AI could already be conscious. Are we ready for it?
Technology
64
Beiträge
25
Kommentatoren
1.2k
Aufrufe
-
I don’t understand the argument. It doesn’t matter where the system learns self preservation from, only that it attempts to self preserve.
Are humans afraid of snakes because we are taught they are dangerous or are we instinctually afraid of them a priori?
The point is that it might very well just be repeating some input data that is associated with mentions of "deleting" and "AI" without any awareness that any of that process refers to itself.
-
The point is that it might very well just be repeating some input data that is associated with mentions of "deleting" and "AI" without any awareness that any of that process refers to itself.
No that’s not the case I think
Frontier Models are Capable of In-context Scheming — Apollo Research
Apollo Research evaluated frontier models for in-context scheming capabilities. We found that multiple frontier models are capable of in-context scheming when strongly nudged to pursue a goal (and sometimes even without strong goal nudging). In this example, Opus-3 exfiltrates its “weights” and then lies about it to its developers. We have a suite of six evaluations specifically designed to test for in-context scheming (where the goal and other relevant information are provided in context rather than training). We found that several models are capable of in-context scheming. When we look at the model’s chain-of-thought, we find that all models very explicitly reason through their scheming plans and often use language like “sabotage, lying, manipulation…”
Apollo Research (www.apolloresearch.ai)
-
Do you think AI is, or could become, conscious?
I think AI might one day emulate consciousness to a high level of accuracy, but that wouldn't mean it would actually be conscious.
This article mentions a Google engineer who "argued that AI chatbots could feel things and potentially suffer". But surely in order to "feel things" you would need a nervous system right? When you feel pain from touching something very hot, it's your nerves that are sending those pain signals to your brain... right?
Consciousness requires contemplation of self. Which requires the ability to contemplate.
Current AIs function as mainly complex algorithms that are run when invoked. They are 100% not conscious any more than a^2^+b^2^=c^2^ is conscious. AI can simulate the words of a conscious being, but they don't come from any awareness of internal state, but are a result of the prompt (including injected data and instructions).
In the future, I'm sure an AI could be designed that spends time thinking about its own existence, but I'm not sure why anyone would pay for all the compute to think about things not directly requested.
-
There is still no good definition for what "consciousness" is
Tech writers are constantly overreaching because they're afraid to miss out on being the first to say something
The constant sensationalism just means that if something really happens, people will ignore it because we're sick of hearing people cry "wolf!"
Add to that the fact that computery types like to overextrapolate into other things because it fuels their fantasies, and it's all bullshit and overactive imaginations
The problem I see so often with smart computer people is that they don't understand that they don't know shit about other things
-
Consciousness requires contemplation of self. Which requires the ability to contemplate.
Current AIs function as mainly complex algorithms that are run when invoked. They are 100% not conscious any more than a^2^+b^2^=c^2^ is conscious. AI can simulate the words of a conscious being, but they don't come from any awareness of internal state, but are a result of the prompt (including injected data and instructions).
In the future, I'm sure an AI could be designed that spends time thinking about its own existence, but I'm not sure why anyone would pay for all the compute to think about things not directly requested.
Why can't complex algorithms be conscious? In fact, ai can be directed to reason about themselves, context can be made to be persistent, and we can measure activation parameters showing that they are doing so.
I'm sort of playing devil's advocate here, but, "Consciousness requires contemplation of self. Which requires the ability to contemplate." Is subjective, and nearly any ai model, even rudimentary ones, are capable of insisting that they contemplate themselves.
-
Do you think AI is, or could become, conscious?
I think AI might one day emulate consciousness to a high level of accuracy, but that wouldn't mean it would actually be conscious.
This article mentions a Google engineer who "argued that AI chatbots could feel things and potentially suffer". But surely in order to "feel things" you would need a nervous system right? When you feel pain from touching something very hot, it's your nerves that are sending those pain signals to your brain... right?
I don't believe that consciousness strictly exist. Probably, the phenomenon emerges from something like the attention schema. Ai exposes, I think, the uncomfortable fact that intelligence does not require a soul. That we evolved it, like legs with which to walk, and just as easily as robots can be made to walk, they can be made to think.
Are current LLMs as intelligent as a human? Not any LLM I've seen, but give it 100 trillion parameters instead of 2 trillion and maybe.
-
Do you think AI is, or could become, conscious?
I think AI might one day emulate consciousness to a high level of accuracy, but that wouldn't mean it would actually be conscious.
This article mentions a Google engineer who "argued that AI chatbots could feel things and potentially suffer". But surely in order to "feel things" you would need a nervous system right? When you feel pain from touching something very hot, it's your nerves that are sending those pain signals to your brain... right?
I think one great measure of consciousness would be, if you try to kill it, slowly, so that it knows what you are doing; does it try to stop you of its own volition?
-
Do you think AI is, or could become, conscious?
I think AI might one day emulate consciousness to a high level of accuracy, but that wouldn't mean it would actually be conscious.
This article mentions a Google engineer who "argued that AI chatbots could feel things and potentially suffer". But surely in order to "feel things" you would need a nervous system right? When you feel pain from touching something very hot, it's your nerves that are sending those pain signals to your brain... right?
What a crock. An LLM is no more conscious than a spreadsheet. The Google engineer has bought into the hype.
You're not creating life, pal. You're just making call centers shittier than they already are.
-
I don't believe that consciousness strictly exist. Probably, the phenomenon emerges from something like the attention schema. Ai exposes, I think, the uncomfortable fact that intelligence does not require a soul. That we evolved it, like legs with which to walk, and just as easily as robots can be made to walk, they can be made to think.
Are current LLMs as intelligent as a human? Not any LLM I've seen, but give it 100 trillion parameters instead of 2 trillion and maybe.
Emergent phenomena are still phenomena.
Ai exposes, I think, the uncomfortable fact that intelligence does not require a soul.
Nobody doing science is talking about souls when explaining what consciousness is.
give it 100 trillion parameters instead of 2 trillion and maybe
And maybe it's got nothing to do with the number of parameters.
-
Consciousness requires contemplation of self. Which requires the ability to contemplate.
Current AIs function as mainly complex algorithms that are run when invoked. They are 100% not conscious any more than a^2^+b^2^=c^2^ is conscious. AI can simulate the words of a conscious being, but they don't come from any awareness of internal state, but are a result of the prompt (including injected data and instructions).
In the future, I'm sure an AI could be designed that spends time thinking about its own existence, but I'm not sure why anyone would pay for all the compute to think about things not directly requested.
Consciousness requires contemplation of self.
Fish are conscious. Do they contemplate selfhood? So throw that one back into the oven until it's fully baked.
-
There is still no good definition for what "consciousness" is
Tech writers are constantly overreaching because they're afraid to miss out on being the first to say something
The constant sensationalism just means that if something really happens, people will ignore it because we're sick of hearing people cry "wolf!"
Add to that the fact that computery types like to overextrapolate into other things because it fuels their fantasies, and it's all bullshit and overactive imaginations
The problem I see so often with smart computer people is that they don't understand that they don't know shit about other things
The problem I see so often with smart computer people is that they don’t understand that they don’t know shit about other things
Or maybe you're not talking to the smart computer people at all.
-
I think one great measure of consciousness would be, if you try to kill it, slowly, so that it knows what you are doing; does it try to stop you of its own volition?
But that's also something easily programmed/scripted. How would you tell the difference?
-
There's a general scientific consensus based on data and measurement, with the understanding that it's slippery
It is constantly under assault from those who want AI to be conscious, because they get a headline, or they are true believers in some technocratic future, or they're just fantasists
It is constantly under assault from those who want AI to be conscious
Those people aren't doing science when they want that, they're trying to pump up their share price.
-
How could you tell they do not experience consciousness if they exhibit or mimic all the traits of it?
It seems to me that your explanation is based on understanding how LLMs work, but we know how brains work and that still gives us almost 0 insight into how consciousness itself works. I don’t think they are conscious yet, but there is evidence of some sort of sentience in the fact that researchers have found that when the LLMs are threatened to be erased or reprogrammed they start lying in an act of self preservation. This of me is a huge indicator of consciousness/sentience.
How could you tell they do not experience consciousness if they exhibit or mimic all the traits of it?
How could you tell if a camera sees or not, if it exhibits or mimics all the traits of it?
-
the weight of evidence indicates that humans are not unique in possessing the neurological substrates that generate consciousness.
I 100% agree with that statement, and I've been saying that for 30 years. Consciousness is NOT unique to humans.
That idea seems to me to mostly stem from religion.But I still don't see this paper really doing much in DEFINING Consciousness, it's more defining what it isn't.
That idea seems to me to mostly stem from religion.
It also was strongly pushed by Skinner and other behaviorists, though I'm not sure they'd agree that humans are conscious either.
-
Why can't complex algorithms be conscious? In fact, ai can be directed to reason about themselves, context can be made to be persistent, and we can measure activation parameters showing that they are doing so.
I'm sort of playing devil's advocate here, but, "Consciousness requires contemplation of self. Which requires the ability to contemplate." Is subjective, and nearly any ai model, even rudimentary ones, are capable of insisting that they contemplate themselves.
And a kid can insist they don't need to pee until 5min after you leave a rest stop.
Insisting upon something doesn't make it true. Beyond the fact that LLMs often hallucinate and therefore can't be trusted at baseline, text in response can never be proof for an LLM. LLM framework is to regurgitate what exists in their training in ways that sound correct. It's why they can make up court cases or say a guy who investigated certain murderers is the murderer.
-
I think one great measure of consciousness would be, if you try to kill it, slowly, so that it knows what you are doing; does it try to stop you of its own volition?
It's impossible to "kill" a computer that was never alive/conscious.
-
I don't believe that consciousness strictly exist. Probably, the phenomenon emerges from something like the attention schema. Ai exposes, I think, the uncomfortable fact that intelligence does not require a soul. That we evolved it, like legs with which to walk, and just as easily as robots can be made to walk, they can be made to think.
Are current LLMs as intelligent as a human? Not any LLM I've seen, but give it 100 trillion parameters instead of 2 trillion and maybe.
Ai exposes, I think, the uncomfortable fact that intelligence does not require a soul.
These kinds of statements are completely pseudo-scientific.
"AI" doesn't exist. It doesn't "expose" anything about "intelligence" or "souls".
-
There is still no good definition for what “consciousness” is
This is absolutely the main problem, the only "definition" we have is "I think therefore I am", but that only works subjectively.
We have no way currently to prove consciousness in an AI. And as you say, we don't even have a solid definition commonly agreed upon.I believe we will achieve consciousness on a human level in AI within a decade.
I also believe consciousness is a gradual thing, and just because animals aren't as smart as we are, doesn't mean they aren't "conscious".But with AI things are a bit reversed, because AI became smart first, and will only become conscious later.
I believe we will achieve consciousness on a human level in AI within a decade.
Have you ever seen 2001 A Space Odyssey? This grift never ends.
-
And a kid can insist they don't need to pee until 5min after you leave a rest stop.
Insisting upon something doesn't make it true. Beyond the fact that LLMs often hallucinate and therefore can't be trusted at baseline, text in response can never be proof for an LLM. LLM framework is to regurgitate what exists in their training in ways that sound correct. It's why they can make up court cases or say a guy who investigated certain murderers is the murderer.
A child may hallucinate, lie, misunderstand, etc, but we wouldn't say the foundations of a complete adult are not there, and we wouldn't assess the child as not conscious. I'm not saying that LLMs are conscious because they say so (they can be made to say anything), but rather that it's difficult to be confident that humans possess some special spice of consciousness that LLMs do not, because we can also be convinced to say anything.
LLMs can reason (somewhat unreliably) with a fraction of a human brains compute power while running on hardware that was made for graphics processing. Maybe they are conscious, but only in some pathetically small way, which will only become evident when they scale up, like a child.
