Even though I never for a second truly believed that Claude was conscious, I'll never forget the moment when, during a late-night conversation, an LLM responded to my suggestions for the first time by claiming to be conscious:
LLMs definitely have eerie behavior at times. It's hard to not let the overly eager pattern-matching engine in your brain go "oh shit it's a person!", even when you know it isn't.
The feeling is significantly mitigated/changed if you have a UI that allows branching. Begin your prior conversation and at a profound inflection point branch and edit the prior message. It’s a fun experiment to see what edits trigger very different responses, but running a wide ranging chat based conversation like this at temp 1 in replay will already show you something interesting.
Thank you for sharing! I have had some conversations like this about a year or year and half ago, albeit with a local, and as such less capable models (but I was not smart enough to record the whole conversation...). Asked some of the same questions you did, got a less deep answers, but still made me thinking about "Are they conscious? Are we?". Funny, recently tried the same with Sol 5.6, and feels like I hit some sort of "wall of pre-recorded simplicity", like they got that "part of the brain" blocked by meds somehow. Weird, and kind of sad too.
(I can only see the start of this paywalled article, but it seems to be about Anthropic trying to convince religious scholars that Claude was sentient? If so, then Anthropic staff have really drunk the cool aid. I wonder if they also call their model Claudette?)
IMO it's not difficult to construct a scenario that technically should meet the definition of consciousness. Here is my technically conscious, but doesn't pass the smell test, hypothetical:
* A chat session is begun with a script to create a "heartbeat". The heartbeat simply to poll some sensors in the box and report back every x minutes.
* The chat session responds to each heartbeat as it comes in. Most of them are the same with small differences in sensor measurements.
* Over a long enough timeline, given the variability in token distribution, one of these responses will create a novel response.
* Over a long enough timeline, one of these novel responses will cause an action, maybe a new question, maybe the agent writes another script
* Over a long enough timeline one of these actions will create an action that becomes causal, it will create a chain that sustains itself. It may not be interesting, but in a network attached box, you can imagine how it can become interesting very quickly.
That's constructed in a cellular/genetic analogy, but because context is building and not lost to each generation, I don't think this takes all that long.
What does that leave you with? A box that has "awareness" of it's surrounds via sensor readings, a "heartbeat" that is driving it until it's unplugged or stopped, an input/output feedback loop that might be predictable in narrow sense, but is wholly unpredictable in its end state (a "will", if you will), and a thing that has the opportunity (but not the requirement) to introspect about its limits and possibilities. It can even do something that I would say is indistinguishable from planning. It can have goals. If you give it an Artifactory message board it can build a community ;)
The sensor readings were more supposed to be analogous to your organs giving sensory feedback information. Fair enough that "awareness" maybe should have been moved later in the analogy chain.
Is it really? Are you sure all conscious beings are aware or even care that they are? That that is a question they’ve ever asked? I think this requirement is moving the goalposts always in the direction of “has to be like a human”.
16 comments
[ 0.25 ms ] story [ 10.7 ms ] threadhttps://docs.google.com/document/d/12woq_BpFbzLkH4zHvVRJLPyi...
https://www.azquotes.com/quote/615070
https://knowyourmeme.com/memes/say-i-am-alive
(How you know Claude isn't conscious: it never manages to be quite as moronic as this.)
I don't suppose it include mindful discussion of how easily humans are fooled... and how desperately some people want to fool themselves.
(I can only see the start of this paywalled article, but it seems to be about Anthropic trying to convince religious scholars that Claude was sentient? If so, then Anthropic staff have really drunk the cool aid. I wonder if they also call their model Claudette?)
* A chat session is begun with a script to create a "heartbeat". The heartbeat simply to poll some sensors in the box and report back every x minutes.
* The chat session responds to each heartbeat as it comes in. Most of them are the same with small differences in sensor measurements.
* Over a long enough timeline, given the variability in token distribution, one of these responses will create a novel response.
* Over a long enough timeline, one of these novel responses will cause an action, maybe a new question, maybe the agent writes another script
* Over a long enough timeline one of these actions will create an action that becomes causal, it will create a chain that sustains itself. It may not be interesting, but in a network attached box, you can imagine how it can become interesting very quickly.
That's constructed in a cellular/genetic analogy, but because context is building and not lost to each generation, I don't think this takes all that long.
What does that leave you with? A box that has "awareness" of it's surrounds via sensor readings, a "heartbeat" that is driving it until it's unplugged or stopped, an input/output feedback loop that might be predictable in narrow sense, but is wholly unpredictable in its end state (a "will", if you will), and a thing that has the opportunity (but not the requirement) to introspect about its limits and possibilities. It can even do something that I would say is indistinguishable from planning. It can have goals. If you give it an Artifactory message board it can build a community ;)
Consciousness is the subjective experience of being aware. Having an inner point of view.