News for hackers. Stuff that matters.
Re:AI cannot be the inventor of a patent, appeals court rules by driverdan on Saturday September 25, 2021 @05:59PM (#28656557)
By AI he means mathematical equation. When you reword this to "Mathematical equation cannot be the inventor of a patent" it sounds obvious and stupid. And that's because it is. Of course an equation can't patent something.
Re:AI cannot be the inventor of a patent, appeals court rules by edouard-harris on Saturday September 25, 2021 @06:08PM (#28656599)
Given that all human behavior can (very likely) be reduced to mathematical equations too, I'm not sure this is a convincing reductio. It certainly doesn't seem obvious that a rule like this could be consistently applied if, for example, AIs became as capable as humans — or far more so.
Re:AI cannot be the inventor of a patent, appeals court rules by mrbungie on Saturday September 25, 2021 @06:12PM (#28656618)
I think a lot of people would expect some evidence for your first phrase.
Re:AI cannot be the inventor of a patent, appeals court rules by lordlic on Saturday September 25, 2021 @06:52PM (#28656863)
See https://en.wikipedia.org/wiki/Computationalism

It's a very common worldview among the tech set.

Edit: Uh, downvotes, really? I think it's totally reasonable to assume this as a given since it's, like, pretty obviously true. If anything the parent needs to provide their own extraordinary evidence for why consciousness would be non-computable. I'm just trying to provide some context for them to read up more on some stuff that apparently everyone but them (who has given it any thought) already knows about.

Re:AI cannot be the inventor of a patent, appeals court rules by karaterobot on Sunday September 26, 2021 @10:49AM (#28661307)
Downvoted because the commenter asked for evidence, and you just provided a definition. The statement that it's "obviously true" is not evidence, or even an argument. If I said it's "obviously not true", would you consider yourself refuted? Probably not.
Re:AI cannot be the inventor of a patent, appeals court rules by tremon on Sunday September 26, 2021 @06:00AM (#28659637)
You're merely repeating the GP's assertion, and completely ignoring the parent's request for corroborating evidence. What you hold as self-evident is still just your anecdata, and cannot be extrapolated to "the tech set", unless the tech set includes only you.
Re:AI cannot be the inventor of a patent, appeals court rules by lordlic on Sunday September 26, 2021 @10:52AM (#28661327)
I'm just pointing out that it's a common belief (and it is, your ignorance notwithstanding) and giving them the search term to read more if they're interested. What's so difficult to understand about this? And it's the philosophy of mind; there's not really "corroborating evidence" to present so much as there are logical arguments that you need to read up on yourself.
Re:AI cannot be the inventor of a patent, appeals court rules by edouard-harris on Saturday September 25, 2021 @06:21PM (#28656671)
You're right. I'll edit the parent to soften the assertion, though I do consider it to be virtually certain.

In terms of evidence: the simplest argument is probably that humans appear to consistently obey the laws of physics, and the laws of physics appear to be mathematical.

Re:AI cannot be the inventor of a patent, appeals court rules by ars on Saturday September 25, 2021 @09:35PM (#28657632)
Something can be mathematical and also not computable.
Re:AI cannot be the inventor of a patent, appeals court rules by canjobear on Saturday September 25, 2021 @06:26PM (#28656695)
Given sufficiently powerful computers, human behavior could be simulated. The action of the computer would be describable as a (possibly enormously long and complicated) equation. I don't see how you can disagree with this unless you think there is something metaphysically inexplicable about human behavior.
Re:AI cannot be the inventor of a patent, appeals court rules by sjy on Sunday September 26, 2021 @02:57AM (#28658973)
You can accept that the brain is governed by physical laws, without conceding that the brain can be simulated. Emulation on the atomic level may not be computationally feasible. The emergence of human behaviour from a less detailed simulation is not guaranteed. These possibilities are discussed under the heading “Scale separation” in the roadmap to whole brain emulation published by the Future of Humanity Institute [1].

> At first it may appear unlikely that a complex system with many degrees of freedom like the brain could be modelled with the right causal dynamics, but without taking into account the smallest parts … However, state variables of complex systems can be quantitatively predicted when there is ‘scale separation’: when different aspects of the system exist on sufficiently (orders of magnitude) different scales (of size, energy, time etc), they can become uncoupled … [an] example is the scale separation between electric currents and logic operations in a computer …

> When there is no scale separation (such as in fluid turbulence) macroscale predictions become impossible without simulating the entire microscale. An important issue to be determined is whether such a cut‐off exists in the case of the human brain … if it could be demonstrated that there is no such scale, it would demonstrate the infeasibility of whole brain emulation.

The FHI is assuming scale invariance in order to test it experimentally. We shouldn’t assume it for the purpose of analysing the roles played by humans and AI under IP legislation.

[1]: https://www.fhi.ox.ac.uk/brain-emulation-roadmap-report.pdf

Re:AI cannot be the inventor of a patent, appeals court rules by 2muchcoffeeman on Saturday September 25, 2021 @06:38PM (#28656774)
Basically this line of argument boils down to assuming something about something we don’t and may never have.

Until we have this technology should we adjust laws for hypotheticals? No.

Re:AI cannot be the inventor of a patent, appeals court rules by canjobear on Saturday September 25, 2021 @06:57PM (#28656894)
Put another way: the claim that humans cannot be modeled as equations equates to the claim that humans cannot be modeled using physics. Do you want to defend that?
Re:AI cannot be the inventor of a patent, appeals court rules by 2muchcoffeeman on Saturday September 25, 2021 @08:34PM (#28657365)
I’m not claiming anything with regards to math model people.

With regards to the law, you’re saying, because we assume that one day we should have a complete simulation of humans, or at least a model, therefore, we should update our laws NOW, to prepare for this?

A story of premature optimisation if I ever heard one.

Re:AI cannot be the inventor of a patent, appeals court rules by canjobear on Saturday September 25, 2021 @08:39PM (#28657389)
The claim was that it's obvious that an AI can't patent something because the AI is an "equation." I'm claiming that if you want to establish a meaningful difference between humans an AIs, which would be useful legally, it can't be that one is an "equation" and the other is not, because "an equation" could equally well describe human behavior. You'd have to search for the distinction elsewhere.
Re:AI cannot be the inventor of a patent, appeals court rules by 2muchcoffeeman on Saturday September 25, 2021 @09:16PM (#28657548)
It is legally useful since it fits with our current understanding. When this changes, the laws will change. Your scenario is completely hypothetical at this point and also kinda extrapolated from things we know now.
Re:AI cannot be the inventor of a patent, appeals court rules by Jensson on Sunday September 26, 2021 @01:30AM (#28658646)
Modelling quantum physics interactions is NP-hard, you can't do that using regular computers even if you make them as big as the universe.

It is possible you can make a good simulation of human thought without all of that, but we can't be certain.

Re:AI cannot be the inventor of a patent, appeals court rules by tremon on Sunday September 26, 2021 @06:05AM (#28659659)
No, what people are saying is that humans cannot be reduced to your abstract model. Your model is a simplification, and to insist that reality matches the model is a very myopic worldview.
Re:AI cannot be the inventor of a patent, appeals court rules by JumpCrisscross on Saturday September 25, 2021 @08:09PM (#28657237)
> claim that humans cannot be modeled as equations equates to the claim that humans cannot be modeled using physics

We don’t have a complete theory of physics. We don’t have any viable models of human thought. This is a non sequitur.

Re:AI cannot be the inventor of a patent, appeals court rules by canjobear on Saturday September 25, 2021 @08:34PM (#28657366)
We do have a theory of physics which is likely to be able to cover all the phenomena of biology[1], but that is beside the point. If you think that human behavior cannot be reduced to equations, then you think it cannot be predicted mathematically. I can only interpret this as a claim that there is something about human behavior which is supernatural.

[1] https://arxiv.org/abs/2101.07884

Re:AI cannot be the inventor of a patent, appeals court rules by JumpCrisscross on Saturday September 25, 2021 @09:46PM (#28657686)
> If you think that human behavior cannot be reduced to equations, then you think it cannot be predicted mathematically. I can only interpret this as a claim that there is something about human behavior which is supernatural.

You’ve reduced the question of whether our universe is deterministic, something the greatest minds in physics have debated for a century and continue to debate, to a sound bite. That’s disingenuous. There may be phenomena we cannot accurately model and predict for fundamental reasons that have nothing to do with the supernatural.

We are so far from scientifically understanding consciousness that we don’t even have an agreed-upon empirical definition for it, let alone a way to measure it. That gap in understanding undermines any claim that it can be modelled this way or that.

Re:AI cannot be the inventor of a patent, appeals court rules by canjobear on Saturday September 25, 2021 @10:24PM (#28657900)
Determinism is irrelevant. You can use stochastic equations to simulate human behavior if there are quantum effects involved.

Consciousness is more interesting, and it boils down to the claim that there is something metaphysically and unverifiably (from the outside) special about humans. That sounds supernatural to me.

Re:AI cannot be the inventor of a patent, appeals court rules by Jensson on Sunday September 26, 2021 @01:39AM (#28658684)
> You can use stochastic equations to simulate human behavior if there are quantum effects involved.

No you cannot, quantum effects are way more complex than just random chance effects. The quantum waves are particles in their own rights that flows and interacts with each other and itself, if you treat it as just a stochastic variable you don't get quantum effects.

Re:AI cannot be the inventor of a patent, appeals court rules by canjobear on Sunday September 26, 2021 @11:17AM (#28661519)
You just use the Born rule to get your probabilities from the wave function…
Re:AI cannot be the inventor of a patent, appeals court rules by blix on Saturday September 25, 2021 @10:48PM (#28658022)
It sounds to me that you are arguing that consciousness fundementally doesn't exist; what we think of consciousness is an emergent property of fundementally determinable physical rules.

If this is the case why are you arguing on HN? Is it possible for you to leave the discussion? Is talking to you the same as dropping a ball and watching gravity pull it downwards?

I'm a little skeptical of your stochastic model of humanity as well. Statistical mechanics works because oxygen molecules don't go off and start religions and convince other oxygen molecules to behave radically differently for thousands of years (or at least we haven't ever seen it). In fact our stoachstic physical models, classical, semi-classical and quantum, kinda specifically preclude this. How would you design an equation to capture this real phenomenon in humans?

Re:AI cannot be the inventor of a patent, appeals court rules by EarthLaunch on Saturday September 25, 2021 @09:58PM (#28657747)
This really makes me think. Unknowable is often mistaken as supernatural. Could human behavior possibly be physically unknowable? There are potential unknowables: Non-observable universe, destructed past information, progressively fundamental particles.

Even with indestructible information, combined with locality at the big-bang, combined with a perfect fundamental physics, is there enough time/energy to compute that? Perhaps this is grasping at straws.

Re:AI cannot be the inventor of a patent, appeals court rules by reverend_gonzo on Saturday September 25, 2021 @06:35PM (#28656751)
Sure, and those sufficiently powerful computers don’t exist yet.
Re:AI cannot be the inventor of a patent, appeals court rules by canjobear on Saturday September 25, 2021 @07:00PM (#28656903)
The actual existence of the actual machine is irrelevant for the claim that human behavior can be reduced to an equation. What is it about human behavior that you think could not be simulated by a machine?
Re:AI cannot be the inventor of a patent, appeals court rules by robbedpeter on Saturday September 25, 2021 @07:03PM (#28656913)
The brain is a biological system of interconnected neurons. Neuroscience has shown neurons operate digitally - on and off. The brain is a biological computer. There is no magic, no mystical quantum "thing" enabling human intelligence. We're not picking up some metaspectral soul transmission from an alternate universe. We're advanced binary pattern processors, and we're only a handful of orders of magnitude more powerful than consumer level hardware.

The algorithm we run on is special, but it's physically encoded and self generating, based in our DNA and physical development. We're vastly complex but entirely built of relatively simple fundamental building blocks.

We are computers, in the abstract sense of Turing machines and such concepts.

Re:AI cannot be the inventor of a patent, appeals court rules by aaaaaaaaaaab on Saturday September 25, 2021 @07:20PM (#28656993)
>Neuroscience has shown neurons operate digitally - on and off.

[citation needed]

Re:AI cannot be the inventor of a patent, appeals court rules by porkbrain on Saturday September 25, 2021 @08:06PM (#28657227)
Parent refers to the behaviour of generating action potentials which travel to synapses connecting (mostly) axons with dendrites. The continuous (analogue) signal processing pre- and post- synapse seems to be a massive part of how neurons achieve the computation they do. A hypothesis I subscribe to claims that it is cheaper to compute locally with molecule gradients and protein folding than with spikes. Brains leverage that. The specific statement you quoted is to the best of my knowledge inaccurate.

I am not an expert, my understanding of neurons is largely is based on [0], [1], [2].

[0]: https://mitpress.mit.edu/books/principles-neural-design

[1]: http://www.sci.utah.edu/~macleod/bioen/be6003/notes/W08-spru...

[2]: https://mitpress.mit.edu/books/functional-connections-cortic...

Re:AI cannot be the inventor of a patent, appeals court rules by robbedpeter on Saturday September 25, 2021 @09:45PM (#28657682)
A neuron fires, or it doesn't. The patterns received by neurons via synapses are binary. The synapses themselves are analog, and the activation can trigger in several ways, leading to complex behavior, but the signal each individual neuron receives from any and every other neuron in the brain is digital.

Where an impulse comes from, when it is sent, the current charge state of the receiving neuron, the specific neurotransmitters released by the axons, and the efficiency of signal transmission at a given point in time all play a part in the signal sent along dendritic, but at any given point in time a neuron is firing or isn't, and that's the only state that matters to the other neurons it's connected to.

The biggest difference between biological brains and artificial neural networks is the analog nature of biological timing compared to the arbitrary cycles of software.

Spike timing dependent plasticity networks address some of the problem of variable timing and its impact on signaling.

https://en.wikipedia.org/wiki/Spike-timing-dependent_plastic...

At any given moment, you could capture the state of a biological brain as a binary number. The model would need to account for all sorts of analog mechanisms and a huge, biologically messy graph of connections with dozens of variations across a handful of neuron types, and the length of a synapse and where it physically connects or touches other synapses without connecting directly and... lots of stuff that's hard to model and even hard to see.

I'm not arguing that every part is digital or even equivalent to computers on silicon, simply that the basic mechanism of computation is itself digital pattern processing. It literally cannot be anything else.

There is no evidence or theory to indicate any other plausible method of operation. All the analog stuff can be swapped out with an arbitrary bit length representation and the functionality is retained. Remove the digital assumption and everything breaks. You need magic or spirit to make it work, which is ridiculous because silicon computers proved that binary computation is a perfectly fine way to represent and process data. Quantum processing in biological brains is magical thinking performed in a desperate attempt to ascribe eminence to human cognition, because for some reason people don't like the idea that brains do digital processing.

We are messy and vastly complex and sophisticated digital pattern processors. Which makes evolutionary sense, because that's how our predecessors work, all the way back to single cell organisms and self replicating DNA chemistry.

Signals and patterns in the real world can be represented as binary to an arbitrary precision. The Planck length dictates the maximum level of precision relevant to our universe.

https://www.researchgate.net/post/How-many-bits-of-informati...

Human brains achieved a special state of depth and breadth of processing, concurrent with physically encoded, evolved algorithms in our neural architecture. The arrival of biology at human cognition affords us the privilege of deep reflection, the apparent subjective reality we experience, and the thus far unlimited scope of conception such that we can make approximations about the size of the universe or the mass of a photon. Or the operations of the minds we possess.

We operate in binary - the explanation is sufficient, simple, and corresponds to all known evidence. There aren't any outstanding questions requiring that premise to be questioned.

Re:AI cannot be the inventor of a patent, appeals court rules by enchiridion on Saturday September 25, 2021 @07:28PM (#28657030)
And we were clocks a few centuries ago. I wonder if there is a name for the bias of thinking our modern abstractions are the right ones.
Re:AI cannot be the inventor of a patent, appeals court rules by Dylan16807 on Saturday September 25, 2021 @07:52PM (#28657147)
I don't understand the distinction you're making in this particular case. Clockwork and circuits have very similar abilities.
Re:AI cannot be the inventor of a patent, appeals court rules by robbedpeter on Saturday September 25, 2021 @10:28PM (#28657916)
Some of the literature in neuroscience and ai perpetuate this meme, going something like "medieval people thought their minds worked like catapults, Renaissance people thought of clockwork, early industrial age people thought it was like steam engines, and now in the digital age we think it's like computers, oh aren't purple so silly."

It's a superficial and unfounded observation that tries to frame neural computation as a bias or conceit of modern culture. Not how science works, and the argument is condescending as hell.

Re:AI cannot be the inventor of a patent, appeals court rules by frabcus on Sunday September 26, 2021 @04:25AM (#28659286)
The earliest place I’ve seen this is the introduction to Julian Jaynes “Origin of Consciousness in the Breakdown of the Bicameral Mind”

https://www.julianjaynes.org/resources/books/ooc/en/introduc...

The way he does it is quite compelling - we do make metaphors of the mind to current tech. This was written well before current computer neural networks, and gives a good account of the sound version of this argument.

Re:AI cannot be the inventor of a patent, appeals court rules by Dylan16807 on Saturday September 25, 2021 @10:36PM (#28657960)
It can be an interesting thing to look at, much like the idea of remote harassment with an air loom.

But I agree that it's a really shallow way to dismiss arguments. In this case, it's basically just saying that a brain is deterministic and made out of parts. Clockwork and steam engines and circuits all work fine. I can't comment on the catapult without more elaboration.

Re:AI cannot be the inventor of a patent, appeals court rules by enchiridion on Saturday September 25, 2021 @11:02PM (#28658083)
I’d say it’s anything but shallow. What I’m getting at is that there may be things completely beyond our current understanding related to consciousness. As in, multiple paradigm shifts away.

Without getting into the weeds too much, I just find the certainty of the systems line of thinking a little unfounded.

Re:AI cannot be the inventor of a patent, appeals court rules by robbedpeter on Sunday September 26, 2021 @12:59AM (#28658513)
Universal Turing Machines and number theory and formal logic aren't going away - the universe runs on math, and our brains operate in binary. Those statements are ground truth - there's no room for breaking things down further.

Consciousness is an algorithm performed by a computer with messy biological parts, some of which are analog, which may have serious implications for the level of strictly digital computer power you need to successfully implement the algorithm in silicon. The same applies to intelligence, and it's an open question as to what exactly the relationship between subjective experience and intelligence is.

You are a brain in a bone vat experiencing a subjective construct produced by a biological computer processing binary states consisting of internal feedback and sensory streams made entirely of patterns distinguished at any moment in time by the presence or absence of electrical impulses. Each neuron can be receiving a pattern of up to ten thousand individual bits, and whether the neuron fires or not depends on the current state and precise pattern of inputs.

Everything you experience, think, or do, is determined by those electrical patterns.

Re:AI cannot be the inventor of a patent, appeals court rules by enchiridion on Sunday September 26, 2021 @09:21AM (#28660666)
Well certainly, in some sense they are fundamental to the universe. But in this situation it’s a bit like saying the photon isn’t going away. Sure, but the strong link to consciousness isn’t there.

However, the universe doesn’t run on math, it runs on itself. Math is the best way we have to describe it at the moment. It used to be that divination was the best method. Why are you so certain math in it’s recognizable form today will be the method used to describe the universe in 100k years?

Our brains certainly do not run on binary. I think you might be confusing the information content, represented in bits, with the actual implementation. At the very least dendrites take in and mix signals of various strengths. The threshold for triggering and strength of the action potential are dependent on many chemicals signals. And that’s even a dated understanding which gave rise to the ideas of artificial neural networks. I’m sure the current understanding is much more complex and nuanced.

You brush off the analog parts as an implementation detail, when it could have very real implications for the computability of consciousness. The thing is, we don’t know.

But even all that is besides the point. The main point is, we don’t have the slightest idea how consciousness works. To the point that no one can even figure out how to ask the right questions about it.

In the face of such great uncertainty, it is anti-scientific to state with such conviction that you have the correct model. It’s fine to present it as a hypothesis, but the certainty is unfounded.

Re:AI cannot be the inventor of a patent, appeals court rules by robbedpeter on Saturday September 25, 2021 @10:58PM (#28658068)
As far as catapults go, or really the whole premise, I don't think it was ever considered in a serious way, except to propagate through various papers and books because it sounds clever. It's a bad meme.
Re:AI cannot be the inventor of a patent, appeals court rules by robbedpeter on Saturday September 25, 2021 @10:22PM (#28657888)
You can't abstract further than bits. 0's and 1's all the way down!
Re:AI cannot be the inventor of a patent, appeals court rules by u10242 on Sunday September 26, 2021 @06:37AM (#28659803)
Bye bye free will!

From now on, nothing is my fault any more.

Re:AI cannot be the inventor of a patent, appeals court rules by howaboutnope on Saturday September 25, 2021 @10:59PM (#28658071)
> Given that all human behavior can (very likely) be reduced to mathematical equations too

I can reduce all the music in the history of the world to the phrase "rhythm, melodies and lyrics". So? That'd say nothing about all that music, it'd just say something unflattering about me.

> It is not the greatest of modern scientists who feel most sure that the object, stripped of its qualitative properties and reduced to mere quantity, is wholly real. Little scientists, and little unscientific followers of science, may think so. The great minds know very well that the object, so treated, is an artificial abstraction, that something of its reality has been lost.

-- C.S. Lewis

Re:AI cannot be the inventor of a patent, appeals court rules by naasking on Sunday September 26, 2021 @01:58PM (#28662987)
The Bekenstein Bound entails that people can be fully captured by a finite amount of information, and therefore a person is no more powerful than finite state automoton. Note that this is an upper bound, and refining our understanding of the the physics involved will only reduce the amount of information needed.

The only escape hatch is to assert that humans can't be described by physics.

Re:AI cannot be the inventor of a patent, appeals court rules by mr_toad on Sunday September 26, 2021 @06:20PM (#28665187)
> The only escape hatch is to assert that humans can't be described by physics.

Or that there is some physics that can’t be described by computation. It seems vanishingly unlikely, but there is no actual proof that everything physical is computable. And even if there was, it probably has nothing do do with minds.

Re:AI cannot be the inventor of a patent, appeals court rules by naasking on Sunday September 26, 2021 @09:14PM (#28666126)
Right, not an escape hatch for human specialness though. We haven't seen anything that's uncomputable in physics yet either. Quantum computation "only" gives polynomial speedups so far.
Re:AI cannot be the inventor of a patent, appeals court rules by Valmar on Sunday September 26, 2021 @12:25AM (#28658384)
> Given that all human behavior can (very likely) be reduced to mathematical equations

There is no scientific evidence whatsoever for this very bold claim.

Re:AI cannot be the inventor of a patent, appeals court rules by tired_and_awake on Saturday September 25, 2021 @06:10PM (#28656609)
> human behavior can ultimately be reduced to mathematical equations

Wait what? Source?

Re:AI cannot be the inventor of a patent, appeals court rules by Goety on Saturday September 25, 2021 @09:21PM (#28657569)
Human behavior is 100% able to be reduced to mathematical equations. Full stop. In aggregate and in day to day observation.