. Tell it whatever facts you want, be amazed how well it might recall them, and then start a new conversation and notice that it has entirely forgotten all of them (again, it has no memory, and it doesn't learn anything). From first principles, we know how the LM works: it has been trained to predict what is the most likely next token in a series of tokens. It has been trained to do this on a gigantic corpus, and is now pretty good at it. You then present it with a prompt, and it starts predicting the most likely token that comes after that prompt, and it keeps predicting for a while until the most likely "token" is to stop. Note that these tokens aren't even words - they are typically letters, or short combinations of letters. To give the illusion of a coherent conversation and memory, the prediction function is applied to the entire conversation so far, not just to the latest prompt. There is more sophistication to avoid various undesirable behaviors, especially in LaMDA (e.g. the model is also trained to evaluate the "safety" of various possible answers, so it doesn't generate a sentence that insults you or discriminates against you or induces you to do something dangerous etc), but this is the gist of how it works.

cwkoss 4y ago
Do you think all humans are sentient? This thread has me wondering if people with dementia, anterograde amnesia, etc. would be able to pass the proposed sentience tests others have suggested.
- tsimionescu 4y ago
  I'm not sure that I would classify humans in the final stages of dementia as being (fully) sentient. I have had the extremely sad experience of helping to care for two different grandparents with late stage memory loss, and while they had moments of lucidity, they had many periods where they were acting almost entirely randomly, or doing nothing at all, barely or not even perceiving the world around them. I struggled a lot to find signs of the person I had deeply loved in them, and there were many times when I can't say I found any signs of a full consciousness in them at all, as troubling as that was. To be very clear, this came after many years of milder forms of dementia, where their sentience could not in any way have been in question, even when forgetting who I was or where they were or other such things. But the disease can progress to a stage where it's not clear what, if anything, they understand of the world or themselves anymore. I can't speak for anterograde amnesia as I have no experience with that disease.

krageon 4y ago

Arrogance and belief in the superiority of humans, even absent any understanding of the source material. It's kind of depressing, because this entire thread is essentially a rehash of absolutely identical discussions I've had with first year AI students: Poorly informed, poorly thought out argumentation for why computers can't think (right now). The only correct conclusion is "I don't understand myself, so I cannot say anything about how similar something is to me". That's unpleasant and I suspect for a lot of engineers deeply offensive, but it's important to understand where your limitations currently lie.