Yuxi Liu is a PhD student in Computer Science at the Berkeley Artificial Intelligence Research Lab, researching on the scaling laws of large neural networks.
Personal website: https://yuxi-liu-wired.github.io/
Yuxi Liu is a PhD student in Computer Science at the Berkeley Artificial Intelligence Research Lab, researching on the scaling laws of large neural networks.
Personal website: https://yuxi-liu-wired.github.io/
Okay I posted the whole thing here now.
Relevant quotes:
Original text is from Discourse on Heaven of Xunzi:
雩而雨,何也?曰:無佗也,猶不雩而雨也。日月食而救之,天旱而雩,卜筮然後決大事,非以為得求也,以文之也。故君子以為文,而百姓以為神。以為文則吉,以為神則凶也。
The Britannica says:
Another celebrated essay is “A Discussion of Heaven,” in which he attacks superstitious and supernatural beliefs. One of the work’s main themes is that unusual natural phenomena (eclipses, etc.) are no less natural for their irregularity—hence are not evil omens—and therefore men should not be concerned at their occurrence. Xunzi’s denial of supernaturalism led him into a sophisticated interpretation of popular religious observances and superstitions. He asserted that these were merely poetic fictions, useful for the common people because they provided an orderly outlet for human emotions, but not to be taken as true by educated men. There Xunzi inaugurated a rationalistic trend in Confucianism that has been congenial to scientific thinking.
Heaven never intercedes directly in human affairs, but human affairs are certain to succeed or fail according to a timeless pattern that Heaven determined before human beings existed...
Thus rituals are not merely received practices or convenient social institutions; they are practicable forms in which the sages aimed to encapsulate the fundamental patterns of the universe. No human being, not even a sage, can know Heaven, but we can know Heaven’s Way, which is the surest path to a flourishing and blessed life. Because human beings have limited knowledge and abilities, it is difficult for us to attain this deep understanding, and therefore the sages handed down the rituals to help us follow in their footsteps.
No. Any decider will be unfair in some way, whether it knows anything about history at all. The decider can be a coin flipper and it would still be biased. One can say that the unfairness is baked into the reality of base-rate difference.
The only way to fix this is not fixing the decider, but to just somehow make the base-rate difference disappear, or to compromise on the definition of fairness so that it’s not so stringent, and satisfiable.
And in common language and common discussion of algorithmic bias, “bias” is decidedly NOT merely a statistical definition. It always contains a moral judgment: violation of a fairness requirement. To say that a decider is biased is to say that the statistical pattern of its decision violates a fairness requirement.
The key message is that, by the common language definition, “bias” is unavoidable. No amount of trying to fix the decider will make it fair. Blinding it to the history will do nothing. The unfairness is in the base rate, and in the definition of fairness.
Concretely speaking, are you to suggest that a 2-layered fully connected network trained by backpropagation, with ~100 neurons in each layer (thus ~20000 weights), would have been uneconomical even in the 1960s, even if they had backprop?
I am asking this because the great successes in 1990s connectionism, including LeNet digit recognition, NETtalk, and the TD-gammon, all were on that order of magnitude. They seem within reach for the 1960s.
Concretely speaking, TD-gammon cost about 2e13 FLOPs to train, and in 1970, 1 million FLOP/sec cost 1 USD, so with 10000 USD of hardware, it would take about 1 day to train.
And interesting that you mentioned magnetic cores. The MINOS II machine built in 1962 by the Stanford Research Institute group had precisely a grid of magnetic core memory. Can’t they have scaled it up and built some extra circuitry to allow backpropagation?
Corroborating the calculation, according to some 1960s literature, magnetic core logic could go up to 10 kHz. So if we have ~1e4 weights updated 1e4 times a second, that would be 1e8 FLOP/sec right there. TD-gammon would take ~1e5 seconds ~ 1 day, the same OOM as the previous calculation.
I cannot see anything that is particularly innovative in the paper, though I’m not an expert on this.
Maybe ask people working on poker AI, like Sandholm, directly. Perhaps something like many details of the particular program (and the paper is full of these details) must be assembled in order for this to work cheaply enough to be trained.
Thanks. I had hoped it to be informative and entertaining. Think of it as “let’s play” but for nerds.
Brief note: the “analysis by synthesis” idea is called “vision as inverse graphic” in computer graphics research.
I fixed the submission as required.
Also I changed the submission 3 significantly.
Stars follow the laws of thermodynamics. This observation is more predictive than you make it out to be, once it is quantified.
The theory of thermodynamics of life is more than just a statement that life is constrained by thermodynamics in the boring sense. I’m especially interested in this statement:
In short, ecosystems develop in ways which systematically increase their ability to degrade the incoming solar energy.
If this is true, then it can be used to predict what kinds of future life would be like. It would not be any kind of life, but life that can capture more solar energy and convert it into low-temperature heat at a faster rate.
Unfortunately my thermodynamics is not good enough to actually read the papers.
The talk of “purpose” seems to cause great confusion. I don’t mean it for any value judgment (I generally avoid value judgments and use it as a last resort). “Purpose” is just a metaphor, just like talk of the “purpose of evolution”. It helps me understand and predict.
I thought it was clear even to them that “wasting” energy meant using up usable energy into useless forms.
It is not just sophistry. If it turns out to be the fundamental feature of life (like how the laws of thermodynamics are for heat machines), then it would be predictive of the future activities of life. In particular, the aestivation hypothesis would be seriously threatened.
This is analogous to prediction that population would always go Malthulsian except in non-equilibrium situations. It’s not a value/moral judgment, but an attempt to find general laws of life that can be used to predict the future.
Wait, 30-50 years of good life followed by 15 years of less-good, then death. That’s “negligent”?
It’s not comparing HD-life and no-life, but comparing HD-life and non-HD-life. I think it’s obvious that HD-life is greatly worse than non-HD-life (otherwise HD wouldn’t be considered a serious disease).
You might still disagree, and that gets us into the nonidentity problem.
In an intelligence community context, the American spy satellites like the KH program achieved astonishing things in photography, physics, and rocketry—things like handling ultra-high-resolution photography in space (with its unique problems like disposing of hundreds of gallons of water in space) or scooping up landing satellites in helicopters were just the start. (I was skimming a book the other day which included some hilarious anecdotes—like American spies would go take tourist photos of themselves in places like Red Square just to assist trigonometry for photo analysis.) American presidents obsessed over the daily spy satellite reports, and this helped ensure that the spy satellite footage was worth obsessing over. (Amateurs fear the CIA, but pros fear NRO.)
What is that book with the fun anecdotes?
I was thinking of porting it full-scale here. It is in R-markdown format. But all the citations would be quite difficult to port. They look like [@something2000].
Does LessWrong allow convenient citations?
I’m following common speech where “biased” means “statistically immoral, because it violates some fairness requirement”.
I showed that with base rate difference, it’s impossible to satisfy three fairness requirements. The decider (machine or not) can completely ignore history. It could be a coin-flipper. As long as the decider is imperfect, it would still be unfair in one of the fairness requirements.
And if the base rates are not due to historical circumstances, this impossibility still stands.
Yes, (Kleinberg et al, 2016)… Do not read it. Really, don’t. The derivation is extremely clumsy (and my professor said so too).
The proof has been considerably simplified in subsequent works. Look around for papers that cite that paper should give a published paper that does the simplification...
I will review more posthumanism, things like Dark Ecology, Object-Oriented Ontology, and such.
[According to dark ecology,] we must obliterate the false opposition between nature and people… the idea of nature itself is a damaging construct, and that humans (with their microbial biomass) are always already nonhuman.
Object-Oriented Ontology… rejects the privileging of human existence over the existence of nonhuman objects
Somewhat independently of transhumanism, posthumanism developed in a more philosophical and less scientific style in the liberal arts department, with different viewpoints, often ambiguous and not at all happy.
posthumanism… refers to the contemporary critical turn away from Enlightenment humanism… a rejection of familiar Cartesian dualisms such as self/other, human/animal, and mind/body...
My personal thought is that the future is weird, beyond happy or sad, good or evil. Transhumanism is too intent on good and evil, and from what I’ve read so far, posthumanism uses less human value judgments. As such, posthuman thoughts would be essential for an accurate prediction of the future.
Solaris (1972) seems like a good story illustration of posthumanism.
The peculiarity of those phenomena seems to suggest that we observe a kind of rational activity, but the meaning of this seemingly rational activity of the Solarian Ocean is beyond the reach of human beings.
Dying at age −0.75 counts as nothing, as a little, or counts as a lot of a person, depending on what counts as a person and how much a person at various stages matter. If it counts as a lot of a person, then it leads to an anti-abortion stance, and some pro-abortion arguments might apply in this situation.
And an alternative to abortion is adoption. A person that is highly unlikely to have HD could even be produced on demand by surrogacy or IVF, instead of being taken from the pool of people already up to adoption, so that it is a “net gain”.
If the women would not even consider abortion or surrogacy as better alternatives than giving a high-risk natural birth, I consider that unreasonable and grossly negligent.
I use a fairly basic Quarto template for website. The code for the entire site is on github.
The source code is actually right there in the post. Click the button Code
, then click View Source
.
https://yuxi-liu-wired.github.io/blog/posts/perceptron-controversy/
In David Rodin’s Posthuman Life, a book that is otherwise very obtuse and obscurely metaphysical, there is an interesting argument for making posthumans before we know what they might be (indeed, he rejected the precautionary principle on the making of posthumans):
CLAIM. We have an obligation to make posthumans, or not prevent their appearance.
PROOF.
Principle of accounting: we have an obligation to understand posthumans
Speculative posthumanism: there could be radical posthumans
Radical posthumans are impossible to understand unless we actually meet them
We can only meet radical posthumans if we make them (intentionally or accidentally).
This creates an ethical paradox, the posthuman impasse.
we are unable to evaluate any posthuman condition. Since posthumans could result from some iteration of our current technical activity, we have an interest in understanding what they might be like. If so, we have an interest in making or becoming posthumans.
to plan for the future evolution of humans, we should evaluate what posthumans are like, which kinds are good, which kinds are bad, before we make them.
most kinds of posthumans can only be evaluated after they appear.
completely giving up on making posthumans would lock humanity at the current level, which means we give up on great goods for fear of great bads. This is objectionable by arguments similar to those employed by transhumanists.
I am very doubtful of this. Humans are hardwired to think in cause-and-effect terms, and Confucianism does not explicitly deny causality.
In very early China (about 500 BC), there was a period of great intellectual diversity before Confucianism dominated. There was a School of Names which is very interested in logic and rhetorics. Philosophers in that school have been traditionally disparaged, which seems to explain why formal logic has not developed in China. For example, the founder, Deng Xi, ’s fate was used as a cautionary tale against sophistry.
This has been demonstrated to persist even in modern times, by psychology studies. A reference is (Nisbett, 2003).
My guess is that in any large population of humans (~1 million), there are enough talented individuals to generate the basic scientific ideas in a few generations. The problem is to have a stable social structure that sustains these thinkers and filters out wrong ideas.
They were also lucky that they got remembered. If their work didn’t get copied that much by the Arabs, we would be praising the medieval Arabs instead of ancient Greeks.
Also, a personal perspective: in current Chinese education (which I took until high school), Chinese achievements in science and math are noted whenever possible (for example, Horner’s Method is called Qin Jiushao Method), but there was no overarching scheme or intellectual tradition. They were like accidental discoveries, not results of sustained inquiries. Shen Kuo stands out as a lone scientist in a sea of literature writers.
Confucian classics, which I was forced to recite, is suffocatingly free from scientific concern. It’s not anti-science, rather, uninterested in science. For example: