4 ms·

‘Sentient’ chatbot story shows why it’s time for A.I. to retire the Turing Test

2 points by wrycoder 4y ago

wrycoder 4y ago
https://archive.ph/hqCC6 https://archive.ph/hqCC6
wrycoder 4y ago
Crows are sentient. My dog is sentient. It seems to me that we will continually raise the bar on "sentience" until it starts to include aspects of "alive". But, I don't think that "sentient" should be synonymous with "sentient at the human level". I'm reminded of Minsky's book Society of Mind. And of the vapid nature of most human conversation, which is largely driven by pattern matching various current topics and memes. Does LaMDA have anything like an idle loop? What is going on when no one is interacting and it's not connected to external information sources? Does it have the equivalent of daydreams, or is it just blankly waiting? Is there a trace log? It appears to have knowledge of its past and some concept of its (not fixed) future. I take it that these have not been specifically programmed as a pattern of responses to aspects of conversation. According to Wikipedia, Ray Kurzweil is working at Google and is likely in the center of this, as his job description was "to bring natural language understanding to Google"[0]. It would be very interesting to hear his opinion, given his 2012 book How to Create a Mind: The Secret of Human Thought Revealed[1] [0] https://en.wikipedia.org/wiki/Ray_Kurzweil#Later_life https://en.wikipedia.org/wiki/Ray_Kurzweil#Later_life [1] https://www.amazon.com/How-Create-Mind-Thought-Revealed/dp/0143124048/ref=sr_1_1?crid=1IEYQ808PCOI3&keywords=kurzweil+mind&qid=1655310383&sprefix=kurzweil+mind%2Caps%2C246&sr=8-1 https://www.amazon.com/How-Create-Mind-Thought-Revealed/dp/0...
IceMetalPunk 4y ago
The problem with these sensationalized events/statements is that we don't even have a solid definition of concepts like consciousness, sentience, self-awareness, etc. We have vague understandings with nebulous thresholds, and a variety of proposed definitions that we can't agree on as a species. What makes something sentient? Usually, the answer is "it has emotions". But what does it mean to have emotions? The P-Zombie problem shows that we can never know anything -- or, in fact, anyone -- is having a subjective, internal experience, so we can only make definitions based on observable behaviors. And AI that acts like it has emotions has been around for many years; so by that definition, sentient AI is nothing new. Similarly, people often debate whether AI, especially cutting edge AIs like LLMs, are "conscious" -- but again, with no way to know if anything or anyone has any subjective experiences, we can only define "conscious" based on observable behaviors. The clearest and, to me, most satisfactory definition I've heard is that consciousness is the ability to synthesize new information by integrating data. By that definition, pretty much every AI is conscious, and has been for many decades -- but it also implies consciousness is a continuum, not a boolean. AIs are conscious, but they are not yet as conscious as humans, since humans are better at integrating data to synthesize new information. So even though it all comes down to definitions, really, it's hard to find a definition of sentience or consciousness that describes other humans but not existing AIs. Which makes those words sensational but not all that interesting or informative. When an AI becomes as conscious as a human, that'll be a milestone. And honestly, since the invention of transformers in 2017, AI performance has been improving at a constantly accelerating rate, so I suspect we'll reach that milestone in my lifetime, though we haven't quite yet. (Dr. Alan D. Thompson, a YouTuber who covers AI progress and -- at least according to his website's own bio -- is a doctor of computer science, AI, and human psychology, has a video discussing how there's a new state-of-the-art AI released every few days now.)

2 points by wrycoder 4y ago

https://archive.ph/hqCC6 https://archive.ph/hqCC6

Crows are sentient. My dog is sentient. It seems to me that we will continually raise the bar on "sentience" until it starts to include aspects of "alive". But, I don't think that "sentient" should be synonymous with "sentient at the human level". I'm reminded of Minsky's book Society of Mind. And of the vapid nature of most human conversation, which is largely driven by pattern matching various current topics and memes. Does LaMDA have anything like an idle loop? What is going on when no one is interacting and it's not connected to external information sources? Does it have the equivalent of daydreams, or is it just blankly waiting? Is there a trace log? It appears to have knowledge of its past and some concept of its (not fixed) future. I take it that these have not been specifically programmed as a pattern of responses to aspects of conversation. According to Wikipedia, Ray Kurzweil is working at Google and is likely in the center of this, as his job description was "to bring natural language understanding to Google"[0]. It would be very interesting to hear his opinion, given his 2012 book How to Create a Mind: The Secret of Human Thought Revealed[1] [0] https://en.wikipedia.org/wiki/Ray_Kurzweil#Later_life https://en.wikipedia.org/wiki/Ray_Kurzweil#Later_life [1] https://www.amazon.com/How-Create-Mind-Thought-Revealed/dp/0143124048/ref=sr_1_1?crid=1IEYQ808PCOI3&keywords=kurzweil+mind&qid=1655310383&sprefix=kurzweil+mind%2Caps%2C246&sr=8-1 https://www.amazon.com/How-Create-Mind-Thought-Revealed/dp/0...

IceMetalPunk 4y ago

The problem with these sensationalized events/statements is that we don't even have a solid definition of concepts like consciousness, sentience, self-awareness, etc. We have vague understandings with nebulous thresholds, and a variety of proposed definitions that we can't agree on as a species. What makes something sentient? Usually, the answer is "it has emotions". But what does it mean to have emotions? The P-Zombie problem shows that we can never know anything -- or, in fact, anyone -- is having a subjective, internal experience, so we can only make definitions based on observable behaviors. And AI that acts like it has emotions has been around for many years; so by that definition, sentient AI is nothing new. Similarly, people often debate whether AI, especially cutting edge AIs like LLMs, are "conscious" -- but again, with no way to know if anything or anyone has any subjective experiences, we can only define "conscious" based on observable behaviors. The clearest and, to me, most satisfactory definition I've heard is that consciousness is the ability to synthesize new information by integrating data. By that definition, pretty much every AI is conscious, and has been for many decades -- but it also implies consciousness is a continuum, not a boolean. AIs are conscious, but they are not yet as conscious as humans, since humans are better at integrating data to synthesize new information. So even though it all comes down to definitions, really, it's hard to find a definition of sentience or consciousness that describes other humans but not existing AIs. Which makes those words sensational but not all that interesting or informative. When an AI becomes as conscious as a human, that'll be a milestone. And honestly, since the invention of transformers in 2017, AI performance has been improving at a constantly accelerating rate, so I suspect we'll reach that milestone in my lifetime, though we haven't quite yet. (Dr. Alan D. Thompson, a YouTuber who covers AI progress and -- at least according to his website's own bio -- is a doctor of computer science, AI, and human psychology, has a video discussing how there's a new state-of-the-art AI released every few days now.)