Posts
Comments
It remains extremely likely that the em drive will turn out to either not work or work only by known-but-improperly-accounted-for physics (see also: Pioneer anomaly).
Ah yes, pausing ghostery seems to fix it.
Clicking the "Donate now" button under "PayPal or Credit Card" does not seem to do anything other than refresh the page.
(browser Firefox 48.0 , OS Ubuntu)
We know that whatever they did led to Paris and Brussels
Correlation / Causation?
I have taken the survey.
As I understand it, "This painting is beautiful" is completely equivalent to "I like (the visual aspects of) this painting".
Definitional arguments are not useful. Even using your interpretation, the point stands: the statement, properly understood, is empirical truth.
Why bring the brain into it?
No particular reason.
"This painting is beautiful" is a statement about the reaction of the speaker
That is what I mean, yes.
Or, paralleling Good_Burning_Plastic, a statement about the reaction of people generally
Whether we define beauty to be the reaction of the speaker, or the reaction of the majority of a certain group of people that are similar to the speaker, is not relevant: in both cases "This painting is beautiful" becomes an empirical truth instead of an "affective" truth.
No. "This painting is round" is a statement about the properties of the painting itself, independent of any observer. "This painting is beautiful" is a statement about the reaction of the speaker's brain upon seeing the painting. The syntactical similarity between those different kinds of statements in English (and all other natural languages that I know of) is unfortunate to say the least.
I do not think we should dilute the meaning of the word "truth" like this.
If I say "This painting is beautiful", I mean "my brain produces a pleasant reaction upon seeing this painting". The latter sentence is empirical truth. See also 2-Place and 1-Place Words
"This place feel right to me" -- true! Affectively true.
Also empirically true!
Shakespeare is truth
If by this, you mean "I like Shakespeare's writing" (an empirical truth), just say so.
All advertising for a product should be produced by a competitor of the company that makes the product. This should be required by law.
Truth in advertising laws should keep the thus-produced advertising more or less factual. It would be much less annoying and manipulative than current advertising.
However, arguments such as "you can't exactly specify what you want it to do, so it might blackmail the president into building a road in order to reduce the map distance"
The reason that such arguments do not work is that you can specify exactly what it is you want to do, and the programmers did specify exactly that.
In more complex cases, where the programmers are unable to specify exactly what they want, you do get unexpected results that can be thought of as "the program wasn't optimizing what the programmers thought it should be optimizing, but only a (crude) approximation thereof". (an even better example would be one where a genetic algorithm used in circuit design unexpectedly re-purposed some circuit elements to build an antenna, but I cannot find that reference right now)
me and a pretty large community of real AI builders who consider a utility-function-based goal stack to be so unworkable that it will never be used in any real AI.
Just because the programmer doesn't explicitly code a utility function does not mean that there is no utility function. It just means that they don't know what the utility function is.
Applause!
I like the minimalist UI - compare with far more cluttered sites like memrise and fluentu .
you mustn´t make a religious belief into a premise for science
I strongly disagree. If religion were true, that would be exactely what you should do.
Of course you can´t mix up scientific work with religion.
Why?
That statement is widely accepted today, but it is only widely accepted because virtually all attempts to do so have failed.
What happened is the following: people did try to base science on religion, they did make interesting predictions based on religious hypotheses. By elementary Bayesian reasoning, if an observation would be evidence for a religion, not observing it is evidence (though possibly weak evidence) against that religion. That is hard to accept for religious people, thus they took the only remaining option : they started pretending that religion and science are somehow independent things.
Imagine - just imagine! - that Decartes did find a soul receiver in the pineal gland. Imagine that Newton did manage to find great alchemical secrets in the bible. Imagine! If that would have happened, do you think anyone would claim that "of course you can´t mix up scientific work with religion" ?
If you can only think of Francis Collins
I did say the only relatively well-known one, not the only one. Would you prefer if I used as an example Frank Tipler or Immanuel Velikovsky, both of whom make up exceedingly implausible hypotheses to fit their religious worldview, and are widely considered pseudoscientist because of that? Or Marcus Ross, who misrepresented his views on the age of the earth in order to get a paleontology phd?
No, today's good theistic scientists, to the extent that they still exist, are precisely those who have stopped to take religion seriously as a scientific hypothesis.
he had strong interests in Eastern religions
Being interested in religion does not a theist make. Nor does merely acknowledging the possibility of an unspecified creator entity, the simulation hypothesis is not theism.
Why did you mention him then?
He is the only well-known example of a modern theistic scientist that I can think of.
Why not mention Erwin Schrödinger or Heisenberg for example?
Both are dead, and I am not familiar with their thoughts on religion.
I looked up Schrödinger on wikipedia, and there it is : "Despite being raised in a religious household, he called himself an atheist.".
obscure hobby quite secret and separated from his scientific work
To someone who truly takes a certain religion seriously as a scientific hypothesis, attempting to extract non-obvious information from that religion's holy book is scientific work! The book was supposedly written by, or inspired by, an omnipotent being. How could they not expect to find important clues in there?
What do you base that last sentence on?
The complete and utter lack of modern theistic scientists looking for a soul-body communication organ, to name just one example.
There many scientists today whom are also theists.
People such as Francis Collins, who claimed to have converted to christianity after seeing a three-part frozen waterfall, which he interpreted as a sign of the holy trinity? Even though 3 is a significant number in more religions than I can be bothered to count ? No, such people are not worth mentioning in a serious discussion of this subject.
Besides, not all old theistic scientists based their science on religious premises.
Very true. They would hardly have made much progress if they did!
Many tried to do it though. Another example is Isaac Newton, who tried to extract scientific information from the bible.
My point here is not that their conclusions were wrong, but that their attitude towards religion was a scientific one, an attitude rarely seen in today's theists.
I see you mention Rene Descartes. He believed in the existence of souls, and, taking that hypothesis seriously, he concluded there has to be a way for the soul to send signals to the body. He went looking for an organ that might fulfil this purpose, and concluded that it is the pineal gland.
This conclusion is false, the true function of the pineal gland is known today, but it illustrates a point : the old theistic scientists tended to take religion seriously, they viewed it as a valid scientific hypothesis whose implications in the real world could be studied. In that sense, they were wholly unlike modern-day theists who all too often seek refuge in unfalsifyability.
The fact that quantum mechanics conserves energy is stronger evidence for the hypothesis that reality conserves energy than the fact that classical mechanics conserves energy. He is saying "our best model of reality conserves energy" which is very relevant.
I make the following prediction : the transfiguration exercise of ch. 104 foreshadows the possibility of safely transfiguring a certain kind of explosive, that relies on containing several components that will explode upon contact. The ch. 104 exercise tells us that containment chambers can be formed first, and their contents afterwards, such that the bomb will not accidentally explode during transfiguration.
binding constraint that you absolutely must make sure he doesn't say certain words at school
What would happen if he did say those words at school? Would they expel him? Does he know what the consequences of saying those words at school are, and does he think these consequences are insufficiently bad to act as an effective deterrent?
There are other options. Especially in cases where requiring this is illegal, forging such a proof may be an option. Or, answer "If I give you proof that my previous salary was X, I will precommit to only accept this job if you pay at least X + 20%".
Why would anyone be able to sell an item with a given pricing scheme like 1/n?
On downloaded, digital goods, this would be simple.
If their competitor is undercutting them on the first item, they'll never get a chance to sell the latter ones. And besides there's no reason such a scheme would be profit-maximizing.
Please see the numerical example in this comment
I imagine the following:
Suppose 2 movies have been produced, movie A by company A and movie B by company B. Suppose further that these movies target the same audience and are fungible, at least according to a large fraction of the audience. Both movies cost 500 000 dollars to make.
Company A sells tickets for 10 dollars each, and hopes to get at least 100 000 customers in the first week, thereby getting 1000 000 dollars, thus making a net gain of 500 000 dollars.
Company B precommits to selling tickets priced as 10 f(n) dollars, with f(n) defined as 1 / ( 1 + (n-1)/150000 ) , a slowly decreasing function. If they manage to sell 100 000 tickets, they get 766 240 dollars. Note that the first ticket also costs 10 dollars, the same as for company A.
200 000 undecided customers hear about this.
If both movies had been 10 dollars, 100 000 would have gone to see movie A and 100 000 would have seen movie B.
However, now, thanks to B's sublinear pricing, they all decide to see movie B. B gets 1270 000 dollars, A gets nothing.
Sublinear pricing.
Many products are being sold that have substantial total production costs but very small marginal production costs, e.g. virtually all forms of digital entertainment, software, books (especially digital ones) etc.
Sellers of these products could set the product price such that the price for the (n+1)th instance of the product sold is cheaper than the price for the (n)th instance of the product sold.
They could choose a convergent series such that the total gains converge as the number of products sold grows large (e.g. price for nth item = exp(-n) + marginal costs )
They could choose a divergent series such that the total gains diverge (sublinearly) as the number of products sold grows large (e.g. price for nth item = 1/n + marginal costs )
Certainly, this reduces the total gains, but any seller who does it would outcompete sellers who don't. And yet, it doesn't seem to exist.
True, many sellers do reduce prices after a certain amount of time has passed, and the product is no longer as new or as popular as it once was, but that is a function of time passed, not of items sold.
I was unclear, of course it is real physics. By "real" I mean simply something that occurs in reality, which quantum nonlocality certainly does.
Quantum nonlocality - despite being named "nonlocality"- is actually local in a very important sense, just like the rest of physics : information never moves faster than c.
Every single physical theory that is currently considered fundamental is local, from general relativity to quantum mechanics.
I dislike the wikipedia article on the subject, it gives far to much credence to the fringe notion that maybe there is a way to exploit entanglement to get faster-than-light information transfer.
The quantum nonlocality article is much better, it correctly points out that
it (quantum nonlocality) does not allow for faster-than-light communication, and hence is compatible with special relativity.
Real physics is local. The graphs, to the extent that there are any, are embedded in metric spaces, have small upper bounds on the number of edges per vertex, are planar, .... generally there is plenty of exploitable structure beyond the pure graph-theoretical problem. This is why I do not think hardness results on abstract graph-theoretical problems will be a great obstacle for practical problems.
Recently, there has been talk of outlawing or greatly limiting encryption in Britain. Many people hypothesize that this is a deliberate attempt at shifting the overton window, in order to get a more reasonable sounding but still quite extreme law passed.
For anyone who would want to shift the overton window in the other direction, is there a position that is more extreme than "we should encrypt everything all the time" ?
What you mean is there's no way to write Pi with finitely many digits, in any basis."
pi=1 in base pi
... but that's not what you meant :)
doctors, who know jack shit about statistical and causal inference
Another is an asymmetry in the average temperatures on opposite hemispheres of the sky. This runs counter to the prediction made by the standard model that the Universe should be broadly similar in any direction we look.
Why doesn't this just mean that we are moving w.r.t. the rest frame of the CMB? The signal is redshifted in the hemisphere we're moving away from, and blueshifted in the hemisphere we're moving towards, so it would look hotter in the hemisphere we're moving towards.
The prior distribution over hypotheses is distribution over programs, which are bit strings, which are integers. The distribution must be normalizable (its sum over all hypotheses must be 1). All distributions on the integers go to 0 for large integers, which corresponds to having lower probability for longer / more complex programs. Thus, all prior distributions over hypotheses have a complexity penalty.
You could conceivably use a criterion like "pick the simplest program that is longer than 100 bits" or "pick the simplest program that starts with 101101", or things like that, but I don't think you can get rid of the complexity penalty altogether.
Solomonoff induction justifies this : optimal induction uses a prior which weights hypotheses by their simplicity.
Two is an odd prime number, because two isn't odd.
Languages : what's a syllable?
what happens if we find all these biologically feasible exoplanets that just don't have any life on them?
That would be evidence for an early filter over a late filter, so it would probably be good news.
Indeed. Truly reversing the measurement would involve also forgetting what the result of the measurement was, and Copenhagenists would claim this forgotten intermediate result does not count as a "measurement" in the sense of something that (supposedly) collapses the wave function.
The Brussels meetup is typically in English.
Make the immigrants pay higher taxes.
Immigrants to the UK pay, on average, more taxes than native Brits.
I wonder if there's any significance to the fact that there is only one (known?) material out of which a space elevator could be constructed on Earth. Most planets, I would expect, will be such that either no material is strong enough, or several materials are strong enough. Earth lies just on the boundary.
the same institutions that educate students also give them their degrees and grades.
In the Netherlands and Flanders, there is a government organisation responsible for maintaining the quality of higher education. It is true that they do not grade the students, but they do look at past exams to verify if the questions asked were sufficiently difficult and if they were graded properly. They use experts from competing universities to help them make their judgements.
The anecdote in this post, about Fermi, Rabi and Szilard considering keeping the possibility of practical nuclear fission a secret, may shed some light on the subject. He thinks that some knowledge is dangerous enough that people who know it may reasonably want to keep it secret.
(much more recently, there has been some controversy about the publication of a way of obtaining a particularily infectious strain of a certain virus, but I can't find any references for that right now)
You said
There is in fact such a thing as a learner with a sub-Turing hypothesis class. Such a learner with such a primitive output as "in the class" or "not in the class" does not engage in world optimization, that is: its actions do not, to its own knowledge, skew any probability distribution over future states of any portion of the world outside itself. ... Now, what we've been proposing as an Oracle is even less capable.
which led me to think you were talking about an oracle even less capable than a learner with a sub-Turing hypothesis class.
It would truly have no outputs whatsoever, only input and a debug view. It would, by definition, be incapable of narrowing the future of anything, even its own internal states.
If the hypotheses it considers are turing-complete, then, given enough information (and someone would give it enough information, otherwise they couldn't do anything useful with it), it could model itself, its environment, the relation between its internal states and what shows up on the debug view, and the reactions of its operators on the information they learn from that debug view. Its (internal) actions very much would, to its own knowledge, skew the probability distribution over future states of the outer world.
Physics is turing-complete, so no, a learner that did not consider turing complete hypotheses could not model the outer environment.
This thing you are proposing, an "oracle" that is incapable of modeling itself and incapable of modeling its environment (either would require turing-complete hypotheses), what could it possibly be useful for? What could it do that today's narrow AI can't?
I agree, of course, that none of the examples I gave ("primitive classifiers") are dangerous. Indeed, the "plans" they are capable of considering are too simple to pose any threat (they are, as you say, not Turing complete).
But, that doesn't seem to relevant to the argument at all. You claimed
a very general learning algorithm with some debug output, but no actual decision-theory or utility function whatsoever built in. That would be safe, since it has no capability or desire to do anything.
You claimed that a general learning algorithm without decision-theory or utility function is possible. I pointed out that all (harmless) practical learning algorithms we know of do in fact have decision theories and utility functions. What would "a learning algorithm without decision-theory or utility function, something that has no desire to do anything" even look like? Does the concept even make sense? Eliezer writes here
A string of zeroes down an output line to a motorized arm is just as much an output as any other output; there is no privileged null, there is no such thing as 'no action' among all possible outputs. To 'do nothing' is just another string of English words, that would be interpreted the same as any other English words, with latitude.