Three questions that keep me up at night
blog.evjang.com
blog.evjang.com
If you believe that the human mind is the result of Darwinian evolution, then you implicitly believe that thinking can arise from interacting with the environment. That does not exactly answer the question, but if anything, it goes beyond it.
On the other hand:
"This could also be addressed via other high-data environments:
1) Observing trader orders on markets and seeing if we can recover the trader's deductive reasoning and beliefs about the future. See if we can observe rational thought (if not rational behavior).
2) Recovering intent and emotions and desire from social network activity."
Before you could even begin to get anything from these streams of symbols, you would have to have some way to give them semantics - to understand what they signify to the actors. I suspect that these would be among the last things you would achieve, not a means to the end.
Q2) This would require the knowledge of what is sufficient & necessary for AGI.
Q3) Minsky, Neumann, and most sophisticatedly, Norbert Weiner answered this question. Read page 33 of Weiner's Cybernetics. The real insight, where in what I will now explain in brevity is at the expense of the insight's resolution, is that in the nature of microscopic phenomena there are an exorbitant quantity of fickle components ("equal particles") whose positions, accelerations, velocities are all impossible to record because of that sheer quantity by the nature of this sphere of study. More so, there are then at least second-order coupling effects between all these fickle components due to their constant interaction. As a resolve, we have developed terms of statistical understanding. Weiner cites "cloud, temperature, turbulence, etc." all being concepts which are specifically defined in a statistical context.
Where as in the macroscopic sphere there are far fewer particles much further apart from each other with near certain initial masses, positions, and velocities.
The result? In the microscopic systems, because their properties are precisely defined statistically the distribution which predicts the future nor past for a given system isn't unique. That is, the future state (predicted by the distribution provided by some initial system) could have also been predicted by other distributions describing many other initial systems unless by "some miracle" there is a "tightly defined statistical range".
So are microscopic physics condemned to this paradox? Not necessarily, there may be some way to construct scientific objects from systems phenomena which don't require a statistical conception because they can be precisely defined from a microscopic systems point of view temporally & spatially. However, this would require a revolution in theory and likely mathematics
First, note that CPT-symmetry can be extended up to the macroscopic level. Einstein's thought experiments about time involve macroscopic entanglement. For example, given that a red ball and a blue ball are each in their own boxes, and one possesses one box, then the box is entangled to have either a red or blue ball inside. We can use linear logic to reason about these situations; one holds the multiplicative disjunction of red and blue.
Second, note that in order to make MWI's numbers work, we need not just branching of possibilities, but also coalescing. Imagine that, in empty space, we suddenly have a single virtual-particle pair; the diagram would look like () with time vertical. At the beginning, we have one universe; during the split, two universes, one for each chirality of the pair; afterwards, one universe. When we split, the probabilities of each possible universe decrease. Now, imagine doing this with more pairs, and verify that the Binomial Theorem manifests. This gives the same rise to macrostate-like entropy as Boltzmann's approach to thermodynamics, but only using QM particles and QM logic; a QM state is likelier when more of the many worlds in MWI would support its existence.
Putting these together, we may conclude that Loschmidt's paradox is an incomplete part of a bigger paradoxical complaint: We can see that there are time-reversible and time-irreversible behaviors, both in particles and in macroscopic systems. The time-reversible behaviors seem to govern resources; the time-irreversible behaviors seem to govern entropy. The two logics for these two sets of behaviors are distinct; linear logic and intuitionistic logic can intertwine, but here they don't appear to do so. Why not?
The confusion here is that my statement has nothing to do with cybernetics. Although it’s taken from his book titled Cybernetics, almost 60% of the book isn’t about cybernetics but instead a philosophy and analysis of Science as it progresses through history.
Though it’s fair to Wiener to preemptively note there is no failure in epistemology in one of the most successful theories which holds ground in neurons, to electrical circuits, traffic engineering, the abstract feedback loop. There’s no assumption that “we know nothing of particles”. Wiener originally wrote the book in 1948 and completed the second edition in 1961. He made significant contributions to quantum mechanics[0]. It’s quite suspect to make such a defamatory remark about Wiener’s cybernetics given his quantum knowledge and the field as a whole.
In fact, what would cybernetics possibly be predicting about the interactions between a particular system of which each component cannot exhibit a property? Surely some property must be being reinforced by means of itself or another auxiliary property.
QM & Cybernetics answer completely separate questions.
You’re missing the bigger picture. The irreversibility in any physical phenomenon is only through scientific paradigms which are probabilistic—such as thermodynamics, temperature, cloud, turbulence etc. These concepts are simply more powerful in dynamical systems, which are always due to microscopic phenomena, where there are no analytical, deterministic solutions for the otherwise favored deterministic paradigms (e.g. three-body problem).
“while when a great number of atoms (of the order of Avogadro’s number) is considered, this energy related to irreversibility becomes so large that its order of magnitude must be taken into account. Consequently, macroscopic irreversibility results related to microscopic irreversibility by flows of photons and amount of atoms involved in the processes.”
[0] https://www.ams.org/books/psapm/052/psapm052-endmatter.pdf [1] https://www.nature.com/articles/srep35796
you mean like, the last question?
http://www.informationphilosopher.com/solutions/scientists/f...
If we assume molecular chaos hypothesis (independent particle velocities prior to collision) and take into account probabilistic collisions occurring at the 2-particle microscopic level, then post-collision velocities are now dependent, thereby imposing an ordering to time.
I have to think more about this though, and I'm still struggling with understanding the circular dependency between entropy "being time" and entropy being "caused" by "motion" (with respect to ... time?)
The way I think about this (from Feynman) is that entropy measures "how special" a macroscopic state is, on average. When you apply many microscopic independent fluctuations to a given state you are unlikely to end up with a "more special" state. That's how states are ordered at the macroscopic level giving the arrow of time a clear direction.
I agree with your observation of what happens to the black and white billiards. My question is - if the interactions are all time-symmetric (implies conservation of entropy), where does the entropy come from in the macroscopic system? My belief is that the microscopic model is flawed; if we simply introduce some randomness into the collision dynamics of two particles (which is believable given particle accelerator experiments / QM), then microscopic interactions are no longer T-symmetric.
It comes down to a philosophical interpretation of whether you can treat a system of two particles probabilistically (i.e. thermodynamically in aggregate) or not.
Counterpoint: hummingbirds.
It might sound cheap, but discussions of the "basic necessities" of anything so often just overlook obvious counterexamples.
What is a "stable grasp"? An entire body of classical robotics literature is devoted to answering this problem. I work on robotic grasping research and defining what it means to grasp something successfully is a little tricky unless you have a broader context of what the grasping is for.
Nature doesn't care whether hands exist or not - only that the animal can feed. That is why we see such a spectrum of appendages in the animal kingdom. Some are hand-like, others not at all, and some in between. In some sense, "hand" is a mental construct of human language.
For me: How do viruses really work? How to antiviruses work? I’ve gone into basics of DNA -> RNA -> Ribosome -> protein -> cell organelles -> Cells. We know how to sequence DNA and it’s pretty cheap right now. How do we simulate compiling DNA -> proteins and how they interact with each other? How do you find weak spots in a virus and figure out what proteins will inactivate them ?
2) what is the best course of action for people who lost their jobs? Taking <20k from 401k/IRA should be free of taxes and penalties. Everyone has lost >30% of value since Feb already. Sure most people don’t have 401ks, but those who do could weather this storm a bit better. Should we stop payments for all loans for 2 months. No mortgage, auto loans, student loans etc. stop those foreclosures from happening. No rent.
3) will US have the most deaths? Most likely. A large population + most states haven’t locked down. This means unless >80% of population gets infected and recovers to build herd immunity, or we have a vaccine, we’re in deep trouble. Should we infect everyone >10 and under <30 and in clusters so we build up herd immunity in weeks? And get most of the young population back to work? Like if you’re gonna be sitting at home for two weeks, may as well get sick and build up immunity while we’re at it.
Deep fakes can be done on older hardware, we just didn't have the code. Does the same thing apply to AI?
This article is an excellent introduction:
http://www.dreamsongs.com/Searle.html
A relevant excerpt:
Searle goes on to contend that intentional behavior must be in response to real contingencies, not formal simulations of reality. If so, it would seem to follow that the capacity for intentional behavior is the result of what might be called phylogenic contingencies, and so would only be found in products of real evolution. In short, what Searle is saying is that, just like other biological activities such as digestion and photosynthesis, thought is intrinsically dependent on the biochemistry of its origin. Just as a formal simulation of digestion is not really digesting, a formal simulation of thought is not really thinking. And to tie this to the gender thought experiment, a formal simulation of gender is not gender, because it lacks biological structural and functional characteristics.
The "Simulation, Duplication and Evolution" section of the SEP article is also useful:
https://plato.stanford.edu/entries/chinese-room/#SimuDuplEvo...
On the other hand, at least at the time of writing the original paper, he clearly misunderstood the 'systems reply', as his response to it -- supposing that the actor had memorized the instructions -- is beside the point, and makes no difference.
We start with Aaronson's observation [0] that the Chinese Room's complexity can be analyzed. Let us assume that the Room responds in (N)P time, and in particular not EXP time. Then the dictionary must be poly-sized. Further, the dictionary is finite but the input strings are unbounded in length, so that we must be able to divide and conquer somehow. Therefore we only need log-many steps to walk the dictionary. The dictionary is ordered, and so indexed, and so walked in log-time.
We can conclude that Searle's definition of intelligence is basically the ability to execute certain kinds of basic katamorphisms, akin to pregroup grammars [1] or other categorical grammars; in other words, Searle says that intelligence is merely the ability to talk convincingly like a human, and not necessarily any of the other features that we see across Mammalia.
Searle might retort that the Room knows facts about the world. This would require us to augment our pregroup/category, enriching it with relational information. For example, in the utterance "I like cats", we might annotate each word with a relation or type, and then look them up in a secondary database-book. We can see here that it would then be the database-book which has an opinion on cats, since "I" becomes merely another symbol in the database. This is reminiscent of Hofstadter's critique of Searle (supported by his other work), in which he suggests that "I" is indeed merely a self-referent symbol which leads to a strange loop, or as he puts it, "I am a strange loop." Here, the strangeness of the loop is provided by the ability of "I" to be both a plain word in the textual language and also a referent symbol in the symbolic database.
Edit: Forgot the main takeaway: When Searle and Turing measure intelligence based on a person's ability to respond to textual queries with textual answers, but the entire interface between the tester and the person is the ability to give textual answers to textual questions, then we aren't measuring intelligence anymore. We've struck Goodhart's Law [2] and we can imagine that the most successful people in these tests will be those who can abuse glitches in text-processing systems.
[0] https://www.scottaaronson.com/papers/philos.pdf
IIRC, Searle's claims to show that the instruction book could not be written, and therefore the room could not exist, so he presumably does not attribute any properties or capabilities to it.
https://philosophynow.org/issues/46/Newtons_Flaming_Laser_Sw...
- "there is every reason to believe that a human brain is a machine"
Given what evidence?
- Platonism died for mathematicians some centuries ago, and simply looks silly.
And yet Frege and Gödel, two of the most accomplished mathematicians of the past century, could be described as Platonists. Clearly this isn't a closed question.
https://en.wikipedia.org/wiki/Philosophy_of_mathematics#Plat...
https://plato.stanford.edu/entries/platonism-mathematics/
Regarding digestion: if the Turing Test is determining whether a computer is thinking like a human being, then focusing only on input and output, and not the 'being', sort of misses the entire point. Remember that Searle's Chinese Room is primarily a criticism of the Turing Test.
As always, the truth is: it's more complicated than it seems.
To me the Turing test gives a falsifiable experiment that defines what "thinking like a human being" means.
It does so by focusing only on the input output. To do that, it puts the system under test out of view of the experimenter and mediates the experiment with written text.
So yes, the input output is the important part, the Turing test, albeit flawed, is interesting, and any claim that machines can't think based on a definition of "think" that does not lend itself to a falsifiable experiment, such as Searle's, is uninteresting.
I will leave it at that and make a quick exit, as Adler suggests.
Well, one could just as well pose the same question to any dualist.
The evidence for minds being physical phenomena comes, at least in part, from the observation that minds appear to be the result of physical brains, which have evolved through a physical process from much simpler precursors, doing physical things. This is far from proof, and is only a hint of an explanation of how it all actually works, but it is as least as good evidence as anything offered by dualists, who do not seem to have anything to offer in the way of a constructive explanation of minds -- they seem to spend all their time arguing against materialism, rather than for a particular idea of how minds work.
If thought is indeed "intrinsically dependent on biochemistry...", then we have a hard journey ahead.
However, AI philosophers should try their own medicine and present claims with real contingencies, i.e. make claims that are experimentally falsifiable.
Put another way, I'm less interested in debating what "understanding" is and more interested in figuring out how to build a Chinese room to begin with. Once we have such a system, perhaps we might realize that the Chinese Room thought experiment was ill-posed to begin with.
He does indeed, but the Chinese Room paper presents no compelling argument for that view; it ultimately depends on an appeal to one's intuitions about the feasibility of a remarkable and unrealistic situation that is far from our own experiences, and one, moreover, where he misunderstood the most common response (the 'systems reply'). Dennett has objected that these sort of arguments (or, as he calls them, 'intuition pumps') are simply unsuitable for proving anything, comparing them to denying heliocentricity on the grounds of it being just obvious that the Earth is not moving.
NO as you are measuring actions and not the process that instigated those actions.
One example - chess - you see the move, but the thinking behind that move may elude. More so a bad move, as their thinking was that it is a good move when it is not.
[EDIT format/spelling]
My gut translates that idea as: Progress is made though War not Markets.
Which seems dubious to say the least.
Given an appropriate reward function, and multiple independent actors seeking advantage, some might converge on an optimal strategy for attaining AGI.
Anyway, that's the way I read it.
https://www.economist.com/open-future/2018/09/10/can-inequal...
If intent could be inferred- the meaningful and material portion of human privacy would cease to exist.
>Q1: Can we imitate "thinking" from only observing behavior?
>Q2: What is the computationally cheapest "organic building block" of an Artificial Life simulation that could lead to human-level AGI?
>Q3: Loschmidt's Paradox: How does T-asymmetry in macroscopic phenomena arise from T-symmetric (microscopic) physical laws?
2. What is your quest?
3. What is your favorite color?
Watch out for that last one.
This must be a reference I’m not getting?
No, YELLLLLLOOoooooowwwwwwww!
2. Really?!
3. You?
The ones on my mind are “how long will large populations be locked down?” “How many people are going to die?” and “how long until this all happens again?”
But life is a long game.
The media is incentivized to distract you. Turn off your computer, read a book, and don’t forget to live.
As an individual, there's only so much you can do and there's no need to keep yourself distracted from other matters.
You can stay tuned in to the world around you without collapsing on the ground in hysterics.
For those who make it through, they'll be wiser and more prepared for similar situations - and their descendants may be too, at least for a while.
Is it really the large populations that worry you or is it just the risk that you might be next or even more hypocritic, that your investment portfolio is going south? Very bad and innevitable horrors have been happenning for a while(i.e the war Yemen, Iraq) and the "large populations" that you mention had little to no insomnia.
And they’ve had very little effect on me. It’s not irrational to care at different levels depending on that.
https://news.ycombinator.com/newsguidelines.html
Your comment might be just fine in an in-person conversation, but if you respond to someone this way in a broadcast medium [1], it comes across as a kind of social enforcement and shaming, especially if (like this one was) it sits at the top of a thread accruing mass—which happens easily because because indignation routinely attracts upvotes [2]. The OP has every right to think about what interests him, nor does it somehow follow from this that he doesn't care about people dying.
[1] Which HN is: https://news.ycombinator.com/item?id=22780144
[2] https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...