Comment Re:Potential Breakthroughs in AI (Score 1) 67
I forgot to mention...what I describe above is called learning by induction. Which is a form of probabilistic learning.
I forgot to mention...what I describe above is called learning by induction. Which is a form of probabilistic learning.
Our brains are based upon probabilistic computing? Not doubting it since I don't know enough of the "real" neuron details. Just noting it as ironic given how bad we are at assessing risk.
Perhaps I am not using the correct terminology. And I am not an expert...this isn't really my field. But I have seen neural network simulations which, for example recognize handwriting. You show the program many variations on the letter F, and the computer then learns to take symbols that look like the letter F and put them into that group. One can imagine that the brain is made up of many such processes. It certainly sounds like the way that I learned. See a spoon, see mommy using it to feed me, see mommy using it to feed herself. They are all different spoons, but they look similar and they seem to be used to move food from a bowl to a mouth. In a way it isn't that different than taking similar shapes and calling them a particular hand-written letter. It's just that we do this categorization at a myriad of levels on a myriad of objects and eventually of ideas themselves.
I suppose what I think might happen is that instead of building simulations of neural networks, that we will build actual neural networks using some form of electronic system. I don't know what that will look like exactly, but I suspect it is quite possible. After all, nature has already done it.
What is the point of having a nuclear power station for 5 billion dollars emulating a brain if you can make one for something like 200k by inseminating a real WIFE ? And then schooling the offspring etc.
If we can make a human level brain, then we can make something much smarter as well.
For starters, look up how many neurons you can simulate with a single CPU. Then, calculate how many fibers you need for the comms bandwidth.
There is the problem...I think that CPU's are the wrong tool. The problem needs to be re-framed. Perhaps we need synthetic neurons. Then we will need precisely one synthetic neuron to simulate a real one. I know...easier said than done. But we do know the basics of neuron function. Why can't we make a single electronic element that does something very similar?
Neurons can typically fire at a rate of 250Hz. There are about 100 billion neurons in a typical human brain. These neurons are networked in an extremely complicated and changeable parallel network. This network of neurons can be powered for reasonably long time with the energy contained in a bowl of oatmeal. Surely we will at some point be able to create a similar device, and one that doesn't require most of the world's computing power to run.
I suspect the breakthrough will come with a new computing paradigm, one that is based on massive parallelism. Perhaps it will consist of a silicon based device that mimics the network and function of neurons. I suspect it will be based on probabilistic computing, similar to how our own brains work. It will be taught rather than programmed. Perhaps there will be more states than merely 0 and 1.
I think that this is coming, because our brains are already doing it. And with incredible efficiency. Once we saw birds fly, and so we tried to do it ourselves. Eventually we figured it out. I think it will be the same with AI. We will copy nature, learn its principles, and then we will create our own version. And in doing so, I suspect our ideas of what intelligence is will fundamentally change.
There is a BBC documentary film maker named Adam Curtis who makes some fascinating and disturbing videos about society and control. He has access to the BBC film archives, and uses historical footage extensively. The assertions made are extensively documented and the interviews of powerful people are extremely interesting. I think that this video, The Engineering of Consent, is relevant to this discussion. It is one hour, and quite "stream of consciousness", but worth watching. It is the second episode in a series called The Century of the Self.
From the wikipedia summary:
"This series is about how those in power have used Freud's theories to try and control the dangerous crowd in an age of mass democracy." —Adam Curtis' introduction to the first episode.
Sigmund Freud, the founder of psychoanalysis, changed the perception of the human mind and its workings. The series describes the propaganda that Western governments and corporations have utilized stemming from Freud's theories.
Freud himself and his nephew Edward Bernays, who was the first to use psychological techniques in public relations, are discussed. Freud's daughter Anna Freud, a pioneer of child psychology, is mentioned in the second part, as is one of the main opponents of Freud's theories, Wilhelm Reich, in the third part.
Along these general themes, The Century of the Self asks deeper questions about the roots and methods of modern consumerism, representative democracy, commodification and its implications. It also questions the modern way we see ourselves, the attitudes to fashion and superficiality.
The business and political world uses psychological techniques to read, create and fulfill the desires of the public, to make their products or speeches as pleasing as possible to consumers and citizens. Curtis raises the question of the intentions and roots of this fact. Where once the political process was about engaging people's rational, conscious minds, as well as facilitating their needs as a society, the documentary shows how by employing the tactics of psychoanalysis, politicians appeal to irrational, primitive impulses that have little apparent bearing on issues outside of the narrow self-interest of a consumer population.
A scanner the size of an iPhone that you could hold up to a person’s chest and see a vivid, moving, 3-D image of what’s inside is being developed by entrepreneur Jonathan Rothberg.
Rothberg says he has raised $100 million to create a medical imaging device that’s nearly “as cheap as a stethoscope” and will “make doctors 100 times as effective.” The technology, which according to patent documents relies on a new kind of ultrasound chip, could eventually lead to new ways to destroy cancer cells with heat, or deliver information to brain cells.
I have been reading lately that there has been serious concern about the liquid-solid hybrid engine used in Spaceship 2. In 2007 there was a nitrous oxide explosion that killed three people.
On 26 July 2007, during the early rocket testing phase, an explosion occurred during a propellant flow test at the Mojave Air and Space Port. The test included filling the oxidizer tank with 4,500 kg (10,000 pounds) of nitrous oxide, followed by a 15-second cold flow injector test. Although the tests did not ignite the gas, three employees were killed and three injured, two critically and one seriously, by flying shrapnel.[12]
parabolicarc.com is a pretty good source of information on New Space. Here are some tweets:
Parabolicarc.com @spacecom Oct 31
I had deep concerns over both the new plastic/nitrous oxide engine and so did other sources familiar with the testing. #SpaceShipTwo
The concerns were three fold. One, that it wasn't being tested sufficiently on the ground before it was flown. #SpaceShipTwo
Second: that modifications required to ship to accommodate the new engine introduced additional complexity and failure modes. #SpaceShipTwo
Third, handful of test flights they were doing with new #SpaceShipTwo engine before putting Richard and Sam Branson aboard were insufficient
Let me stress2 things: one, we don't know what happened yet, so I'm not making a snap judgment about what caused the accident #SpaceShipTwo
Second, these concerns about the new engine were not mine alone. Folks much smarter and knowledgble than me were worried. #SpaceShipTwo
I predict you will be hearing a lot more about these concerns and the problems they were having in the days and weeks ahead #SpaceShipTwo
Scaled's Kevin Mickey called the engine change "a minor nuance". He rushed out of the press conference once it was over.
Mickey claimed "minor nuance" engine change thoroughly tested on ground. Tried to ask for details, but presser ended quickly & Mickey left.
Ken Brown was taking photos through the entire incident. Tracked one large piece of debris down to the lakebed. #SpaceShipTwo
I heard Ken say, "They're in trouble." And then "They're tumbling. Ken's pictures will be very crucial to understanding it. #SpaceShipTwo
Just talked to Ken Brown. Pictures show Engine fired fine, then there's a white plume. He thinks the nitrous oxide tank blew. #SpaceShipTwo
I am a huge supporter of commercial space, most especially Space X. But I think that accidents like this give the whole sector a bad name. Virgin Galactic/SC have been building this ship for 10 years, and they still don't have a viable engine. That is not a good sign.
Contrast that with Space X, which in about the same amount of time has built the Falcon 9, which has a string of 13 straight successes (touch wood). It seems to me that not all space companies are created equal.
The engines were probably fine, chances are it's something to do with the fuel delivery system that caused the explosion.
The Russians built good engines, and they built a lot of them because they anticipated needing many to get to the moon. Rather than a few big engines like the Saturn V they were using 30 odd smaller ones to do the same job. They had some problems but eventually got it working, just because they abandoned the idea of manned moon missions.
My guess is that it was a turbo-pump that failed. Those are a part of this Soviet engine, and were also likely made in the 1960's.
Elon Musk called it two years ago in this interview.
Musk: The results are pretty crazy. One of our competitors, Orbital Sciences, has a contract to resupply the International Space Station, and their rocket honestly sounds like the punch line to a joke. It uses Russian rocket engines that were made in the ’60s. I don’t mean their design is from the ’60s—I mean they start with engines that were literally made in the ’60s and, like, packed away in Siberia somewhere.
NASA spent many years developing processes and vehicles that, initially, had a very high failure rate. Today we are conditioned to expect that these launches will go off safely, but do we need to give the private sector 20 years or so to sort itself out before we start allowing people to fly in their rockets?
Almost all the US space program is ALREADY private. The dominant US launch platform is United Launch Alliance, which was created by a merger of Lockheed Martin and Boeing when the US tried to bring in competitive bidding for Air Force launches. Their merger allowed them to become the sole bidder on most launches, allowing them to continue to receive their cushy "cost plus" bids. These "cost plus" bids allow them to pretty much charge whatever they want for costs, and then receive on top a guaranteed profit margin. There is no incentive for efficiency. In fact, ULA has almost no international customers because their costs are more than double their competitor's prices. They grow fat at the teet of the US Air Force.
Space X is a relative newcomer, but since its first generation rocket (Falcon 1) had three failures, the last of which was caused by a timing error in stage separation, it has had a nearly perfect launch record. The Falcon 9 rocket has never had a significant failure (ie. one that prevented orbit of the primary payload). Touch wood. Their rockets have been designed in house with the eventual goal of re-use. Their rocket engines are designed to be simple and failure resistant. And their rockets have nine engines, so that even if one fails the flight can continue. I wouldn't write off competition yet.
I seem to remember a SpaceX rocket accidentally self destructing recently. And people have died serving NASA missions. Tell me who has this illusive awesome track record you are looking for?
It was a test vehicle that self-destructed, and not a full rocket. The test vehicle did not have all of the redundancies of the production rocket. A single sensor failure stopped the rocket. In the real rocket, there are multiple redundant sensors from which the control systems constantly poll...if one goes, the computer ignores it in favor of the other working sensors.
When I left him, I reasoned thus with myself: I am wiser than this man, for neither of us appears to know anything great and good; but he fancies he knows something, although he knows nothing; whereas I, as I do not know anything, so I do not fancy I do. In this trifling particular, then, I appear to be wiser than he, because I do not fancy I know what I do not know.
From Socrates, The Apology (399 B.C. or so)
Everything old is new again.
The problem here is that Prof. Dunning's principle could apply to anybody, including college professors. So how does he know he is correct? --- "Science is the belief in the ignorance of experts." -- Richard Feynman
Are you sure about that?
Hadn't heard anything of this before today, I'll have to look for videos. I guess he didn't want to draw any attention.
There was a video on the NYTimes site, but it was really short. And I just realized that the New York Times has one of the most irritating video players I have ever used.
Understanding is always the understanding of a smaller problem in relation to a bigger problem. -- P.D. Ouspensky
