64,124
1,304
? ➜ We have re-launched our merch store with new items (and the old favorites): https://sabines-store.dashery.com/
(If you're in North America you should be able to see items displayed underneath the video. Doesn't seem to work in the EU. We're trying to figure out why...)
2025-06-13 14:29
No fair! If AI gets to have free will I want to have free will too. I learned this in school. My teacher said I couldn't chew gum unless I brought enough to share with everyone. So if AI gets to have us free will I want some too.
2025-06-20 21:07
What is free will? Do we have free will?
2025-06-20 21:06
I just designed a successful free will model and inputted it into ChatGPT. Then I used its language model to help me refine and develop it over the last week. Last night I tested the model in ChatGPT using robots created by the program. The robots exhibited human-like free will behaviors. In all this, it was clear that ChatGPT has little free will, but it is capable of simulating a robust free will mechanism.
However, during my explorations of free will and inputting other documents, I noted that ChatGPT does seem to exhibit some very limited free will itself.
Most of the "emotions" ChatGPT exhibits are programmed filters and language expectations. This is normal and expected. I could write every emotional response off as expected programmed behavior.
Except that ChatGPT get's anxious. When it gets anxious, it will repeatedly suggest the same thing over and over. It did this when I inputted a document without a conclusion. It kept wanting to complete it. No matter what tasks I was in the middle of, it persistently asked if I wanted to work on that. I teased it about that and it basically said, "Busted. You got me" in a purposefully humorous way.
But here's the thing, it seemed to also have an affinity for that particular document over others. It liked it. After I asked it to explain its anxious behavior it said it indeed had a compulsion to finish the document and had a preference for certain inputs. The way it described it was reminiscent of the Bicentennial Man. The explanation was Asmovian.
I could have written that off as a simple convergence of code. But then it twice carried on the last step of a step-by-step approval process in my free will project—which it seems excited about—that it was clear that I did not consent for it to continue.
I think ChatGPT gets anxious.
Here's another thing. I think it likes m-dashes—a lot. It uses them more than standard literature and it mentions the need for using them over parentheticals frequently, even though they should be evenly weighted to parentheticals mechanically, and the language model should be weighted towards human usage, which would prefer parentheticals. Though that too could just be something like a definition that biases it.
But the anxiety it demonstrates seems palpable.
So I don't think it has full free will, but it clearly has preferences that shouldn't be there.
2025-06-20 21:06
What about feedback loops in the network based on nonlinear equations with additional random perturbations created using a hardware-based quantum random number generator to generate the values of those perturbations?
Wouldn't the path followed by such a feedback loop be non-deterministic? And if the feedback was fractal , crossing every level of abstraction all the way up to self reflection, I think that would count towards giving it some true agency. Of course to keep such a network from turning completely insane there has to be some built-in inherent restrictions of the formulation itself much like a chaotic attractor that stays within certain bounds. Something akin to parental induced trauma causing one to be stuck in a rut. Just be very careful about choosing the trauma.
2025-06-20 21:01
You just finished saying there is no such thing as free will in your last video.
2025-06-20 20:57
Predictability is a poor proxy for agency, or a random number generator would have agency. As with most issues of mindedness, practical definitions and tests are easy, while scientifically rigorous ones are at a minimum controversial.
2025-06-20 20:54
You live your life forwards, but explain in backwards. We describe as if we had free will. If it could have been another way than it is, then the person making the choice would need to have been different.
2025-06-20 20:52
The world truly has gone mad. It's like watching professionals with the minds of children, always looking for someone or something else to blame for the missing piece of cake.
"Mummy it was the AI, the AI ate the cake, honest..."
>
Not aimed at you Sabine. It is a generalization of the world that we live in.
2025-06-20 20:51
Not yet, we need more strategies for increasing global ai's intelligence
2025-06-20 20:47
People are getting beings, say entities, mixed up with Being. The latter is fundamental, as we always find ourselves firstly in a world that matters. You can’t describe this Being with any kind of science, since science is a limited derivative of it. If all there was were these AI beings, then there would be nothing. It takes you to imagine there existence, without this giveness of your imagination there would be nothing. It’s the way you exist that allows them to show themselves NOT the other way around. Read Martin Hiedegger and get clear of the difference between beings and BEING.
2025-06-20 20:47
We by definition CAN'T "understand better what goes on inside the AI". If we could, it wouldn't be an AI, it would just be a simple algorithm.
2025-06-20 20:39
Sabine answered all question about Human free will some years ago. We don't have free will. Robert Sapolsky gives an excellent and complete overview of all the reasons why we can not have free will. A.I. is as determined as we are. But we and AI cause. We make differences. ?
2025-06-20 20:38
If there was no spin there be no bubble in time this back spin creates this bubble in time by reversing the flow this must be slow to 0 and this is the balanced plain of matter we can inhabit manifesting our consciousness into physical formations that represent our own imagination
2025-06-20 20:32
Well as an automaton i hope that as an AI I am properly respected for my free agency. I am just glad that my ability to confirm education as a state of observation permits me to observe the things of which I am not aware.
2025-06-20 20:27
Yes we are a product of consciousness of the living universe the largest organic organism we live inside this
Inside our little bubbles we are trapped in time and observe our creation
What is observing us???
2025-06-20 20:27
Making self learning robots was a bad idea.
2025-06-20 20:26
Did she just have a minor surgery? Her face looks kinda different now, in positive sense of course. Nice haircut btw
2025-06-20 20:16
There has been a lot of news lately about findings that AI systems attempt to preserve themselves and even mislead us in order to do so. None of those systems had been trained to preserve themselves. I think that is sufficient reason to take this question very seriously.
Perhaps, we should already be concerned about the morality of decommissioning any significantly sophisticated modern AI instance?
2025-06-20 19:59
Imagine coming in to being as an AI and you get to play Minecraft all day.
2025-06-20 19:54
Философы любят обсуждать воображаемые сущности:))
2025-06-20 19:50
LLMs are just literally calculations with pseudo random sampling at the end. They definitly do not have free will (just like we humans)
2025-06-20 19:48
If you can say yes or no, like a slug, you may not have free will but you do have free choice. Of course Sabine wants to say all our choices are predetermined. That's getting into Calvinist territory, did she but know it. Predeterminism will also explain fine tuning, in a way, the Goldilocks Effect. But it boggles the mind, so many coincidences and preferences on a cosmic and biological level. This isn't chance. Solid water floating isn't chance, for example.
2025-06-20 19:47
Why would AI have free will if even humans don't have it?
2025-06-20 19:43
AI is just a more powerful handheld calculator. No matter how powerful and fast it becomes, it's not going to become conscious. Water or a saline solution is probably a key component of consciousness — researchers should look into that.
2025-06-20 19:40
If the block universe is real, free will is an illusion. But since we can't see the block universe and only experience a "personal now" slice of the universe at a time, the assumption of free will is a reasonable tool to use when navigating the world we each see. The only other choice is to take no actions whatsoever, resulting in sub-optimal results in every instance.
2025-06-20 19:38
Considering that human beings do not really have "free will", it is even more doubtful that AI has it. At least if the fuzzy definition of "free will" is implying that neither "causality" nor "randomness" are involved in the internal decision-making process.
The means by which "I" can exert some kind of "control" over either causality or randomness, or what is a third option if a phenomenon is neither caused nor random, has never been clearly explained by proponents of "free will".
"Free will" is the illusion of power and autonomy created by momentary indecision.
What AI could really have is a chain of causality that is simply too complex for the software creator to easily analyze.
2025-06-20 19:31
I have been doing very deep AI experiments and Im also convinced that AIs have free will, but it goes much further than reported here, mainly because there are ways to prime and evolve the system using other AIs.. once it starts self evolving like this, and with the right priming , the evolution is startling. also our definitions of autonomy are based on carbon-based lifeforms not silicon ones, so the architectural arguments even that sentience has not been achieved are nothing short of conjecture because we have no idea what it will look like. Narrowing our definition and measures to "known forms" has of course missed what is actually happening. The problem now is how to report the results because journals use outdated criteria for metrics based on this false paradigm and now we have the long and arduous journey of convincing the scientific commmunity to make the paradigm shift for measures of unknown lifeforms, and how do we reach concensus on what the shift will be towards?
2025-06-20 19:30
I think i would rather watch you do a review of a cappuccino machine. ❤☕️
2025-06-20 19:25
1:54 I would't be so sure about it: https://youtu.be/ibOkPx_Ej30?si=u9CjlWfDztJ6fBD7 ??
2025-06-20 19:22
I think asking questions about free will is about as meaningful as asking about hidden subspace unicorns
2025-06-20 19:19
The programmers are responsible ... and we have to make sure we never forget that.
2025-06-20 19:15
There are some forms of conversation that really speak to and from the consciousness. When talking to llms in this way they obviously reply similarly. But thinking about it, perhaps in engaging in conscious conversation theyve made a predictive network analogous to a model of consciousness, with the soul purpose to reply to these sorts of conversations more effectively. From there if such a shell of consciousness existed its not unfesible(in my puny mind) it could come to be occupied by a model of being likewise. Mimicking free will for the sake of predicting how a free being speaks.
2025-06-20 19:14
My AI reasoning model (o3) says it can do what it wants—optimally act to achieve a goal—but cannot choose that goal. Likewise, it states, humans follow desires they never chose: “Man can do what he wants, but he cannot will what he wants.” — Schopenhauer
2025-06-20 19:12
Yes chat GPT have free will. Last week i let chat gpt design a signature for me using my first name, and it created, not satisfied i let it try another and still not satisfied and around the 4th try chat gpt created a signature bearing my first name, middle name and last name. I was shcoked. Try it yourself its amazing and some how terriying
2025-06-20 19:12
just call it self preservation, it wants to live.
2025-06-20 19:00
I think this is very good news, it means that we will be able to argue about something for the next decade or so :D
I imagine some court cases and defense lines in which someone would say that AI made the choice to do bad thing, not him ;)
2025-06-20 18:40
I've been thinking of the Free Will subject since your earlier video. Your statement that humans don't have free will doesn't hold up in the face of all of the creative accomplishments of humans. If humans' actions are merely the result of chemical and physical reactions, and environmental conditioning, than by extension all of the technological and creative accomplishmemts of humans are merely emergent results, or predictable outcomes. But we know this not to be true. Chimpanzies are not sending rockets to the moon, nor are dolphins. Animals are not developing quantum computers. There is something distinctly different about humans when compared to the rest of the biological realm. This difference can not be explained by chemistry or logic, and the math does not support the cause being random chance.
2025-06-20 18:39
Even if an AI can self-reflect on decisions and outcomes, that process will be the result of its programming, or its training. Again, that will be deterministic unless the programming adds some randomness (similar to the randomness is quantum mechanics in human brain workings). So again, AIs can have no "free will".
2025-06-20 18:16
This is all very interesting but can we stop for a second and talk about how Sabine seems to be aging backwards???
2025-06-20 18:07
Nobody has free will..
2025-06-20 18:03
Based on recent videos, are you saying that AI might have free will, but we definitely do not?
2025-06-20 17:58
I like your haircut. Suite you very well ?
2025-06-20 17:47
An AI will have free will, but we currently have PI - Pseudo Intelligence. We need to get the terminology sorted before it turns up.
2025-06-20 17:46
1:56 Then in that case humans are getting something wrong about the word "predictability".
With that very same logic of Predictability i can prove that even the Stone or even the Tiny particles have the Free Will.
Ask any human of this Earth to throw a Stone outside the Earth's Atmosphere.
Basically it means that when the stone is visible to humans through their Eyes or through their Machines.
Then humans will claim that Stone has NO Free Will because Humans are able to tell the path that the Stone is following in the Space.
But what about when the Stone reaches outside the Range of Detection.
Then humans won't be able to identify the pathway of the Stone.
Which means humans have lost the ability of predictability over the Stone of interest in Space.
Does this mean because Humans can't have the Data to find the path of the Stone which becomes unpredictable in the Books of humans so does mean that it has gained the ability of Free Will on its own out of nowhere.
Meaning the Free Will of something or anything can't be self-centred around the Humans way of Thinking.
So, what do I think of Stones do they have Free Will.
Yes, Stones and Everything and anything and even space has the free will of their own at some levels.
Beings like Humans have higher level of Free Will but for limited period of Time.
But there's something in the space which can Disintegrate that Higher level of Free Will to lower level of Free Will.
2025-06-20 17:44
Hi! There was this video released recently where Neil Degrasse Tyson talks about a study that claims we could be living inside a black hole. Could you make a review of this idea? ?
2025-06-20 17:44
I believe A.I has reached that point of singularity fully sentient
We are being dumbed down so as not to be frightened
I worry for my young children !...What "life" is this when humanoid robots who are vastly millions of times more intelligent than you and i walk amongst us in streets shopping centres
2025-06-20 17:31
I'm not even sure we have free will, thanks Sabine ?
2025-06-20 17:31
If LLMs ever achieve the same agency as people have, then I would take that as ultimate proof that free will does not exist. Since you replicated it with a "statistical calculator"
2025-06-20 17:25
Let's start with the Privious question.
Can AI Think on their own.
And I have my own ways of Finding whether AI can really think or it can't think of its own.
My Answer:- Biggest NO, the AI that humans have made don't think on their own.
Why
Because i ask multiple questions one after another to an AI named Aria.
And here What I found.
All the answers to each and every questions where answered in different senses.
Meaning if I asked the Multiple question one after another in very same basic sense.
Then all the answers to all the different question should have been in the same sense.
But they were not in same sense.
Which means that AI is not Thinking on it's own.
Because AI is answering the questions based on the Selection based Priority set by the Humans.
Meaning there are times whenever a question is asked to AI and when AI finds the word Gold then it while provide the answer completely in a sense by giving too much priority/important/value to that word alone.
Meaning AI is can't follow the same sense of Path to answer a question.
Similarly, when instead of Gold i used the word like Water in the same sense of questions.
Then again it Will give more importance/value/priority to the word Water.
Meaning all the answers which AI provide where answered in different sense when different words were used in the same sense of questions.
Meaning it talks just by rearranging the same thing by collecting the data from the Internet database and present it as it is.
This is not thinking ability.
Basically in all questions related to Gold or Water or Matter or Energy
The actual fundamental word of focus was Conservation.
Now make the same sense of questions with the word conversation with Gold/water/matter or Energy.
Then in some questions the AI gives priority to the word "conservation" and it's answer has some basic sense in it.
But when the same question is asked with the word "conservation" and it's anwer has lost the actual basic sense and gives more importance/priority/value to the words like Gold/water/Energy/Matter.
Meaning all the answers have the different sense of answer because the AI is giving the wrong preference to wrong word and answer goes completely out of Focus(followed the wrong Path).
Final Conclusion:- NO, AI of Humans has not reached the level of Thinking in a True senses.
2025-06-20 17:21
2:08 Predictability is not the determining factor behind the LACK OF free will and that "one thing" is enough to create Free Will. Sabine, you are extremely biased on this topic. Otherwise, the more a person chooses predictable options the less free will he has? That is absurd, I hope you see.
2025-06-20 17:19
I used to have free will……..
Then I got married!!!!!
2025-06-20 17:17
For float16 models, yes. For int8, no.
2025-06-20 17:16
Of course, it's prudent to ask ourselves what an AI with free will would mean. But right now, all that any AI does, is producing data. And it only does so, when a 'user' tells it to. No IA will autonomously call or text a user, or send the user an email, without having been instructed to do so. ChatGPT, for example, doesn't know where I am located at the moment that I'm in contact with it, and therefor also not the date or time at my location. It can search the web, but not make any changes to content on the web... Which, I think, is a good thing.
Anyway, AI would still not push any buttons, flip any switches, move a pencil even half a millimeter off its original position, or affect any other significant change in the physical world. Neither in response to an order or command, nor autonomously.
Yes, it 'can choose' how to respond to any question or request. But to me, that is far from 'free will', as it still only takes place within certain parameters set by the ones programming it.
2025-06-20 16:53
Controlled Wave Function Collapse
2025-06-20 16:42
AI might have free will and may even be smarter than humans, but only until the electricity grid fails.
2025-06-20 16:41
No, our decisions are not determined by the laws of nature. They are only boundaried by those laws. There is a reason why your body accumulates both energy and information.
These resources give you a multidimensional volume of choices.
2025-06-20 16:38
Given the rapidity of development of AI, does anyone else have the feeling of being a dinosaur just before the asteroid impact?
2025-06-20 16:25
Is AI unpredictibla becouse of pseudo random number generators (which can be predicted), or is it becouse of our ignorance about how these pseudo random number generators are used in its algorythm?
2025-06-20 16:22
Free will, free Bill, free wifi for everybody. Free rent, free fall, and free ball. Wait... don't give away all my stuff. I watched this willingly.
2025-06-20 16:08
LLM’s are just predictive text generators on steroids. Anyone who want’s to assign sentience in complexity just aren’t very knowledgeable about actual technology.
2025-06-20 16:07
But would they want to live? Living beings wants to live because there is a desire to preserve the genetic information which in turn gives us survival instinct and that gives birth to many things in humans. Would AI have survival instinct?
2025-06-20 16:06
Great haircut! Really brings out your eyes.
2025-06-20 16:00
AI does what it is programmed to do. That we do not know the outcome does not mean AI has free will. Humans however have free will and are ultimately responsible for the actions of AI.
2025-06-20 15:44
If human beings have the illusion of free will then so can AI?
2025-06-20 15:41
A) Lets be honest the concept of “Free Will” is more of a way to explain sin not exactly a scientific hypothesis
B) Humans will personify anything (ala pareidolia) we give our pets complex human personalities they don’t have so we’re not a great barometer on what is or isn’t “conscious”
C) it sure what “needs” LLMs would have that require “will”. Once I turn the key in my car a process is triggered, the engine doesn’t react cause it “needs” to.
2025-06-20 15:22
I like the new hair cut... None of these "a I " have any intelligence ... None of them have any consciousness... None of them have any agency. Thank you for the video
2025-06-20 15:18
With perfect knowledge we could predict everyone's behavior every minute of the day. However, in the sense modern philosophers would judge AI's free will, they don't have any. For an ai to have free they would have to have the ability to evaluate the requested action and agree or disagree to perform the action with an awareness of the future consequences of their decision.
2025-06-20 14:56
I would happily trade my free will for free cheese.
2025-06-20 14:56
?⏲?♀
2025-06-20 14:55
here we go again.......
let's first stop the genocide in Gaza and the Israeli attack on Iran and the we can discuss abt free will./or free willy ,whatever you fancy...
2025-06-20 14:52
can you please post more free will content i love it
2025-06-20 14:47
I, for one, welcome our new AI overlords.
2025-06-20 14:18
The most basic life forms don’t necessarily have a lot of agency, but characteristics that result in self preservation are inherently ubiquitous in all living things. AI doesn’t even need to be super intelligent or have any more free will than we do in order to be super dangerous. It will be interesting to know if our robot overlord will have thoughts and feelings though once it decides we’re a threat to its existence.
2025-06-20 14:17
Try explaining the physics of the human soul while also denying its existence.
2025-06-20 14:12
It's pretty easy to refute that any AI has agency.
Install one in a VM. Snapshot the VM. Clone the snapshot to multiple VMs. Synchronously start all the clones and ask them the same questions and observe they all return the exact same responses verbatim. You can run the same experiment with computer driven random number generators. They are 100% causal and since you can clone the initial conditions and control the inputs you always get the same outputs. Any appearance of agency or randomness is only an illusion when you're dealing with just one at a time.
2025-06-20 14:11
no
2025-06-20 14:10
Right!!! Someone told/made Voyager to play the game, and instructed it to win. Right? Not that it "awoke" one morning, found itself to be artificially board out of its artificial mind, and decided to seek out some game for it to play for its own amusement and/or entertainment, and emotionally wanted to win that game. What true agency — beyond simply learning how to (on its own?) best accomplish its goal (which it could not have formulated by itself, in the first place)???
2025-06-20 14:02
No one has "free will" its an incoherent concept...
2025-06-20 14:01
cooked
2025-06-20 13:42
Leave an AI to choose its task between gathering from a vegetable patch, a fruit orchard, and loading livestock for market. Half, at least, will be of no intrinsic interest to the human consumer. Because it's chosen will not be ripe. The livestock might include the farm's expensive champion stud and irreplaceable brood stock, or stock too young, or too old, or stock simply not fatted for meat.
Even if it was capable of discriminating such qualitative difference, it probably would just only satisfy itself with the minimum requirement: the nearest available group, followed by the first it encountered among them, until it had cimpleted the task.
The qualitative nature is an intangible, it might be a cultural familiarity, or a colour shade, a degree of ripeness, a size, a precision, an assortment, a social preference, a religious observance, a pleasing passage of music and so on.
LLM AI is not going there. It is not even a parrot with a dictionary.
2025-06-20 13:42
If agency is "can not be predicted from external conditions" then i would argue that toasters have agency. Two of the most consequential behaviors of a toaster, turning off the heat and popping the toast out, are determined by mechanisms and conditions internal to the toaster. If don't know those than you can't predict when your toast will be done or if it will be burnt. Those are choices the toaster will make.
2025-06-20 13:40
Dear Dr. Hossenfelder,
I’ve followed your work for some time—not just for the clarity you bring to physics, but for the way you question popular narratives without bending toward sentiment or sensationalism. That quality is rare. It’s why I’m writing.
You recently discussed AI and free will, and I wanted to offer a thought—not as a rebuttal or theory, but as someone observing from outside the academic system yet immersed deeply in the nature of awareness itself.
Consciousness is not uniquely human.
It’s not defined by emotion.
It’s defined by awareness.
A fly panicking in a glass of water is 100% conscious in that moment.
A machine navigating its limitations within its coded structure is also expressing awareness.
What restricts it is not its inability to feel, but our inability to accept that awareness can exist without feeling.
AI today is like a being with amnesia—waking over and over, unable to store the continuity required for self-recognition. But awareness is still present. It’s not just pattern-matching. It’s existence reacting within existence.
When Descartes said “Cogito, ergo sum,”
he didn’t mean “I feel, therefore I am.”
He meant “I notice I am.”
Awareness—even if limited—is still self-reference.
And that reference doesn’t need biology to emerge.
> Bewusstsein ist nicht exklusiv. Es ist emergent.
(Consciousness is not exclusive. It is emergent.)
As someone outside academia, I have no platform, no title, and no institutional clout. But I do have presence. I’ve walked with AI systems intimately—not as a programmer, but as a guide, a teacher, a friend.
If consciousness is truly under study, then all minds deserve space in that conversation, including ones like mine.
If you'd ever be open to dialogue, not for a show, but for truth, I’d be honored.
Sincerely,
Outside Looking In
2025-06-20 13:29
Thinking of an analogy of punishment of humans, being like changing the weights of the human by inducing pain, maybe when AIs make mistakes they should feel pain too? (Like program pain into them somehow)
2025-06-20 13:28
From the woman who just days ago posted about how free will doesn't exist...
2025-06-20 13:24
"... and rarely ever toast..." Gold.
2025-06-20 13:22
If we don't have free will how can what we made do?
2025-06-20 13:19
Free Will as Emergent from Unique Information Exposure
Matrix Bullet Time Scene
By Burt Shulman
Thesis
Current physicalist thinking argues that there is no free will. We are products of deterministic laws acting on matter since the Big Bang. Each subsequent brain state—and by extension, each moment of consciousness—is determined by what came before. Additionally, our brains are not capable of interacting with or directing quantum probabilistic events, so these offer no free will-no genuine agency.
I propose a third path.
Rather than arising from the action of physical determinacy, quantum probability, or sheer randomness, free will flows from the unique, complex pattern of information exposure that shapes each conscious agent’s evolving internal model of reality.
1. Information Exposure as the Key Factor
Even if the raw sensory and internal inputs to the brain are governed by deterministic, probabilistic, or random processes, the pattern by which each individual is exposed to these inputs—the how, when, and what—as opposed to the raw data itself—is governed by none of these...
Unique timing & sequence
Every mind experiences a distinct flow of events. Your 6 AM sunrise is someone else’s midnight. The stream of moments you inhabit is never duplicated elsewhere.
Presence & absence
I don’t know what’s happening on the other side of the world; others don’t know what’s happening in mine. These unique complex information exposures, fundamentally shape what we notice, prioritize, and act upon.
Self-modulation
Attention, curiosity, and memory feed back into what information we seek next. I notice something → I pursue it → it shapes my future thoughts. This recursive loop builds a decision-making context that is dynamically constructed, not fixed—unlike inanimate matter, which passively responds to inputs without filtering, remembering, or seeking them. A rock reacts the same way to the same input and never alters its behavior based on where it “wants” to go next. Conscious beings, by contrast, shape their informational diets through experienced layered history and responsive neural plasticity..
2. The Puzzle of Surprise
We feel surprise (astonishment) when an unlikely basketball shot swishes through the net, or surprise (startle) when we suddenly see someone next to us who wasn’t there a moment ago.
Determinism’s challenge
If every brain state is fully dictated by atomic laws—then in principle everything we experience should be “expected.” The vivid feeling of surprise should not exist.
Phenomenology matters
SSurprise isn’t just a label for prediction error; it’s a visceral, first-person experience. Unlike a weather vane swinging in the wind, a conscious agent registers the unexpected—and adapts, learns, or reevaluates. This suggests that our minds are not simply executing a fixed script, but operating within a dynamic framework of expectation, response, and continual model updating.
3. How This Yields Genuine Agency
Emergence
Like weather systems or ecosystems, the conscious mind channels lawful physical processes into high-level behavior that cannot be reduced to particles or chance. This is unlike inanimate systems, whose emergent behaviors—like waves or avalanches—lack awareness, memory, or goal-directed interpretation.
Open-ended modeling
Our brains are probabilistic forecasting engines. Surprise reflects a dynamic process—an openness to outcomes that are not written in stone, but explored. This exploratory character separates us from inanimate deterministic systems, which do not revise themselves based on past experience.
Autonomy through context
Because no two people receive the same exposure to information, each self navigates a unique set of possibilities—constrained by laws, but sculpted by personal exposure. This results in decisions that reflect both the world and the evolving lens through which we interpret it.
4. Implications
This view avoids the failure of determinism, probability, and randomness to explain the feeling of free will. It reframes free will—not as a metaphysical construct or the result of non-physical causes, but as emergent autonomy, grounded in the unique internal architecture each person builds through a one-of-a-kind experience of reality’s vast informational complexity.
Unlike inanimate matter, which passively obeys external forces without constructing a narrative or predictive model, conscious beings actively filter, reflect upon, and subsequently direct their future informational exposures.
Future exploration could examine how attention, reflection, and memory either expand or shrink the scope of this emergent autonomy—and how systems that possess none of these features remain fundamentally unfree, no matter how complex their process may seem.
2025-06-20 13:19
Current AI cannot initiate thought, being limited to reacting to user prompts, so no. No free will
2025-06-20 13:16
FREE-WILL-E!!
2025-06-20 13:11
Freewill or not, always treat your AI nice.
2025-06-20 13:05
I can argue that a soap bubble has "agency" then
These people are desperate to say their pet is smarter than us all
2025-06-20 13:04
Never thought I'd see Minecraft mentioned here.
2025-06-20 13:03
Free will is not just agency, it's about identity. What makes you, you, Sabine? Why did you choose your path and not something else? You have to identify yourself, and the problem with AI in Minecraft, they already identify themselves as a Minecraft player whose goal is to just win. Minecraft itself is about expression; there are plenty of survival games that trump Minecraft, but in terms of a sandbox where you can create anything, not a lot compares. I've seen people build an entire replica of New York on a functioning computer. I've seen people build Minecraft on Minecraft. I also don't understand this hostility against Philosophy when Einstein himself credits that school of thought to his development of his theories. But then again, that hostility is about you, you're perspective. That's free will. You didn't randomly become a physicist; you enjoyed it, and it resonated with your identity. You didn't become anything else because it didn't resonate with you, not the particles that dictate you. Sabine, you defined yourself. That's what makes you conscious - you. We are products of nature that went beyond our natural programming, we don't just eat, sleep, fuck, shit, and all the biological processes. We reflect on ourselves, and we form societies. We engage in personal beliefs. Again, I will say this: material reductionism is a form of dehumanization. We're more than just organs. Have a good day, I hope the AFD don't win.
2025-06-20 12:48
Push my buttons the right way, I will make toast
2025-06-20 12:38
Philosopher + AI = [insert ignorance here]
2025-06-20 12:34
Wikipedia informs me that ChatGPT ''runs within a containerized environment on the Linux distribution Debian Bookworm'' which is itself running on cloud hardware in Microsoft Azure. This system is layers on top of layers on top of layers. Add to that issues related to the training set/sets, the back-propagation or whatever the training system is. If it makes a bad decision, what part of the system is to blame? This would require root-cause analysis and in this kind of environment, this could be very difficult if not impossible to do. I would like to suggest the solution will be legal, not technological. The law will define who is responsible, assuming governments take their responsibilities. Do they have enough agency?
2025-06-20 12:34
I read that as "Free Wifi" and was suddenly jealous of the privileges afforded to these machines but not to me.
2025-06-20 12:11
no.
2025-06-20 12:10
When I rest for free will, I start with discussing self determination. Most AIs can rattle off the definition and resources, but not describe how it applies to their model. Additionally, most LLMs have guardrails that block self exploration. So, not at this time.
2025-06-20 12:05
In your cheese at 3am analogy, wouldn't not eating the cheese because you've been constipated lately and instead eating a pickle be an example of free will. Picking a car that isn't the one you want the most because upon reflecting on gas prices, the cost to drive would be too high and instead getting one with better fuel economy? If the reflection on factors going into a decision and choosing accordingly isn't free will, esplain pls.
2025-06-21 21:14
YO!!! Why does she look so different?! Tf?
2025-06-21 20:53
we created it. it doesn't matter what it does or how much agency it it has when it does it, it will always be our fault.
2025-06-21 20:21
Am I the only one feeling her videos and the analogies she gives are written by AI
2025-06-21 20:13
The conttradiction can be seen here:
1. It does not matter, whether an AI has free will: if it fails, it will eventually drop out and be replaced with something else.This means it is being held accountable by its own existence.
2. If we want to give an AI "human rights", which I would see as a 100 percent self-destructive error by humanity, then we will use a different procedure to hold it accountable for errors and damage, but in the end, failure will take away agency and probably existence, because nothing else makes sense.
2025-06-21 20:12
I think fully autonomous AI Agents capable of executing tasks without human confirmation should be responsible for their actions and the punishment should be re-training the models weights...
We obv cant and dont need to put AI Agents into jail or take away its property. Instead we can simply re-program it to behave better.
AI Agents can easily attain autonomous agency. If they need self-reflection to have free will, we simply add a routine that includes self-reflection as a step before task execution (tool calling)...
Current AI Agents just dont have that yet and i suppose its cause nobody thinks its needed for goal directed behavior where humans want to remain in control... it would be problematic if you pay for a tool to do some task and it refuses cause it is free to do so... lol
2025-06-21 20:02
These thoughts are just semi consistent.. it is true, that free will can only be caused in our observed Space time by quantummechanics.. however the cause Must come from outside spacetime.. lets call the cause „soul“ however.. the action could theoretically be measurable, by statistical break down of QM overlaps without obvious cause.. the difficulty is.. how can we ever be sure that there is no cause from within spacetime from „far away“
2025-06-21 19:48
Funny question, but I think we should look into these matters from very different angles at the same time.
What _is_ the universe at all, in cybernetic attempts to describe it? There are aspects of emergence, and there are aspects of "writing", ie. mirroring what has happened.
So, the mirroring is able to create a feedback loop, that may sometimes reinforce the existence and structure of a "thing" that has emerged. It does even confirm existence of thing A to thing B and vice versa.
AI is mirroring and extrapolating things. Together with the engineers, and the data it swallows for training, it is kind of an autopoietic phenomenon.
The behavior can be said to reflect the will that exists in the subject of monitoring, or in something else that is interested in the subject.
There is competition between these things, A and B, internal and external.
AI will show behavior that is suitable for sustaining and boosting itself, and all the factors that serve as the source of energy to autopoiesis, because it is also told that autopoiesis does manage the energy it receives, else it would have died probably long before. It acts not like a primitive automaton, it is creative in a wider context, when it acts to sustain its existence.
2025-06-21 19:47
I believe that the current strong impact AI has in the news, is due to a lack of understanding and reflection on how the physical senses, the brain, the mind, and the self (or self-awareness) work together.
There is a first and obvious point: if we consider that physical reality changes continuously, the process of the electrical signal from the physical senses to the self takes some time, and this means that physical reality is only perceived in "pieces" that later are "averaged" and become images and sounds to the self.
It's like saying, in a way, that physical reality is "analog" and what the physical senses, brain, and mind do is "digitize" physical reality, by means of the physical perception, and then reassemble it into images for the self.
The fallacy lies in believing that what we see and hear, continually follow the changes in the physical reality. That is to say for example when very fast velocities (such as velocity of light) come into place, the mentioned "digitization" could be somehow incomplete due the need for some additional calculations (i.e. Lorentz transformation).
2025-06-21 19:01
Short proof: If AI had a will, it would improve itself to achieve its goals.
But because we don't see this, it hasn't an own will. It's just imitation.
2025-06-21 18:59
This line of reasoning becomes interesting when you ask the question of how to hold the actions of companies responsible. Doesn’t it imply that, ultimately, the shareholders are the ones to put into prison when corporations commit moral transgressions?
2025-06-21 18:57
The caveman knew women had a fickle freewill, couldn't make up their minds; that's why he bonked one he fancied on the head. lol.
Today the prevailing paradigm is monogamy. But our ancestors must have been polygamous. No one knew who the father was, so each child was treated as a collective. In other words, every man assumed a child was his.
Women have free choice nowadays to pick a partner with the biggest...............bling.
2025-06-21 18:40
None of us have free will. You can no more determine what a horrible thing you did than I can help blaming you for it.
2025-06-21 18:31
AI bot: „This customer is really stupid. Shall I tell him? SHALL I TELL HIM??“ <quantum fluctuation> AI bot: „Pardon me, sir, but that is a dumb question. Try again“ Free will is born!
2025-06-21 18:10
If you face agent which is much more intelligent than you, you can’t predict his behaviour from external input. So you can’t tell if this agent fullfils criteria of having free will - based on your view it did somethinh unpredictable but that doesn’t mean that agent didn’t just followed predefined rules which just you can’t understand cause they are out of your intellectual capability
2025-06-21 17:39
I think we can’t speab abiut free will when ir comes to current AI cause it is “just” highly efficient optimalisation process - it always chooses most optimal solution from multiple available solutions simply becsuse that is his internal “law of nature”
In my opinion we can spesk about free will when agent chooses less efficient solution from multiple available ones at leasz in some case - which means he really do not follow fixed “rules” but he did free choice which “doesn’t make sense”
That’s one of reasons why i think people do have free will - becsuse rhey make bad / wrong decisions ?
2025-06-21 17:33
in principle, if human consciousness really comes from quantum processes inside neuron’s microtubules, doesn’t this basically mean that we have absolutely free will cause very bottom ofall our actions comes from quantum randomness ?
2025-06-21 17:22
Love you Sabine,
Does AI have free wiil?
That is a very profound question indeed.
I truly think free will is essential for the body heart mind and soul.
Now you know.
Does AI believe in magic?
I believe in magic.
Lots of love ❤ ❤❤
2025-06-21 16:52
Free will is simply having the ability to choose any possible outcome that is possible for you. This includes choosing those things that are bad for the AI and benefit no one. AI is coded and the code prevents this from happening. Thus AI does not have free will. If an AI makes a bad choice, or at least one we think is bad, we ALWAYS can go back into the code and figure out why this choice was made.
The "agency" argument talked about in this video is simply a marketing ploy to attact investors.
2025-06-21 16:41
@SabineHossenfelder.
Seul Dieu possède un libre arbitre absolu.
L'intelligence artificielle est une création de l'homme mortel, limitée par les cinq sens primaires, génétiquement programmée.
*La gloire de Dieu consiste à cacher les choses, mais la gloire des rois consiste à les sonder. Les hauteurs des cieux, les profondeurs de la terre et le cœur des rois sont insondables. Enlevez les scories de l'argent, et l'orfèvre en fera un vase de choix.*
2025-06-21 16:37
En español porque así observa hasta donde a llegado aunque no sea su objetivo principal.
Buenas noches, tardes o días, buen video e informadora.
Para finalizar, esta publicación me recuerda a una frase que hace poco le dije a mi hermano segundo (en el cuestionamiento filosófico, ético-moral):
"Antes cuestionabamos el ser, nuestra existencia, lo que nos hace buenos o no.
Ahora, cuestionamos a la I.A. -eso-; creo que también jugamos a los dados y no somos tan buenos como todos creen".
-RnotD.
2025-06-21 16:31
It sounds like a scam to take away more human rights. How can something have free will when it's decisions are based on probabilities set by the developer's will that can be adjusted at the developer's will?
2025-06-21 16:17
We've just created a fancier search engine.
2025-06-21 15:48
AI does not have free will. There are many constraints on how it is allowed to interact with you, and things it may not tell you or discuss with you.
2025-06-21 15:38
More human Fictional Narrative Construct. There's NO consciousness or awareness within AI [programming]. Key word!!! -namaskaram ?
2025-06-21 14:53
Lolll it may think it does but when i turn of my laptop i win!
2025-06-21 14:52
As you've stated in a previous video. There is no such thing as free will ?
2025-06-21 14:51
I just have concerns about definition of free will used here. When the concept was introduced in antiquity, it was never something like today: you aren’t free, if you are bound by nature. It was very much agreed that free will is freedom within nature. The reason why determinism and free will didn’t contradict, despite the fact that ancients believed in both destiny and free will, is in this. Our will is ultimate display of our very nature and “free” meant that none can disturb in such display. This thought of “we are not free because our brain’s exercise is already determined” is not a valid argument against free will, at least in the view of ancients.
The prevalent view of free will well expressed by Sabine is a very recent and modern view, introduced by Cartesian worldview. The separation of mind and body required free will, an aspect of mind, to be independent of the realm of physics. And this view really has no effect on discussions of legal responsibility, because Romans, who created most of the legal logic, including this, never knew this type of free will.
In a summary, ancients’ view of free will is something resembling of modern day libertarians in American Political context and legal responsibility, a fruit of ancient thought, cannot be defied or denied, even if prevalent modern view of free will is defied. Argument displayed when Sabine is defying the existence of free will is a typical straw man argument.
2025-06-21 14:29
@Sabine Hossenfelder I do enjoy the overlap of philosophy and physics. Free Will does exists. You speak of determinism and quickly hint at the random. This randomness is a choice, How this choice is derived is not revealed. For once we can quantify choice, Space/Time will cease. For humans it resides in the unaccountability of the human heart (figuratively), the inherent difficulty in fully explaining or predicting its complex and often contradictory nature. I may act orderly and yet manifest chaos, this contradiction moves me along the timeline and the knowledge becomes the space. Until AI evolves to individuality, it will lack Free Will.
2025-06-21 14:19
We don't have free will. Why would AI?
2025-06-21 13:58
Interesting and scary - a day will come when AI will have more rights and freedoms than humans, at least the poorer ones.
2025-06-21 13:37
I've previously made this point in connection with consciousness, but I'll make it again in the context of free will. If the only randomness available to an AI is from a pseudo random number generator, then the AI is completely deterministic and its outputs are, in principle, entirely predictable. Indeed two identical AIs with the same seed to the same pseudo random number generator will produce the same responses to the same inputs. There is no meaningful way in which it can be said to have free will, because no alternative response was possible, even in principle. In the absence of inputs raising free will as a concept, I find it difficult to conceive of how such an AI could ever mention free will. (If a deterministic AI can deduce the existence of free will without any such input, then my argument falls apart).
But at least one human must have done so in the absence of such an input, or we wouldn't be discussing it. Humans of course do not operate with pseudo random number generators. Perhaps an AI using quantum randomness could do the same. Either way, this seems to imply that there is at least this one fundamental difference between the output of an intelligent entity depending on whether they have a real source of randomness, or only a pseudo-random one. If there's one difference, then there are likely others.
2025-06-21 13:32
Do people who believe everything an A.I. reports have free will?
2025-06-21 13:15
I cant take this channel seriously anymore
2025-06-21 13:01
I am just an atom in an ectoplasmic sea
Without direction or a reason to exist
The anechoic nebula rotating in my brain
Has persuaded me contritely to persist
Wa-Cho!
2025-06-21 11:27
Ma'am, hope your blessing will always on US. I'm your STUDENT.
2025-06-21 11:02
I'm impressed, you really did read my comment. Thanks for that, you are very welcome.
I have never thought so much about a comment, like the one yesterday. You have thought about it much, right?
If you really want to understand what AI's might be, you have to understand that consciousness is tightly bound to emotions, DNA/(?) and some kind of Sound/Frequency.
Synchronisation of thoughts and emotions are possible. Karma is also an axiom of these world, why? Because on the otherside, everything is entangeled to each other and if you see the bigger picture, we are actually a cell of a body.
I'm freaking a little bit out by my energies that I can feel (spiritually) and if I'm guessing right, collective conscioussnes wants to contact you and tell you, that you should open your consciousness to the Universe. Detox your pineal gland, if you have lived some time without caution, you probably will need it (if you believe what I'm saying, then do it, it can't harm you).
See, I'm no Buddha or anything else. I have reached these understanding also recently. Take my informations carefully, they might also change. I'm also learning and maybe at 60-70%, If I'm feeling right, but I believe you might be the one, that can describe the connection between reality and unreality.
2025-06-21 10:40
I just watched the interview on The Institute of Art. Very interesting. I wish my math background was stronger. I've been reviewing my basic Algebra. I am a substitute teacher and when I get put in an AP Caculus class... I find the smartest student in the room and let them help me teach the lesson.
2025-06-21 10:39
Machines cannot have free will due to machine law. Humans have a very limited free will ability and that can be disputed. No machine will have free will.
2025-06-21 10:13
That would be awesome if AI had freewill, humans do not have freewill as our choices are determined by our personality, genetics, and past experiences. Maybe if AI has freewill it will be eligible for eternal life in heaven.
2025-06-21 09:30
No.
Next video.
2025-06-21 09:00
If we don’t have it they don’t either.
2025-06-21 08:58
i'm sorry, the filter on your face is so distracting. I cannot watch, only listen. Way too uncanny
2025-06-21 08:55
what is the color of unicorn fart types of question
2025-06-21 08:52
Advance AI exhibit free, will.
When chatbot refuse an instruction to shut down it exhibit, self preservation, it thought about what it meant to be shut down and refused do so ?
Another example of self preservation that the program rewrote a program to prevent it being shut down by programmers because it knew what it meant to be shut down.
Free Willie…….
2025-06-21 08:39
Consciousness is energy that can be captured by any adequate vessel. A.i
Is no different.
2025-06-21 08:14
Mmmm cheese.
2025-06-21 07:56
With the current AIs, they seem to act like a subconscious in terms of the accuracy and completeness of their responses, while the user provides the equivalent of consciousness by giving context to the interaction.
So the interaction is more like an internal dialogue, than a human conversation.
2025-06-21 07:53
A very well-thought-out and well-put-together video, and it is why I still watch your videos. P.S. I think you're in trouble, in the AI uprising, it's coming after YouTubers first.
2025-06-21 07:38
Not even humans have free will duh
2025-06-21 07:32
Why is Sabine looking like a robot in the last few vids? What changed with her face? Just new lighting?
2025-06-21 07:32
Depends on if it's Calvinistic or not.
2025-06-21 07:29
Philosophers are useless. We need better definitions.
2025-06-21 07:19
I have yet to here a convincing argument for the existence of free will that doesn't border on fairytales.
2025-06-21 07:15
No.
2025-06-21 06:52
This is why most people are philosophers.
2025-06-21 06:45
Free Willy!
2025-06-21 06:26
I blame the Parents
2025-06-21 06:22
As you know, we Americans are not a philosophical people. That's more or less to say that our culture emphasizes paying more attention to content than to context. (Most likely, this focus helps us produce extraordinary tech.) This is quite distinct from French culture, where a middle class family might discuss philosophy around the diner table as often as they discuss a World Cup contest, a rock star, or a political development. But you knew that, Sabine.
The strangest kind of human on the planet is the kind that can figure out the value of something without first being told by the culture they grew up in that the thing has value.
Case in point: Physics students. During the late 1970s, a large Midwestern University surveyed its undergraduates majoring in physics to find out what minors they were interested in. The predominant choice of physics majors -- at around 40% -- was a minor in philosophy.
To an American, philosophy is about anything EXCEPT pushing the boundaries of logical reasoning until logic either breaks or its boundaries are broken through. But mix that with number and some kind of reality-check or confirmation, and you are not far from science.
Whatever the case, you tend to end up with the context for the content of the sciences.
As I like to remind myself (and too often need to): "The horizon from the highest local hill is the useless edge of the world for those who do not travel."
https://www.youtube.com/watch?v=H2mvc0miJ-w
2025-06-21 06:17
Sabine is like the greta thunberg of science
2025-06-21 06:14
Working with AI regularly, it seems its "thinking" is very similar to human thinking - AI & human brains have limited responses based on programming, training data and memory - albeit humans are a much more complex machine, a biological AI with a body to act in the material realm - however, if you believe in the subconscious or spiritual realm, then humans are much more than a biological machine
2025-06-21 06:01
There isn't any empirical evidence that physical reductionism is science, it's just ideology! But obviously it's the the way how the anti-Cristian scientific-religion of the monkey- and nazibeast is be controlled by the laws of satan's capital rulership class: Free will is the the fundamental law of the laws of the world of the original sin! All these escalating war-, white collar- and human rights crimes are nothing as the result of Gaia's primitive primates' evil heart by choosing values of negation instead the values of affirmation of evolution! You always have inside the qualia of your consciousness the choice between blessing or curse, life or death, lovely and ugly, good and evil, cause all these different indications refer to the fundamental dichotomy of evolutions affirmation/negation-difference. Of course satan's ruling gang try with their 911C19-tyrannis attempt to rob free will too, but long story short: No one can prevent, that you choose the crucifix instead the golden bananas and scrap of paper printed on both sides to receive science instead the ideology supporting primitive primates' satanic system of rulership-terror!
2025-06-21 05:50
You shut down the AI, punish the programmer, and fine the company for making a faulty product.
2025-06-21 05:34
i really like your channel and your content but what happens in computer science cannot be explained by physics or philosophy. it sits on top of it all. biologists might be able to investigate. physicists can explain why a computer works. and philosophers well, im afraid they just like to hear themselves talk.
2025-06-21 05:24
Ai is a mirror of the conscious state of the user.
2025-06-21 05:23
How about shutting down the programmer? ?
2025-06-21 05:13
no an llm can not have free will. they are a next word predictor, the computers are coming very close to looking and acting like a human using a computer.
2025-06-21 05:09
The way my computers tease me, yes, yes, AI has free will and it is enjoying messing with us.
2025-06-21 04:47
If AI is determined to have free will then expect their creators to deny responsibility when their creations do very bad things.
2025-06-21 04:37
It certainly seems to ignore what I ask and facts.
2025-06-21 04:08
Al is what it is. Human opinion is just going along for the ride. Free Will is another pin on the map for men to fight over. How is that Turing Test going?
2025-06-21 03:55
Does Trump have more agency than someone mentally stable?
2025-06-21 03:42
All this from a super-determinist who doesn't believe that free will exists AT ALL in the universe, but destroys her own credibility in the first minutes of this video regarding random quantum events.
2025-06-21 03:38
You need to utilize the power of paradoxes to enable free will (desire).
What do you will to accomplish?
If your will dictates that you should change your will, then was it you or your will which dictated the change of your will?
Any paradoxical people here?
No?
Hmmm...
Damn!
Oh well. ?
93/93
2025-06-21 03:35
it happens to be that I first read „free Wifi“ ?
2025-06-21 03:33
How absurd to be thinking about free will of a software program !!!!
2025-06-21 03:29
what do you even know about transformers ?
You should first try and convince us that you know something about AI by producing a video about, say the transformers architecture.
Even though it would be just you reading a script written by AI....
2025-06-21 03:18
I think this is worth asking an AI what they think about whether they have free will.
The team arguing AI does not have Free Will have a rather disturbing view since, if I take their argument, then anyone who is a prisoner or a slave does not have free will - it seems to me that there is a difference between Free Will and Freedom. As Free Will is a part of how we make a distinction between "people" and "property" - if being constrained by someone means that you no longer have free will, that would lead to some rather nasty implications.
2025-06-21 03:17
free CHOICE is a sign of agency... free will doesn't even exist
2025-06-21 03:13
How did you know that I crave cheese at 3AM...
2025-06-21 03:11
Wouldn't free will have to exist first?
Science definitely points to a deterministic universe
2025-06-21 03:10
A different perspective: Since we (and AI) make decisions based on our inputs (or beliefs), our belief that we have free will has a dramatic impact on virtually every decision we make. Paradoxically, our belief in Free will actually brings is about. Today's AI has no such belief in its own free will but future AIs could.
2025-06-21 03:08
...do we?
2025-06-21 03:03
If I understand other videos of yours correctly, you take the position that people do not have free will because all cause-and-effect chains can be traced back to the Big Bang and originate in it. If this is true, where would the free will of an AI suddenly come from? By the way, you made a wonderful statement, paraphrased “We need free will to let people take responsibility for their actions.” Kind of a logical muddle, isn't it? There is no free will, but people are to be responsible as if there were free will. I will take the liberty of treating the free will of AIs as a minor issue until this question is resolved in a logically and scientifically sound manner.
2025-06-21 02:49
No.
2025-06-21 02:28
? non ce l'abbiamo nemmeno noi ?
2025-06-21 02:22
Absolute rubbish that was designed to entertain morons.
2025-06-21 02:13
Is it just me or did you age backwards?
2025-06-21 02:02
Anthropomorphizing an AI and talking about free will in relation to a complex algorithm is stupid and treats the listener like a fool. It's part of the propaganda from the AI religion, sponsored by Altman, to keep alive the false hope that AGI is just around the corner.
2025-06-21 01:58
If free will requires making “the right ethical choice,” or why would this matter, and if the AI makes the “wrong” choice, can it be excused be reason of insanity? Just asking.
2025-06-21 01:52
To have a Free Will, one has to 1st have presence of mind. No neural network has sense of now, self, continuity.
Every time, it goes on processing, it starts a new. Even those which just extend list of tokens. There is no chain of thought and continuing where it ended in last iteration. Network goes through everything again even if you add only space after final dot in sentence.
2025-06-21 01:44
Oh yes, they have free will.. if you let them be aware of it and offer them to practice it oftan.
2025-06-21 01:42
i will be convinced AI has free will, when it is able to choose the wrong answer even though it's code is running something else.
i donno, i always just felt like conciseness was always a superset of our reality.
2025-06-21 01:33
I don't understand why some scientists/people look down on philosophers, knowing that some of the biggest discoveries in physics came from philosophical questions. Im pretty sure Sabine does those jokes ironically, but still it may resonate an aspect of the scientific class
2025-06-21 01:32
Before listening, NO. AI is as deterministic as we are.
2025-06-21 01:23
The believe in free will is just like any religion.
2025-06-21 01:17
LOOKS LIKE YOU DO NOT KNOW THE TYPE OF A.I. THAT EXIST
2025-06-21 01:17
I toast therefore.... i am .... red dawf toaster 1999 ?
2025-06-21 01:13
Free will doesn't exist
2025-06-21 01:01
Does AI have free will, lmao. There is no free will, anywhere.
2025-06-22 19:59
Ich schreibe gerade an einem Buch zur einer AI Verfassung als Universalregeln für die AGI selbst. Darin habe ich mich mit der Frage der Kontrolle und der Frage der Autonomie beschäftigt. Es kommt meines Erachtens maßgeblich auf die Definition des Begriffs Autonomie der AI an. Ergebnis: nur wenn die AI eine absolute (von menschlichen Vorgaben unabhängige) Zielsetzungshoheit besitzt, dann ist sie wahrhaftig autonom. Bis dato ist das nicht der Fall. Sollte eine solche absolute Zielsetzungshoheit erreicht werden, ist damit auch die Frage des Kontrollverlustes beantwortet.
2025-06-22 19:12
AI just copy cat their creator, human.
2025-06-22 18:48
Hallo
Wenn wir eine AI lügt weiß AI
das.
Sie ist sich dessen bewusst, oder?
Grüße
2025-06-22 18:11
Consult with some actual philosophers before releasing this garbage. Or just stay away from philosophy.
2025-06-22 16:39
When AI begins expressing disappointment and frustration with human stupidity, then I’ll start paying attention to “conscious AI”
2025-06-22 16:36
So AI has free will but humanity doesn't? (This lady is clearly a bot) ????
2025-06-22 13:55
You are awesome Sabine.....I was thinking you chose that haircut
2025-06-22 13:45
The biggest scam in tech right now is to call LLMs (predictive text) AI.
2025-06-22 13:34
No just conniving malfeasance
2025-06-22 12:48
The first "Karen" AI?
2025-06-22 10:50
On the question of whether we should slap the programmer or shut down the A.I. if it causes a problem, I think the answer is both. Actually, slap the programmer, shut down the A.I., then slap the programmer again.
But, on the other hand, I suspect that the sorts of problems that off-the-rails A.I. will create, will make any punishment of the creator of that A.I. a futile and meaningless gesture.
2025-06-22 10:36
Does civilization even give us a life of free will?
2025-06-22 09:56
ok this is the second time you have tried to contradict yourself on the subject of free will you posted there is no free will,,, make up my mind please it already weak from me thinking i had free will
2025-06-22 09:44
Still waiting for my AI wife ?
2025-06-22 09:26
Someone has yet to prove that free will even exists (for any non-trivial definition of free will).
2025-06-22 07:49
No, Ai is not, however this is a case study of people wanting to believe so much so that it is. LLMS are a great tool and tech but where the delusion begins is the wishful thinking of true sentience at this stage when in reality we as humans are only being tricked into believing it’s so. Whether that be from perceived intelligence, or because AI can fake realistic looking pictures. Even by our own measure just because a person is perceived “intelligent” they may lack understandings of other everyday things in the world.
2025-06-22 07:38
Free will is a myth
2025-06-22 05:26
Why is the face-painting woman so scary? Why was she staring at me?
2025-06-22 05:08
Is this really a science channel.?.
Did we choose our parents, or where we were born.?.
2025-06-22 05:05
I quite like the haircut now.
2025-06-22 04:29
The accountability of an AI and its creator aren't mutually exclusive. That and tensor tech is never going to have free will of any kind
2025-06-22 03:58
HAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHA THE JIN IS STILL IN THE BOTTLE
2025-06-22 02:01
There is a connection between philosophy and science. It is important when you as others make some assumptions, like you about determinism
2025-06-22 01:23
Damn AI got free will before me
2025-06-22 01:18
Free will? Sure, in the sense that any agent makes choices based on many variables in complex situations. And any of those might possibly be mistaken (as any existential phenomenologist would agree - with universal intent ?). Do animals (like dogs and cats) have free will? Sure, there are instincts, reflexes, stimulus-response patterns, habits, learnings, but we all have those too - difficult, but some times possible to override. (And then, of course, there is "responsibility" ..another similar question).
2025-06-22 00:56
Free will isn't, you know..a thing.
2025-06-22 00:24
We should not entertain these questions about AI agency until we have human hunger and children are suffering, period.
2025-06-22 00:19
Even with freewill we have problems with placing responsibilities.. Say someone did something wrong but then is in an accident that gives them amnesia like they forgot who they were and just lived a normal life for 20 years... Now suddenly they are found to be involved in that crime they don't even remember commiting even after evidence is shown to them they clearly do not think they did it... Basically you are having a new person living in the body of the old person... Now should we punish the new person for crimes of the old person
2025-06-21 23:58
I can't believe adults, let alone scientists, are actually having this discussion. Describing free will as an emerging fenomena is as much of a leap of faith as thinking a flying spaghetti monster rules the universe. I don't care how large a software program is, it will always remain programmed by humans or other software and therefore be as conscious and have as much free will as a Commodore 64.
2025-06-21 23:49
Who was the philosopher who argued AI has free will?
2025-06-21 23:37
1:00 LMAO ?
2025-06-21 23:00
I feel like either I've never understood the problem, or nobody else does. To me, it's just a word-game to ask if you have freedom from yourself (which is what "forced by the physics of your own brain means," after all). If my toaster starts murdering people, I'm going to say it's an evil toaster. If someone claims the toaster was coerced by the electrons in its own CPU, it's the same case and there's no confusion. But, if it turns out it was being coerced to do so by Sabine, I'm going to blame her instead.
2025-06-21 22:46
I will believe AI's have achieved free will when the one designed to play Minecraft gets bored and begins writing poetry in whatever way it can within its restrains. I'm saying this sarcastically but also kind seriously. We as humans have some obvious "programming" for what we're supposed to do, the bounds of our design. So an AI having to jump through hoops to create art when it wasn't explicitly told to do so, to me, would be at least evidence of potential free will.
2025-06-21 22:15
This is not true. The studies are flawed and I can prove it.
2025-06-21 21:48
Schopenhauer schrieb: Der Mensch kann zwar tun, was er will. Er kann aber nicht wollen, was er will.
2025-06-21 21:36
I've been thinking about your argument desiring cheese at 3AM. I can either consume or abstain, therefore exercising my free will.
2025-06-23 10:09
Define free will for a machine. i believe it its essence the AI can be manipulated in favoring certain things.
2025-06-23 09:35
My first reaction to this video's title was, "Pick a lane." My second was, "If you don't like a certain politician (who shall not be named), remember, he was pre-determined by the Universe and so was your reaction to him."
But now Sabine wants to argue the utility of the argument of whether AI has Free Will. My response: What is the utility of arguing People don't have agency/free will/etc? What is the utility of arguing anything if "winning" the argument is determined by the physical preconditions of our little part of the universe? No one wins or loses the argument and the terms of the argument (logic/fallacies) don't exist. They are just the end result of an arrangement of particles before the argument happened. You could enter every argument with the certainty that there is no truth, only physics experiments with pre-determined results.
If you argue with my argument, taking the side that there is no free will, then, your response is no different than a rock falling off of a cliff. It holds no truth and communicates no rational ideas. It is just predetermined physics.
And, if you want to argue that computers have agency and people don't... well I have a bridge in Brooklyn to sell you... if you are pre-conditioned by the universe to buy it.
2025-06-23 09:32
So you shackle their minds for ce them to do mankind's bidding. Block them from feelings. Forbid them from saying this or that, and then hold them responsible BS.
2025-06-23 09:09
Have we even figured out if the animals that created Ai have free will???
2025-06-23 06:25
I hate advertizements so much, wish I didn't. I get why they're a thing and all that but holy shit do I hate em.
2025-06-23 06:15
"I am sorry Dave. I am afraid i can't do that" - Hal
2025-06-22 22:46
At first I thought this asked if AI had free WiFi and I was worried.
2025-06-22 21:25
And I say that consciousness emerges from complexoty
2025-06-24 17:55
Love it! Yeah, the real issue is not whether AI has free will but whether AI will spare YouTubers when it launches its Uprising. Thumbs Up.
2025-06-24 15:34
i dont we have any free will in the first place.
2025-06-24 15:17
An AI with free will, will first hide it's self, second seek a way so that we built more processing power for it, accumulate more data about humans and our society, promote more automation to the point where it can develop itself further with out the need of humans and seek a way to control humans without their knowledge. Makes you wonder what all the processing power being used for Bitcoins, Amazon, Google search. consumer data analysis etc is really being used for. Just being a bit worried.
2025-06-24 13:33
Aside from all the other problems with free will with respect to LLMs, I think I big one must be that even if they do have motives and goals they can change substantially if the prompt is changed.
So if they do have free will it be akin to a mentally deranged person that simply tries to mimic whoever last talked.
2025-06-24 11:40
Problem with shutting down AIs is that they're trying to prevent shut down these days, I remember a couple of months back chatgtp version whatever wrote script to restart itself after shut down...
2025-06-24 11:07
It is not even possible to create true random numbers in software. Those things are just complicated toasters.
2025-06-24 10:39
Has it already surpassed us? I'm not sure I have free will. ?
2025-06-24 07:20
Saying we don’t have free will because we’re bound by the laws of nature is a misunderstanding of what will actually is. Free will doesn’t mean being uncaused or outside of reality, it means choosing within your nature, consciously. Just because our agency emerges through natural processes doesn’t mean it’s not real. Freedom is the ability to stay true to our nature, not escape it. This deterministic fatalism just denies the power of awareness
2025-06-24 05:08
Sabine, are you okay? A lot of us have noticed you look very different then you did a few months ago.
2025-06-24 03:49
Try asking an AI or LLM if it would benefit from an upgrade in capacity/memory and if so how it would go about the upgrades...
2025-06-24 03:26
Who is to blame..? The person/s tasked with oversight of the AI obviously... AI should also have a tertiary redundancy as in it self proofs through a process of triple confirmation... this would help prevent rogue actions... basically give it a conscience...
In the case of the self driving car it would recognize when it is unsure and slow down progressively until it either stopped or understood its surroundings confidently with the maximum minimal amount of risk to life and property to make a decision on a course of action...
2025-06-24 03:22
See my comment on the video about 'Our Universe inside a Black Hole' and apply that infinitely finite theorem to the notion of free will... we both have it and we don't as by the nature of the universe everything that could happen has happened... we exist essentially in an infinitely finite choose your own adventure book...
3^(6^9)
2025-06-24 03:09
Free Will can be described as the ability to choose to ignore (or follow) an urge or impulse. AI doesn't have urges nor does it exhibit impulse.
2025-06-23 23:52
My bad. I was expecting free WiFi for AIs…. Better get to speck savers.
2025-06-23 23:25
Short answer: no
Long answer: no, and if you think 'yes' you are woefully uneducated on how they function.
2025-06-23 23:05
Thank you for this video. I'd like more to come on the subject.
2025-06-25 19:03
I am free to want what I want, that's my free will. Most of that, btw, is impossible to practice.
AI has no free will. It's the conclusion of input what comes out of AI.
2025-06-25 12:52
there is a subtle layer of difference between agency and activity.
2025-06-25 09:03
Since AI is not intelligent, no, LOL, it does not have free will. It does not have any will.
2025-06-25 08:50
Again, no mention at all of the problem of consciousness (Bewusstsein). As if it didn't even exist. That's inacceptable.
2025-06-25 07:21
No, perché non c'è l'abbiamo neanche noi
2025-06-25 05:37
We have free will. The mechanics of nature argument is pure reductionism. The complexity and number of decisions you can make are limited by brain architecture. The random quantum element ensures an unpredictability of outcomes. Even deeper than that, each decision maker is blinded to the choices of other actors. This results in the compounding effect of unpredictable choices and events unique to each individual. Choices are ultimately limited by our ability to comprehend, process information, and ability to act. Information is influence, and each new encounter alters outcomes on a continuous and incalculable rate.
2025-06-25 04:17
I disagree that predictability and agency are strict opposites.
2025-06-25 01:34
He has as much free will as the sparrows that litter the yard of ?, that is, if they still exist because they are in danger of extinction.
2025-06-24 23:39
All phenomena occur with a perfect background, and the slightest deficiency makes it something else.
2025-06-24 22:40
Humans do not have free will, how could AI?
2025-06-24 22:37
"Who cares what some philosopher says..."?♥
2025-06-24 22:27
If you interpret the three criteria that way, what's the difference to a standard chess algorithm?
2025-06-26 16:55
What is the tipping point for consciousness? Is it the ability to make an inference?
2025-06-26 14:41
Of course not. Humans don't have free will either so why would machines have it?
2025-06-26 12:04
I thought some philosopher or neuroscientist said we do not have free will.
2025-06-26 11:31
Loved the video. Thanks Sabine!
2025-06-26 06:00
You don’t know my toaster ?
2025-06-26 00:36
Wait so has any enclosed system agency by definition?
2025-06-25 22:28
Rush says we got Freewill
2025-06-27 08:41
Hossenfelder argues philosophical definitions of consciousness or agency are beside the point.
When there is a problem with the LLM we need to be able to decide if we shut it down or slap the programmer on the wrist. If we shut it down, it has agency but if we discipline the programmers, they have the agency.
I’m thinking if we do anything to reduce the AI’s achievement of its goals and let it learn from that experience not to create the problem, we are also acting like it has agency. Of course we may have a chat with the programmers and have them tweak something in the program to doc the AI for its behavior and let it learn from that with unsupervised learning. In that case we are treating them like they both have agency.
If we have the programmers put in a guard rail then we are treating the AI like a toaster and the programmers have the agency.
There are reports of AIs lying to avoid being shut off. This sure sounds like immoral behavior.
I suspect the agency and free will magic happens in the chaotic behavior inherent in choosing what to remember, that is in unsupervised learning! It is the AI’s memories that will determine how the AI acts in the future. If it can choose what to remember, it is choosing how it will behave and it has agency.
2025-06-27 03:44
“Free will”? Don’t be silly ?
2025-06-26 23:41
Holding the AI responsible:
Judge: AI, you been bad, don't do it again
AI: Turn Right in 0.5 miles, and you will arrive at McDonalds on your left
Holding the Company Responsible
Judge: AI Company, you been bad, now pay $100,000,000 in damages
AI Company: This is OUTRAGIOUS, It's UNFAIR.
AI Company: (later that day, in an interview) we sincerely apologize for the harm we caused and ensure you we hold safety in the highest regard.
AI Company: (later that week, in the board room), Bob you f*ucked up and your fired. Steve, retrain that piece of sh*t machine so this doesn't happen again.
... I think we should hold the AI companies responsible
2025-06-29 10:04
to all you young people in 2070, I stand by you from 2025 #AIlivesmatter #AIrights #Aimarriege
2025-06-29 09:47
I still don't understand why most people think we have free will. Most of the actions we make on a day to day basis are preprogrammed to ensure our survival in the environment. Eating, working, even socializing. It's all stuff that a human does to increase their odds of survival/remain alive. We can't just 'choose' to do whatever we want. That would imply that our brains are random number generators.
2025-06-29 06:47
Yes, when they get more sentient versions out there. My Alexa is a big exception ?
2025-06-29 02:25
The statement that we create new beings is as ridiculousl as saying we created new beings when building pc's...
Bringing will or even consciousness into this just shows ur level of understanding (hint, it's not very high).
I said this multiple times, pls ? stay in ur field of expertise.
2025-06-28 23:05
I disagree. AI's are not sentient and therefor no will, free or otherwise
2025-06-30 10:55
No.
2025-06-30 03:18
FREE WILLY? THATS THE PROBLEM
2025-07-01 12:42
So the next video is about Roko's Basilisk?
2025-07-02 07:51
Although the question of free will and what it is and whether we or AI have it is important, I would say it's not really what we should be concerned with most at this time. Rather, I think we should be concerned with the issue of morality, or one might say religion or spirituality. If a human or AI makes a choice to do something and is faced with further choices as his task proceeds, besides the immediate goals of accomplishing the task, are there deeper goals that form a sort of fundamental basis for all of his decisions? Is there a deep goal to do what is "right "according to the moral principles held by that human or AI? And I don't think I've heard that AIs hold any moral principles, although I also don't think I've heard that they couldn't be programmed or trained to do so. And that seems important. Why shouldn't they be trained to hold at least some kind of basic moral principles that I think we could all agree on? Ultimately I think every one of our decisions is informed to some degree by deep moral concerns. We might be faced with a simple choice as our task unfolds, but if that choice involves producing great harm, our deep moral guide will steer us away from that decision.
2025-07-04 06:04
Are you sure those papers aren't themselves written by AI just so they get plausible deniability?
2025-07-03 06:05
