Four of us got up way too early this morning to drive down to Stanford to join 2,200 others at the Singularity Summit, a free pow-wow to discuss Ray Kurzweil’s The Singularity is Near, When Humans Transcend Biology.
Peter Thiel, founder of PayPal and FacePlace, was MC. Chess-playing computer programs appeared in the late 70s. In the 80s, people said computers would never play chess at the master level. Then it happened. In ‘96, IBM’s Deep Blue beat Garry Kasparov, a grand master and world champion. Chess, now something a machine could do, lost is lustre. (Sad for Peter who had impressed folks with his play on the Stanford chess team.)
Ray Kurzweil. Applying exponential math to change. Awesome list of awards. The Singularity: A hard or soft takeoff?
Dan Farber’s coverage | Renee Blodgett | Singularity Institute | Mike Treder | Kurzweil AI
Earlier posts at Internet Time Blog: The Singularity is Near | Ray Kurzweil | Changes Ahead
Ray begins by taking a photograph of a page of his book. The camera device converts the text to speech; a natural sounding woman’s voice reads the text to the audience. [Tumultuous applause.] The device was developed for the blind. Five years ago, software couldn’t do this.
Seven years ago, computers couldn’t tell the difference between cats and dogs. Humans excel at pattern recognition. This was Kasparov’s strength: patterns, not decision trees. Now we’ve got pattern recognition and the data to train it. Google has 3 million pictures of dogs (and more of cats.)
The Law of Accelerating Returns. It’s specific to information technology. A scientific theory that’s part of a broader theory of evolution.
The paradigm shift rate is doubling every decade. Evolution feeds on itself.
The Singularity is Near is a counter to criticism of The Age of Spiritual Machines, When Computers Exceed Human Intelligence. Critics who said Ray was not believable were thinking linearly. Software critics failed to realize that software’s complexity and performance is improving; generic algorithms are more complex, more capable (one self-organizing paradigm among many). Reliability critics said software was too brittle — but software flies airplanes; decentralized systems are solid. Criticism from the complexity of brain processing. But that amount of technology is manageable.
We are the species that goes beyond limitations. We are the only species that creates knowledge, art…
Exponential growth is soft… gradual… incremental… smooth…. Mathematically identical at each point.
Douglas Hofstadter. Trying to Muse Rationally about the Singularity Scenario. On April 1, 2000 Doug hosted a symposium: The Age of Spiritual Robots. What we will be in 2093?
Hofstadter was set up to be the foil to Ray’s optimism. Was this picky pedant really the author of Gödel, Escher, Bach: An Eternal Golden Braid? “Ray is the master of playing the lines. There’s a lot of hand-waving in Ray’s new book; he can throw you a curve.”
Is Ray’s Law of Accelerating Returns really a law? It’s a tendency, a trend. Lots of disbelief. Might not the mind be more complicated? Machines are still not writing Bach.How realistic is it? (Ray’s vision.) People throw cold water on it - but they haven’t read the book.
Hofstadter cartoon of Ray hitting the wall and Ray’s reaction.
Too much science fiction. It’s hard to take his concept that he’ll be immortal seriously. In the past book, utility fog. Engineering below the level of quarks. This sort of stuff contaminates the book. Ray confuses genome and genetic code; exponential curve does not have a knee. He reduced my level of belief but I don’t deny it’s not possible. Listening to Ray is like hearing one side of a divorce. It’s time for science to give it a serious response.
Nick Bostrom
Cory Doctorow. Singularity or Dark Age? How the copyright wars threaten technological progress
How do 3,000,000 cats get on the net? Agency.
For 500 years, scientists did not share. They were called alchemists. One day a guy decided to share his findings with others - and thus began the Enlightenment.
Agency is the difference between Frankenstein and I-robot.
DRM takes away human agency - even when we own the device. CDs can be repurposed. Changing DVDs is against the law. Compare AOL to real internet email. For a while, AOL subscribers could email only other AOL subscribers.
The executives of the media business aren’t luddites. They are technophiles. Cory has never met a group so enthusiastic, just like him, willing to pay to beta-test other people’s mistakes. They are hungry for money.
Nothing stays around forever. Vaudeville. Radio. TV. What technology gives, technology can take away. We need rules and regulations that encourage new art but copyright is yesterday’s tool. We can encourage the Law of Accelerating Returns. It’s a world where abundance is celebrated.
Nobody knew that saving the whales, eating wholesome food, and emission controls were all the same thing until the word ecology was invented. We need a word for saving agency.
Eric Drexler: Productive Nanosystems, toward a super-exponential threshold in physical technology. e-drexler.com
Productive nanosystems will treat atoms like bits. Enabling research is advancing rapidly. Components being assembled. Engineered proteins. All of this is atomically precise. Molecular machines.
Trends don’t invalidate physical limits. Entropy. Quantum measurement limits. Speed of light. Discoverings can revise limits up or down.Technology is advancing rapidly but people are only now focusing on production. Bogus criticism has fallen out of fashion, a tech roadmap project is underway, R&D agencies are seeking planning advice.
Apparently, Drexler has been a broken record on this topic for decades.
John Smart. Searching for the Big Picture: Systems Theories of Accelerating Change
Accelerating.org. John spewed information too rapidly for me to keep up.
Eliezer Yudkowsky. Singinst.org. summit notes
I feared this fellow was going to be totally academic, but he ended up being the highlight of the day. Intelligence: the most powerful force in the known universe. But ten different scientists will give you ten different answers because it’s spread across many, many disciplines.
If I am ignorant about a phenomenon, this is a fact about my state of mind, not a fact about the phenomenon.
Confusion exists in the mind, not in reality. There are mysterious questions, never mysterious answers.
Brain’s biological bottleneck. Neurons run at 100hz, no read access, no write access, no new neurons, existing code not human-readable. (Not a problem for AI.)
Lightspeed >10 ^6 times faster than axons, dendrites. Possible to build brain at least 1,000,000 times as fast as human brain.
People talk as if all AI is the same tribe, an ethnic stereotype. The design space of minds-in-general. We are only a few genetic leaps from a petunia.We have to reach into mind design space and pull out a mind such that we’re glad we created it…
AI isn’t a prediction problem, it’s an engineering problem.
“Friendly AI”… the challenge of creating an AI that cures cancer rather than wiping out humanity.
The intelligence explosion: Enough power to…make the world a better place?
Bill McKibbon: Minsky: we learn but 2 bits of data per second! The Age of Missing Information, watching 24 hours of 100 channels, concluded that the most important being in the universe is the viewer.
World is driven by fear of death. Living should be enough for us, not living forever.
The average European uses half the energy of the average American.
Are we happy? Only 25% of us make such a claim. This is a fraction of the results from the 1950s.
Ray Kurzweil. Ray laboriously rebuts the arguments advanced by some of the panelists, particularly Hostadter’s disbelieving critique.
Event
This was an odd one. The price was certainly right. (Free.) I had a fun time carpooling down to Stanford with some interesting people. Ray was right on, although I’d heard his message before.
I don’t think the speakers coordinated their presentations prior to the event. Hofstadter directly critiqued Ray’s book, but most of the panel presentations were at best tangentially related. Nick Bostrom gave a rap on catastrophic disaster which didn’t fit. Eric Drexler sold the concept that nanotech is here today; people haven’t been paying enough attention. Cory Doctorow got off some great one-liners but his major message was that you must remove the constraints (e.g. DRM) to permit innovation to flourish. Max More talked about wisdom but didn’t relate it to Ray. Bill McKibben gave a great presentation, questioning where we are headed, spirituality, and the desirability of immortality. John Smart pointed the futurist firehose at the audience, losing his message from overdosing us. Eliezer Yudkowsky talked about the choices we must make (because AI isn’t going to end up being friendly unless we goose it along.)
The speakers were arrayed in front of us on stage the entire day. What fun, eh? Cory and John pulled out their laptops and began doing their email. Hofstadter and others stared at the floor.
The final Q&A was a disaster. Neither the questions nor the replies were under control. Yack, yack, yack, off topic, yack, diatribe, yack, yack, religion, yack. Peter Thiel, who earlier had cut off Hofstadter, saying he had but 20 minutes to speak (it was actually thirty), let the event run 30 minutes over time, mostly with stuff better left unsaid.
6 comments ↓
Actually, Ray didn’t rebut Hofstadter’s concerns at all. Doug’s key points were:
1) mixing wild speculation and science fiction (foglets, timetravel) with what purports to be science or at least informed discussion about science and scientific progress does a huge disservice to the cause.
Ray didn’t address this in the slightest.
2) inaccuracies such as Ray’s use of ‘genome’ or discussion of the ‘knee of the exponential curve’ (there is no such thing) likewise do a huge disservice.
Ray didn’t address this in the slightest.
3) the comment about listening to Ray being like listening to one side of a divorce is spot on, and again it does the cause a disservice. Ray presents an extremely one-side vision that has the singularity occurring by 2029. It is obviously not as simple as Ray makes it out to be, or else there would not be such a range of opinion even among the summit attendees on if and when such will come to pass. Ray’s tone is like that of a fundamentalist preacher, who has all the answers, no questions, and inhuman confidence.
Hofstadter’s points were that the singularity is completely ignored by mainstream scientific and academic people, and this is partly explained by the above points.
Joseph, perhaps a rational interpretation falls between our two points of view. Hofstader didn’t just call Ray on a vocabulary error regarding the genome, he said this was akin to confusing an electron with a nucleus (I think those were the particles.) Technically an exponential curve doesn’t have a knee, but it’s a handy metaphor for describing what feels like a phase change as we go vertical. Ray is resolute, to be sure, but I doubt if I would understand him at all if he showed me the math and logic that got him to his conclusions. Without his simplicity of vision, I would not have had access to what he was saying. Ray is pounding a stake in the ground because he wants the same as Hofstadter: discussion and debate in mainstream communities. YMMV.
jay
[...] Related post Technorati Tags: [...]
What exactly “didn’t fit” about Nick Bostrom’s speech? I’ve found him to be quite clever in the speeches I’ve seen of him online– TED, etc.
Libertariana, I didn’t say Bostrom wasn’t clever, I said it didn’t fit with the theme of the event he was speaking at. Why? This was nearly two years ago. The misfit, if any, was insufficient to make it into permanent storage.
Joseph-
1.) First of all, ALL science was once science fiction. The fact that we don’t have a particular technology now is in no way evidence for its impossibility in the near future. Moreover, we do have now the necessary technology working in labs to realize everything Ray talks about, including quantum computers. This is a huge weakness in Hofstadter’s critique, if you can call it that. He was actually just venomously attacking Ray basically as a being stupid, arrogant and uninformed - which is a charge that you just can’t legitimately level against Ray Kurzweil, of all people in the world. He even questioned his sanity at one point during his remarks - just childish stuff - nothing of substance.
2.) These are not ‘inaccuracies’ but ways to simplify the argument so that most people can understand it. Do you REALLY think that Ray Kurzweil doesn’t know the difference between the genome and genetic code and a ‘gene’? This is ridiculous. If you read his book, you would see that he is very well aware of how genes work and what they are. About the ‘knee’ of an exponential curve - he explains in the singularity is near, that this is just a metaphor, that the term ‘knee’ is used for efficiency in the argument and that all points on an exponential plot appear linear.
3.) Ray does not predict the singularity will happen in 2029, but in 2042, which is a conservative estimate, like all of his predictions are (by design.) This is a pretty basic fact the you and Hofstadter (ironically) get wrong. Hofstadter’s comments here are at best petty and childish, and once again do not criticize ANY of the science of the argument, but rather serve to attack the character of Ray himself.
Actually, when I sat down to watch Hofstadter’s ‘critique’ of Ray’s ideas, I was really excited to see a well-thought-out, reasoned explanation of why the singularity is NOT near. I have an obsessive curiosity about these ideas, and have read hundreds of articles and scientific papers from various authors on the subject. I have yet to find a valid criticism of the ideas or any of the science and engineering behind them. The most interesting idea I could imagine would be why the singularity CANNOT happen, and I have honestly been searching for that argument pretty intensively for a couple of years now.
Unfortunately, Hofstader just attacked Ray’s character, intelligence and sanity, and basically gave an embarrassingly insufficient argument against the idea of a technological singularity. Ironically, it was Hofstader that had ‘inhuman confidence’ that is, such confidence that you do not even have to provide evidence for your position.
At first I wondered why Ray didn’t just slam the hell out of Hofstader for these comments and attacks, but then I then I realized he was just being polite. I was very impressed with Kurzweil’s professional demeanor in resisting a very easy intellectual smack-down of Hofstader on stage, which would have embarrassed him, and there was no need for that.
Leave a Comment