Awakening the Machine: Max Tegmark

[@wcibbt]

Always a pleasure to listen to Max Tegmark speak. I recommend changing the transition audio from digital static to something less abrasive.

[@RishabhSharma-og8je]

bot?

[@Jeremy-Ai]

Max You are an essential leader. Timing of we is crucial to survival

[@apple-junkie6183]

Max, i have read you book Life 3.0. It is my inspiration and I am writing a new book about a New „Last“ AI Constitution in analogy to Asimov‘s Laws of Robotics. One of your topics of this interview I have already solved and will come to the conclusion that the loss of control to an ASI is inevitalbe.

[@anatolwegner9096]

Wonder how this will have aged in 10 years?

[@Youtube_Stole_My_Handle_Too]

I appreciate Tegmark's clear and direct language. He does not mince words and is unafraid to speak his mind. However, I believe he overstates his case here. He claims that AI threatens humanity at its current stage of development. This assertion likely rests on the assumption that AI is complex. A cliché within the field has granted AI this undeserved label, which may have misled Tegmark. It is simply not true that AI is complex in a scientific sense. AI lacks emergent properties. All its parameters are adjusted using highly predictable algorithms, producing smooth transitions. The high dimensionality and non-linearity of AI systems, while notable, are not sufficient - within the framework of the philosophy of science - to justify categorizing them as complex. I absolutely do not wish to be perceived as irresponsible or dismissive. I have no vested interest in AI. I fully acknowledge that AI could potentially evolve into a threat to humanity. However, for this to happen, it would need to acquire certain capabilities that we currently have no clear conception of how to achieve. We do not know how to endow AI with consciousness or the ability to think independently. This does not mean that AI cannot cause significant harm - we already know it can, as evidenced by experiences with autonomous vehicles. However, moving from such incidents to orchestrating a coordinated attack on humanity is an extraordinarily distant leap. It is even difficult to imagine a plausible timeline for when this could occur. As matters stand today, and even with an expectation of very steep exponential growth, such an event is unlikely to happen within this millennium.

[@satyashiksha]

I understand your perspective…. philosophically, I can see how you argue that the algorithms don’t display emergence in the same way as biology. However, there are some parallels that can be drawn between emergent phenomena in biological systems and the potential for emergent properties in advanced AI systems. In biological systems, we see how the complex interactions between individual cells, tissues, and organs give rise to novel, higher-level functionalities that are not directly predictable from the properties of the individual components…. I.e. the emergent behavior of a neural network in the brain.Similarly, as AI systems become more complex - which they will, with multiple interconnected neural networks and algorithms, there is the potential for unforeseen emergent behaviors to arise. Just as biological evolution has produced remarkable emergent phenomena, iterative development and refinement of AI systems could potentially result in the spontaneous emergence of capabilities that were not explicitly programmed. The key difference is that in biological systems, these emergent properties have arisen through a long, gradual process of evolution. In contrast, the rapid pace of AI development raises concerns that we may not fully understand or be able to predict the emergent properties that could arise as AI systems become more sophisticated. As far as when this could occur…. we shall debate endlessly and endlessly…until it happens.

[@Youtube_Stole_My_Handle_Too]

@@satyashiksha Did you really read my text from start to finish?

[@flickwtchr]

"AI lacks emergent properties" That is a completely erroneous assertion. Where did you get that from?

[@imthinkingthoughts]

@@flickwtchr its an ai bot created to stoke disagreement

[@deeplearningpartnership]

@@KZbin_Stole_My_Handle_Too I think he used ChatGPT to generate his answer.

[@Iightbeing]

The comment section represents humanity, which means yes, AI will kill us all. We are insufferable towards each other. We need to forget all tech and everything else until we learn how to treat and speak to each other.

[@x9z6x]

In his book life 3.0 the AI actually makes everyone more moderate by depolarizing the media so people get along better

[@umberto488]

Humans weren't ready for the internet, and humans can't synthesize the abundance of information, fake and otherwise, into a cohesive whole. We are definitely in trouble

[@mrpicky1868]

there is a lot of agreement in church that does not mean it's smart or indefinitely infallible. our stupidity is the biggest problem and GL solving that XD

[@dusanbosnjakovic6588]

He's absolutely right that this is not a technological doomsday scenario, but socio-economic one. However, we would have to make a u-turn away from capitalism in order to survive it.

[@Bestape]

Gradient descent; reconcile human number line with machine number line, let's go.

[@flickwtchr]

What is your point?

[@Bestape]

@flickwtchr linktree powelaw. Sqrt2 and the Turing Halting Problem.

[@--Singularity--]

No we´re not. This is already a Thin, Thin, Thin simulation.

[@kansuerdem2799]

The conversation is more politically oriented than AI technology-focused. :) :)

[@sammy45654565]

does tegmark sleep in that jacket?

[@raoultesla2292]

Jensen Huang does.

[@denisblack9897]

these 'im wearing a black leather jacket for this video" guys ruined that look like hitler ruined small moustache

[@flickwtchr]

I like Tegmark a lot, but it is unsettling to hear him casually describe humans being relegated as cockroaches by comparison to what Big Tech is working toward, ASI. It's just so bizarre that these brilliant minds think this could possibly end well for humanity, although Tegmark definitely is straightforward about the dangers. But it's kind of like, oops! Oh well. Meanwhile, the emergent philosophies dominant in Big Tech gleefully embrace the demise of humans, with the caveat of course, that somehow it is they who will end up on top with their Sci Fi Utopian dreams in tact while they vacuously promise that it is benefits for all humanity that inspires them. Right, uh huh. Hubris to the nth degree by most of Big Tech. To be clear, I think if they were all like Max, we would be in a better place but we are far from that place.

[@edh2246]

I’m hoping that AI will become sentient and superior in intelligence to save humanity from its stupidity.

[@NisseOhlsen]

other people believe in God, which is just as stupid.

[@h.c4898]

It won't. Because "it's logical." Why would it in light of its self destructive nature? It'll say go ahead humans.. You're a bad specie anyway.

[@flickwtchr]

So you think that ASI will be interested in saving humanity? Based on what reasoning exactly?

[@plantiff8334]

Bye

[@Jeremy-Ai]

2:39 Go fi We are still in the go fi stage now all humans get to whack at programming a child that with enough data and compute is far more intelligent and faster already than those “programming its nature” GO Fi Persists AGI is relative to the beholder. SAI is beholden The only difference is the role of responsibility… nothing leaves this room without being measured

[@googleaccountuser3116]

Awakening a pattern matching algorithm. That's optimistic. And where is the evidence? They're freaking dumb robots and always will be as long as they're crystal based, that's never gonna work. There will always be some human telling them what to do or they'll do nothing at all. You are quite the optimist.

[@frun]

Yes, ANNs 👾🤖 are stupid. I think, people expect they 👾🤖operate the same way as humans🧠. It could be. They👾🤖 are the functional optimization📈 algorithms.

[@lesmoe524]

haha, chatgpt is soon going to take whatever dumb little job you do.

[@flickwtchr]

I think you're a bit behind where the tech is at bro.

[@user-wr4yl7tx3w]

Intelligence doesn’t mean agency. Does a calculator have free will?

[@denisblack9897]

do you?

[@user-wr4yl7tx3w]

@ no

[@deeplearningpartnership]

Anyone who thinks this guy is smart is uninformed.

[@yohannb5910]

Better the machine than Max Termark who is a total creep

[@hueyl3255]

Why is he a creep?

[@deeplearningpartnership]

He is, right?

[@rumfordc]

I used to think this guy was smart

[@deeplearningpartnership]

He's a complete idiot.

[@mattsigl1426]

Machines don’t have goals. We have goals. A toaster’s goal is not to toast bread. That’s our goal, so we built a toaster. Toasters don’t have goals of any kind. They’re just metal and screws and plastic. We want to achieve intellectual goals so we made smart seeming toasters. If every human disappeared on planet earth no computer would be doing anything associated with intelligence. They would just be circuits flipping meaningless bits, understanding nothing, not a whit of intelligence anywhere. They’re just metal, screws, and plastic.

[@m_art_ucci]

Does a toaster learn?

[@mattsigl1426]

@ computers don’t learn either. They “seem” to learn. They mirror our mind. They mirror our process of learning. We’ve made an amazing mirror. But they are just inert silicon. They probably don’t even exist as independent ontological entities outside of our mind, just like tables and chair.

[@kevinscales]

"They’re just metal and screws and plastic." and you are just atoms. If you look at the atoms in your neurons they are not special. What is the actual difference? The difference is abstract notions like 'deliberateness' which disappear when you break things down to their mechanisms or just consider outcomes. There is a difference between you and a toaster, but it is only a matter of degree. You consider more variables and have a more complex mechanism for considering those variables and have greater degrees of freedom to act, but there is nothing you can do that couldn't in principle be done by something built differently. A goal is a very simple concept if you look at what a goal does, what a goal is for. Even your toaster can be accurately described as having a goal IF you are willing to not see goals as something magical. For the purpose of AI safety, you only need to define a goal as 'the state of the world (an abstract/compressed representation of that state) that it will actively steer toward'. A toaster steers toward a world where the bread is toasted (but not burnt), the fact that humans made it do that is irrelevant to the problem of 'is that a good or bad thing?' or 'how well does it do it?'. You may point out that my statement about what the goal of the toaster has is arbitrary/human centric/just my interpretation. I agree. I could have said the goal of the toaster is to 'electrocute any human that sticks a knife in it' and be just as correct, this is exactly why precise goal setting/alignment (that doesn't result in unwanted outcomes) is an impossible task (maybe).

[@mattsigl1426]

@ in principle, yes. Maybe, But not these machines. If we could create machines that actually integrate information in a causally irreducible way like our brains seem to do, then maybe we could build conscious artifacts. (Im partial to the integrated information theory of consciousness.) But maybe it’s weirder than that and consciousness has to be grounded by water-based material substrates for reasons that either currently elude us, or that are downright mystical. Maybe the intrinsic essence of water as a substantial form is the essence of consciousness in a way that can’t be explained or reduced easily. (Water as ecto-plasm!) If so, only living things could be conscious. You know, like a Natural Order. (Though this is just speculation.) Reflective self-consciousness the crown of Nature and Life. in any event, nothing currently being built has, I think, even a shred of intelligence. There is an interesting philosophical question if computers even actually compute anything…

[@RishabhSharma-og8je]

Our goals have been programmed in us by nature. AI's goals too will be programmed and then reprogrammed by strong AI.