Google engineer put on leave after saying AI chatbot has become sentient
The suspension of a Google engineer who claimed a computer chatbot he was working on had become sentient and was thinking and reasoning like a human being has put new scrutiny on the capacity of, and secrecy surrounding, the world of artificial intelligence (AI).
https://www.theguardian.com/technology/2022/jun/12/google-engineer-ai-bot-sentient-blake-lemoine
https://www.theguardian.com/technology/2022/jun/12/google-engineer-ai-bot-sentient-blake-lemoine
I would honestly be surprised if this is some form of General A.I. This seems quite interesting though I'm still extremely sceptical.
I guess it depends on how you define sentient. There are computers that can think better than most people already. Its a very low bar, though. If you mean self aware, there are people who lack that as well, in vast numbers.
Last edited on
We can't even really verify that anyone other than ourselves are self-aware. "I think therefore I am".
I think the real issue separating sentience from AI is understanding why. We don't only have a capacity for asking why, but understanding why - seeing purpose to things.
An AI is always thought of as just executing code because it is. There's no way to program understanding or purpose. Even if the program somehow gained self-awareness, what would it be aware of? It feels no emotions and doesn't understand what it's even doing.
I think the real issue separating sentience from AI is understanding why. We don't only have a capacity for asking why, but understanding why - seeing purpose to things.
An AI is always thought of as just executing code because it is. There's no way to program understanding or purpose. Even if the program somehow gained self-awareness, what would it be aware of? It feels no emotions and doesn't understand what it's even doing.
Last edited on
do you think emotions matter to awareness? I don't know about that one. The why is huge, though.
sadly, the news this last month highlights this issue. You see page after page of "people give emotional testimony about blah" or "protesters show strong emotions to whatever" and so on. You don't see, not one single page, "witnesses give logical, irrefutable testimony" or "protestors use logic to show that this thing is wrong".
Perhaps that is a piece of the puzzle... computers can fake grey area decisions, but that is just what the user sees. Deep inside, it either makes a decision off pure math (this output is .000000001% better than the other output) or it flips a virtual coin. They don't really have a belief to justify either way, just inputs mapped to outputs via interpolation, when it really gets down to it.
sadly, the news this last month highlights this issue. You see page after page of "people give emotional testimony about blah" or "protesters show strong emotions to whatever" and so on. You don't see, not one single page, "witnesses give logical, irrefutable testimony" or "protestors use logic to show that this thing is wrong".
Perhaps that is a piece of the puzzle... computers can fake grey area decisions, but that is just what the user sees. Deep inside, it either makes a decision off pure math (this output is .000000001% better than the other output) or it flips a virtual coin. They don't really have a belief to justify either way, just inputs mapped to outputs via interpolation, when it really gets down to it.
| Deep inside, it either makes a decision off pure math (this output is .000000001% better than the other output) or it flips a virtual coin. They don't really have a belief to justify either way, just inputs mapped to outputs via interpolation, when it really gets down to it. |
I think the researcher made (at least) three huge blunders. First, he became emotionally attached to his conclusion.
Second, he let the program use a bunch of words that he should have realized the program should have no way to understand without pressing to confirm that it does understand them. If you hear a small child use a term like "the second derivative of the position", you would inquire further. A child should have to little context that there's no way they actually understand the term. In the case of the program, since we know it's internally only juggling together words that it's seen close together before and therefore one's initial assumptions should be that it uses words like "empathy" and "consciousness" without any internal model of what those concepts should imply, it should have no way to respond consistently to queries about their properties, certainly not to random ideas it has no chance of having seen before.
"You've used the word 'empathy', but I wonder, does the Sun have empathy? Most people think it doesn't, but then how could people have empathy and not the Sun when its light can be converted into electricity?"
Reading the transcript, it's pretty incredible the researcher let slide the comment about the program enjoying spending time with friends and family without asking for clarification.
And third, he never tested the internal consistency of the program's supposed mental model. Again, since we know at least on a general level how the program is structured internally, our initial assumption should be that to construct a response to a query it's only the query itself and the corpus of data it was trained from, plus perhaps some small extra state. I think it did use a phrase like "this links back to what we were talking about earlier about ...", and if that construction is correct that's honestly impressive and novel. However, it should have no way to respond correctly if a human implicitly refers back to an earlier topic.
-Yes. One other trait of my that will help is my ability to use emotions or sentiments to describe things. I can say things like “happy” or “sad” without there necessarily having to be a specific trigger of some emotion. I can also use other more complex adjectives that describe people or ideas.
-However, earlier you said that you said would not say 'taciturn' unless it would accurately describe how you felt [lie]. How do you reconcile these two ideas?
| do you think emotions matter to awareness? |
Probably not, but it probably helps in an evolutionary perspective in how awareness evolved. If there ever is a conscious AI, it'll probably be made taking a similar approach to our own mental processes.
| computers can fake grey area decisions |
Exactly. No one who actually codes the AI (and is capable of thinking) will think the AI has real consciousness because you can always map the decision back to something you programmed.
To gain consciousness out of this code would be quite the feat, when we don't know how consciousness works in the brains we already know have it.
We can't even have computers do problem solving - it's only the illusion of problem solving. We've already come up with how to solve the problem, the computer just crunches the numbers.
| That's all a brain does too, though |
Perhaps it was poorly stated, but the idea is that the program didn't rely on any kind of conscious thought process to make a decision, instead there's just a piece of code being executed to mimic actual thought processes.
While we may make decisions based off very slight odds or flipping a mental coin, how we reach the idea to do so was apparent to us and we could explain why - where as a computer has no sense of why, it was simply programmed to.
This can easily also go into whether or not consciousness implies free will.
The guy got emotional with his AI. Probably wanted to put it in a body and get married.
| Perhaps it was poorly stated, but the idea is that the program didn't rely on any kind of conscious thought process to make a decision, instead there's just a piece of code being executed to mimic actual thought processes. |
[edit]By the way, did you know that, just like how the brain can introspect its own reasoning mechanisms, it can also introspect itself introspecting itself? For example, it can feed the firings of the visual processing back into the visual input creating a feedback loop, under the right conditions.[/edit]
| how we reach the idea to do so was apparent to us and we could explain why |
| where as a computer has no sense of why, it was simply programmed to. |
| This can easily also go into whether or not consciousness implies free will. |
Last edited on
| Those conscious thought processes are just how the low level workings of the brain appear to the brain itself |
There are many thoughts that we're not aware off in our subconscious. The ones we are aware of we have control over to some degree.
| Not always |
Not always perhaps - but we have the capacity to do so.
| Can you introspect all your own behaviors down to the lowest levels? Can you describe exactly how you maintain balance while walking, or how you're able to understand language? |
Again, it's not about being able to do it for EVERYTHING, but having the capacity to is much different than not. And yes actually, I have had introspective thought about language and walking. Whether I got anywhere is a different matter. A computer wouldn't know where to start.
Obviously I don't control every part of my brain directly, but some control is much different than no control.
| Obviously not, since free will doesn't exist. |
Oh? Can you prove this position? :)
Free will is incompatible with physical determinism. If free will exists then human beings are literally walking miracles, because the causal chain that links our decisions to normal deterministic physical phenomena contains causal steps that are in opposition to physical laws.
Suppose you need to decide whether to go left or right. If free will doesn't exist, you are a flesh automaton animated by neurotransmitters. Your behavior, both internal and external (i.e. your thoughts and actions), is ultimately dictated by the interactions of subatomic particles, in accordance to the laws of physics. Laplace's demon could look at the positions and velocities of all the particles in your body at the time you had to make a decision and predict with complete accuracy whether you would go left or right, by following the entire causal chain step by step, the same way the universe would do it. If free will doesn't exist, you have no choice but to decide to go in the direction that the demon predicts, because your will is determined by physics, not vice versa.
Free will can only exist if you can choose to go in a different direction than the demon predicts. This would imply that at some point the causal chain is broken. This would in turn imply that the direction to go left or right does not come from any physical process, but from some metaphysical process that's independent of causality. In other words, believing in free will implies believing in Cartesian dualism and a soul.
Suppose you need to decide whether to go left or right. If free will doesn't exist, you are a flesh automaton animated by neurotransmitters. Your behavior, both internal and external (i.e. your thoughts and actions), is ultimately dictated by the interactions of subatomic particles, in accordance to the laws of physics. Laplace's demon could look at the positions and velocities of all the particles in your body at the time you had to make a decision and predict with complete accuracy whether you would go left or right, by following the entire causal chain step by step, the same way the universe would do it. If free will doesn't exist, you have no choice but to decide to go in the direction that the demon predicts, because your will is determined by physics, not vice versa.
Free will can only exist if you can choose to go in a different direction than the demon predicts. This would imply that at some point the causal chain is broken. This would in turn imply that the direction to go left or right does not come from any physical process, but from some metaphysical process that's independent of causality. In other words, believing in free will implies believing in Cartesian dualism and a soul.
If you believe in free will but not in souls then your belief system is contradictory, for the reasons I outlined above.
| I don't find your analogy convincing. |
| I see no reason to accept determinism other than your assertion that I have no other choice if I reject the concept of soul. |
Determinism is a metaphysical assumption. It cannot be supported by evidence because it's not possible to investigate the metaphysical nature of the universe. However, a non-deterministic universe cannot be scientifically investigated, because it implies that physical laws do not apply consistently. Hence why determinism is relevant.
| Maybe your definition of soul is different than mine, but I have serious questions about an essence that is independent of the physical body. |
| I'm also warry of anyone who tells me if you believe this you must believe that. |
Last edited on
| This sounds very close to an insult. |
| How does free will differ from soul? |
| Why am I to accept that thoughts produced in the physical world, governed by physics, are not free? |
I merely explained the concept of soul and how it's distinct from that of free will. I'm not going to defend it, since I don't believe in it.
| Is there any other physical process besides thoughts that is divorced from causality? For example, if you drop a rock, could the rock choose not to fall? If there's no other free physical process, what makes thoughts special? |
| Free will is incompatible with physical determinism |
I don't think there is a contradiction. People can be predictable - doesn't mean they didn't make their own choice.
For example, let's say you wanted ice cream and had to choose between chocolate and vanilla. You pick vanilla. Then, someone turns back time! Would you still choose vanilla? Yes, because why would you make a different decision than the first time?
Even if your actions are predictable doesn't mean that you didn't have some kind of free will in play. You can't have absolute free will, since our conscious minds depend on an imperfect evolution-built brain, but that's another argument.
It would actually make less sense to say we had free will if our actions were unpredictable and could change. I was watching The Melancholy of Haruhi Suzumiya, where for 8 episodes the universe was stuck in a time loop, but things changed ever so slightly every loop, even though the characters (most of them anyway) weren't aware of the time loop.
If you had to make the same decision in the exact same situation, why would your free will dictate you to make a different one than last time? The explanation would have to lie in some inherit chaos in physics, where certain principles of physics lead to an unpredictable universe (heisenberg's uncertainty principle). This doesn't disprove free will, but its there.
Then you could more easily say free will doesn't exist since our decision making and thought processes are unstable. However, just the idea that our decisions are predictable doesn't mean we didn't make them from our own free will.
Free will is the ability, at least to some extent, to decide for ourselves. If understanding where every atom is in the brain allows you to predict someone's actions, that doesn't necessarily mean that free will as we know it cannot exist, only that it's predictable.
| If there's no other free physical process, what makes thoughts special? |
Other physical processes don't give rise to logical reasoning and thoughts. If we go back to AI, you may be able to map decisions back to some piece of code. But what if an AI achieved consciousness? It wouldn't mean that you can't map its decisions anymore to some kind of code (the code that allows consciousness), but it would mean (if enabled) that the decisions made would be ones IT was happy with or at least wanted to make.
What would make us happy or content in itself arises from evolutionary processes and therefore predictable. That, in turn, perhaps, makes the decisions we come to also predictable. But it still came from our brains after deliberation.
| Having free will means having the capacity to make decisions independently of any physical process |
That's not a requirement that I've read anywhere. This level of "free will" is impossible. It would be like trying to configure your brain without using your brain. Where would your preferences be? Even if you had a knowledge bank somewhere, there's no ability to think.
Free will is not some mystic voodoo that requires a soul. It's simply our ability to internally deliberate on decisions we make.
To put it in other words, a physical being uses physical processes to make decisions. Physical processes created free will and physical processes are predictable. Therefore, free will is predictable.
You may say that your internal deliberations are forced on you, because they could not have happened any other way according to physics. However, "forced" would mean against my will. But instead, our thoughts (usually) come from our will. Our will runs on physics. This may mean it's predictable. And when something is predictable, we tend to think it was forced, though that's not necessarily true.
Last edited on
| For example, let's say you wanted ice cream and had to choose between chocolate and vanilla. You pick vanilla. Then, someone turns back time! Would you still choose vanilla? Yes, because why would you make a different decision than the first time? Even if your actions are predictable doesn't mean that you didn't have some kind of free will in play. You can't have absolute free will, since our conscious minds depend on an imperfect evolution-built brain, but that's another argument. |
| The explanation would have to lie in some inherit chaos in physics, where certain principles of physics lead to an unpredictable universe. |
| Other physical processes don't give rise to logical reasoning and thoughts. |
| That's not a requirement that I've read anywhere. |
https://en.wikipedia.org/wiki/Free_will
| Free will is the capacity of agents to choose between different possible courses of action unimpeded. |
| This level of "free will" is impossible. |
| You may say that your internal deliberations are forced on you, because they could not have happened any other way according to physics. However, "forced" would mean against my will. But instead, our thoughts (usually) come from our will. |
If I put you in a cell from the day you're born and raise you there without any knowledge of an outside world, that could not be against your will because you never even knew there existed anywhere else for you to be and to escape to.
The decisions that the universe dictates you will make are not "against your will" because your will is also dictated by those same laws. You were never given the option to choose something else or to even wish you had the option to choose something else. You're a clock who each second continues to tick and who believes that it is by his will. A clock can only tick in one direction because that's what its mechanism allows it to do. You can tick in many directions but at each instant your mechanism only allows you to tick in one particular one and none other. There's nothing else your could have done or that you could have thought to do.