| Blast from the past |
[Dec. 2nd, 2009|01:06 pm] |
1988 docudrama about "the ideas of Douglas Hofstadter". It was created by Dutch director Piet Hoenderdos. Features interviews with Doug Hofstadter and Dan Dennett. Dennett also stars as himself. Original acquired from the Center for Research in Concepts and Cognition at Indiana University. Uploaded with permission from Douglas Hofstadter. |
|
|
| Who's squirming whom? |
[Nov. 27th, 2009|11:47 pm] |
One recurring meme that's been popping around over the last couple of days is the idea that climate researchers (and those who trust climate research as a science) are some how in fits of contortions squirming about trying to rationalize this sudden cognitive dissonance. It's a good meme, tells a fun story, and like many memes is generally hogwash.
I spent some time today reading a thread of comments at esr's blog:
http://esr.ibiblio.org/?p=1447
There's a lot of discussion there, but for the most part it's a predictable display of global warming skeptics being unskeptical about anything which supports their position, and backers of science asking for more information, more context, and more data. In a poetic turn of events, one line of code has become the anti global warming movement's "hockey stick". The mere existence of commented out code somehow proves once and for all that climatology is just a house of cards, and if they could only pull that bit of code out and show everyone it would topple.
It's almost embarrassing to watch. But at the same time, it's sort of hilarious watching the skeptics try over and over again to prove by assertion that this data dump is the end of climate science. In the end, I expect that they whole lot of them will convince themselves that global warming is a fraud (your classic no-op) while generally sane people will continue to monitor scientific publications and/or broad surveys of results (also a no-op).
The most disturbing aspects of this whole debate are the parallels with creationism/intelligent design. Both anti-global warming zealots and creationists have done a great deal to dress up their rhetoric in the garb of science, but have had very little success in coming up with scientific models which validate their hypotheses. I wouldn't be surprised if 150 years from now, we still have tons of anti-science crusaders railing against climate science, despite the many probable leaps and bounds the field will make in that time. There's not much that can be done when it comes to certain kinds of irrationality it seems. I understand why some people flock to anti-Darwinism, but for the life of me I have no clue what it is about global warming that gets people's undies in such a bunch.
Does anyone have any idea what it is that attracts the wing nuts to anti-global warming?
[eta: quote which illustrates esr's own wing nuttery: "Creationism is certainly politicized science..." Yes, that's right. Some software guy railing against global warming thinks creationism is a science.] |
|
|
| Bad Scientists |
[Nov. 25th, 2009|11:45 am] |
Not all scientists are perfect. Some are actually pretty flawed. Here's a list of things potential flaws to look out for when considering whether or not a scientists is of the highest caliber:
- Biblical literalism
- labeling contemporaries as frauds
- withholding data and research
- publishing anonymously
- extended investigations into pseudo-science
- a belief s/he is a modern prophet of God
- research results denounced by other scientists
If you find a scientist who suffers from many of these foibles, take care when evaluating their work. If that scientist is unfortunate enough to suffer from all of these foibles, then he's probably Sir Isaac Newton. |
|
|
| More "cat brain" fail |
[Nov. 24th, 2009|10:49 am] |
Apparently, I wasn't the only one who thought the cat brain neurons were a bit skimpy at 16 bytes:
1. These are point neurons (missing 99.999% of the brain; no branches; no detailed ion channels; the simplest possible equation you can imagine to simulate a neuron, totally trivial synapses; and using the STDP learning rule I discovered in this way is also is a joke).
...
this is just a hoax and a PR stunt.
All this whole brain simulation research is just window dressing. Until we can get a realistic simulation of a neuron, the chance of getting emergent AGI is zero. But for those who like Kool-Aid, you can buy it in the paper form for $13.00 at Amazon.
(bonus video: Junky strung out on the Kool-Aid. (He kind of reminds me of Lazlo Hollyfeld). money quotes: "He's shown a lot of graphs." "Why'd I pick ten [years to the signularity]? Frankly, I have ten fingers on my hand") |
|
|
| What science is, and what science is not |
[Nov. 23rd, 2009|04:52 pm] |
Generally speaking, science is the process understanding observation in such a way as to construct models of understanding which lead to the ability to predict outcomes. In order to be a scientist, you have to know how to ask questions, and interpret answers.
Skepticism is the questioning of ideas and relationships that others take for granted.
Both are noble and honorable ways of trying to understand the world we live in. But only one is science. Science gives us the best model with which we may understand the world. It is not a static model, and is often not the true model. But it is in general the best model we have so far. Over time, our understanding of things may deepen, and our scientific model of things may change. But it is not skepticism which drives the change, it is science. While being skeptical about things may lead you to questions that others may not ask, until you partake in answering the questions, and using the answers to make reliable predictions, you are not undertaking science.
In the natural world, the scientific method universally trumps naked skepticism.
http://www.csicop.org/specialarticles/show/deja_vu_all_over_again/ |
|
|
| Everything old is new again |
[Nov. 19th, 2009|03:16 pm] |
Truly obnoxious and patronizing advert for the "dumb terminal of tomorrow":
"Stateless, which is ... kind of a big ... word ..."
Because five letter words with a common suffix are a bit above the target demographic for this dumb terminal. |
|
|
| Singularity fail (?) |
[Nov. 18th, 2009|10:52 am] |
Ray Kurzweil: 2010, Supercomputers will have the same raw computing power as human brains
Actual Reality: 2009, part of a cat brain is simulated on an IBM supercomputer. http://www.modha.org/C2S2/2009/11182009/content/SC09_TheCatIsOutofTheBag.pdf
human brain: 10^10 neurons, 10^14 synapses
"cat brain": 10^9 "neurons", 10^13 "synapses"
Off by a factor of ten. That doesn't seem too bad until you look at how they simulate a "synapse". They are over 500x slower than a human synapse, and the "synapse" state is modeled using 16 bytes of information, somewhat less that it would take to model a physical synapse -- off by 2-6 orders of magnitude(?). Obviously, they don't spend any computational time simulating quantum effects. This is no "Schroedinger Cat". Unfortunately, they don't spend much time analytically discussing how their model of a synapse compares to actual synapse. My guess is not so well.
I may be parsing their findings wrong, but it looks like their "KEY SCIENCE RESULT" is that they built the damn thing and it didn't crash. Predictably they invoke exponential growth to conclude: "with further progress in supercomputing, realtime human-scale simulations are not only within reach, but indeed appear inevitable (Figure 8)." Figure 8 predicts human performance by 2018, which pushes Kurzweil's predictions bacy by only eight years.
It does make me wonder just how "human" a machine with 16 byte synapses will be. I'll go out on a limb and optimistically guess that it'll only take another factor of 1000 or so to get something which actually computationally equivalent to a human brain. That pushes Kurzweil's 2010 date back to about 2040. Given that he made the prediction in 2005, it's is off by a factor of 7 so. From there can extrapolate to get a crude estimate for the arival of the "Singularity" as somewhere around the year 2300. To be squishy, let's say the singularity might occur sometime between 2100-2500 AD. (Hey! That's just in time for Data from SG:TNG!) Of course, that assumes that we can maintain exponential growth till then.
On the plus side, they did find that their "cat brain" computer was very cat-like, being unresponsive to user commands and prone to long naps. |
|
|
| Everyone's favorite game! |
[Nov. 15th, 2009|05:58 pm] |
Guess the idealogue:
These are the folks who blame the CRA for the collapse of the economy; ARAs tend to be hardcore idealogues; many are rabidly partisan. All too many are deeply uninformed. They breathe cognitive dissonance they most people breathe oxygen. When confronted with facts, data, reality that challenge their ideology, they make up new facts.
Of course, if you know what ARA stands for you already know the answer. |
|
|
| The math of "The Singularity" |
[Nov. 6th, 2009|02:07 pm] |
It was suggested by ![[info]](http://l-stat.livejournal.com/img/userinfo.gif) protopost to watch the Singularity Summit talk given by Stuart Hameroff. It was a good presentation of our understanding of consciousness and where it might be found. But unlike most singulitarians Hameroff was much less optimistic about the resources required to achieve true artificial intelligence. In the talk he discusses the computational capacity of the human brain. The human brain does roughly 10^25 operations per second (bops). This gives us a rough estimate for when we might have the computational capacity for simulating the human brain.
bops/brain: 10^25
flops/bop: 10
flops/brain: 10^26
fastest supercomputer in 2009: 10^15 flops
If we assume Moore's Law does not fail (40% annual growth), our fastest supercomputer should have the computational power to simulate a human brain by the year 2085. Compare this to Kurzweil's prediction of 2010 as the year when super computers will have the processing power, and you can see why I'm more than a little dubious.
With such a distant goalpost, Hameroff's suggestion is to not simulate the brain in the computer, but rather on top of an organic substrate -- microtubules.
He also has other good advice for researchers about where to start: "I think AI people should try to simulate something like this [centrioles [very simple cells] moving towards light] before worrying about a whole brain." |
|
|
| The Singularity |
[Nov. 5th, 2009|03:11 pm] |
Recently patrissimo posted some videos from the Singularity Summit: http://patrissimo.livejournal.com/1252373.html
I knew that the "Singularity" had something to do with anti-aging and eventual immortality. Yet despite working with, around, and on AI problems for most of my life I wasn't aware that the "Singularity" was also about artificial intelligence. Specifically, the technological singularity is the point at which artificial intelligence becomes self improving.
I watched a few of the videos, and while they were entertaining, I can't say that I learned very much from them. The panels I watched were light on scientific content, and heavy on speculative ramblings. The most scientific of the bunch was Aubrey de Gray exploring hypothetical advances in systemic age rejuvenation.
Maybe it was the panel format of many of the videos, but the impression I got was that the summit was more akin to a science fiction convention than a scientific conference. On the other hand, since the notion was popularized by Vernor Vinge, maybe it's no surprise that the cultural trappings of conference are drawn more from science fiction than science.
While this formulation of "The Singularity" as it relates to "artificial intelligence" may be interesting, I doubt it will ever come to pass. There are a lot of reasons why it won't come to pass, but I'll list two reasons why I think it won't happen the way the various singularitians think it will.
(1) People confuse current rates of exponential growth with sustainable long term rates of growth.
I've discussed this before in the context of economic resources.
http://prock.livejournal.com/2006/05/15/
The mathematics behind what's going on is covered in the various models of predator-prey interaction.
http://en.wikipedia.org/wiki/Predator-prey_equations
The basic conclusion is that given finite resources, exponential growth is unsustainable, and growth must eventually halt. If the singularity depends on the advancement of technology at an exponential rate, then the chances are real good that exponential growth will peter out before we get there.
As an not at all intuitive and hand-wavy proof of this, consider the complexity class we are talking about. If we are able to advance technologically at a rate of O(e^n) then there is no effective limit on what we can achieve. On the other hand if bounded resources indicates that advancement is even optimistically polynomial, the the relative domain of things that can be achieved approaches zero as the years go on. For example, if we assume that we can model our sustainable accumulated technology as n^2 "units" over n years, then after 25 years we'd have 625 units. On the other hand, if we could accumulate exponential to e^n, we'd have 72 billion units after 25 years. Unfortunately, 625/72000000000 ~= 0.
With exponential growth, we can confidently say, "The difficult we'll do right now. The impossible will take a little while." Without it, even the difficult will remain difficult.
To better illustrate the problem, here is a more concrete example. Imagine the problem of storing all of civilizations knowledge. There are roughly 10^50 atoms on the planet. Let's assume that we currently have 1 gigabyte of information per person on the planet. Let's also assume that we'll eventually only need 10 atoms to store a bit of information. Using those parameters, we are currently using 3*10^21 atoms to store all knowledge. If total knowledge grows at the same rate as Moore's law (40% per year), we will run out of atoms to store our knowledge in less than 200 years.
(2) Most modern approaches to synthetic intelligence are not organic
The other reason why the "Singularity" may be vanishingly difficult to achieve is that currently our approach to developing artificial intelligence is anathema to the way intelligence actually works. Specifically, intelligence is not about system design. Rather it is about emergent properties which result from massive parallelization of locally acting agents. Until our energies are invested towards the actual evolution of synthetic primordial life, most likely through artificial selection, the chance that we'll create an actual synthetic intelligence is probably very small.
"We live in a universe which permits the evolution of molecular machines as intricate and subtle as we." - Carl Sagan
In the end it may wind up that developing any such "synthetic" intelligence will require an organic process parallel to the one that constructed us.
Of course, I say all this being not only a mathematical curmudgeon, but a plain old curmudgeon as well. Such is the contrarian nature of my being.
There is actually an interesting counterpoint to my dismissal of the "Singularity". Specifically, some might argue that we actually passed through the singularity thousands of years ago years. While individually we may not be immortal, our body of scientific knowledge has become larger than any individual could ever hope to fathom. It may be said that the invention of libraries represents the cultural singularity through our species passed to become an intelligence which is capable of self improving. Or one might argue that it was writing, or even speech. Which ever technology it was, we are certainly on the other side of a singularity which has allowed us develop our intelligence to the point where we have an amazingly deep understanding of the universe.
"We are a way for the cosmos to know itself." - Carl Sagan [1]
To sum, I'll return to the subject of the of "Singularity Summit" videos that I watched. The single most disheartening aspect the videos and the various responses to them was the denigration of fields like physics and mathematics. If there is any hope of escaping the finite resources to which we are bound, it will be through advances in exactly those (and related) fields.
[1] Yes I did just watch "Cosmos" on hulu.com [2]
[2] But really, I got the quotes from here: http://www.youtube.com/watch?v=XGK84Poeynk |
|
|
| Scientists loose on the streets! |
[Oct. 28th, 2009|11:00 pm] |
I the past week I've come across two quotes from scientists that involve confronting total strangers certain insights.
I know that the molecules in my body are traceable to phenomena in the cosmos. That makes me want to grab people in the street and say, have you heard this?
- Neil deGrasse Tyson
'Well-educated' reminds me of Peter Medawar's wickedly astute observation that: 'the spread of secondary and latterly tertiary education has created a large population of people, often with well-developed literary and scholarly tastes, who have been educated far beyond their capacity to undertake analytical thought.' Isn't that priceless? It is the kind of writing that makes me want to rush out into the street to share with somebody - anybody - because it is too good to keep to oneself.
- Richard Dawkins
I guess scientists are becoming a bit more confrontational (and violent?) in this modern era. |
|
|
| navigation |
| [ |
viewing |
| |
most recent entries |
] |
| [ |
go |
| |
earlier |
] |
| |
|
|
![[info]](http://l-stat.livejournal.com/img/community.gif){kind=small}
