Originally, I was supposed to talk about LID at SD Forum today, but instead we moved it to October 27, because Ray Kurzweil is in town to speak about how we will all become immortal through technology. So I’m sitting at SAP Labs in Palo Alto instead, listening, being amazed, dazzled, inspired and I don’t know what, all while attempting to blog it. Bear with me …
Steve Jurvetson is introducing him.
So here’s the talk in bullet-point form
- Software is indispensable already today, and it will increase in its capabilities.
- by 2020 we will understand the software of everything, including the software we (people) run on. Will allow us to re-program, such as by turning off genes. We understand our biology as software now.
- Nano-technology will allow us to change us — but the value will be in information.
- Software will take over the world, and eventually the universe. Need to recognize exponential growth even of the growth rate. Will make 20,000 years of progess of the 2000 speed within the 21st century.
- E.g. it does not take 100 years to map the brain (which would be the year 2000) rate of growth
- Got interested in this as an inventor. 95% of investment projects that he and Steve Jurvetson see are fine, but the timing is wrong, so they would fail. Enablers aren’t there. So he became a student of technology trends, to come up with models for where technology would go.
- Can you predict the future? Certain aspects are very much predictable. E.g. cost of a MIPS in 2012, or cost of a base pair, or resolution of a brain scanner.
- If you have a sufficiently complex system, with many interacting and random parts, you can predict system behavior quite well. General answer: double-exponential trends.
The Paradigm Shift Rate is now doubling every decade.
- Many examples for double-exponential growth. Applies to biological and technical evolution.
- Technology is a tool-base that evolves. Double-exponential comes from the simultaneous evolution of the tool and the product. [my terms]
- Spends some time explaining what linear and exponential means (I skip this it’s standard math)
- Makes a side joke on how useless the predictions about social security are, if the singularity occurs around the same time social security runs out
- Moore’s law is the fifth (!) exponential-growth law in computing (vacuum tubes etc.). It doesn’t matter whether Moore’s law runs out, a sixth exponential-growth paradigm will kick in (e.g. molecular/3-dimensional techniques).
All kinds of information technologies double their power every 12 months (right now). Moore’s law is only one example of many. But the rate of growth is growing, too.
- Functional emulation of the brain will probably be possible by 2013, maybe earlier, with the then-best supercomputer.
- IT is now 8% of GDP (software, electronics)
The biotechnology revolution: the intersection of biology with information technology
- There is no more "drug discovery", it’s now strategically planned intervention at specific stages in specific reaction phases.
- We are gaining an entire new toolkit by reprogramming biological processes.
Every form of communication is also doubling every year in price, performace
- It wasn’t surprising to see the internet coming. He had a graph on it in the mid-80’s already, because the trend was clear.
- Also, miniaturization.
- We have reverse-engineering red blood cells, and have nano-bots that can do the same thing.
- He’s talking about blood-cell-sized robots that exist today for animals. They will shortly download new algorithms from the internet.
He’s now getting into specific objections that he’s hearing against his projections about the singularity.
- Where will we get the software for running an artificial brain?
- Non-invasive brain scanning is doubling every year in capability. There is now a real-time brain scanner for live brains that can observe clusters of neurons firing.
- But brains are so complex? But it’s encoded in the genome, which is only 800 million bytes. If you eliminate redundancies, its about 30-100 million bytes. The brain is about 30-60 millions bytes.
- Example: cerebellum. Looks very complex, but is created from just a handful of rules. Analogy: Mandelbrot set.
- Models often get simpler at a higher level, not more complex.
- The bulk of human intelligence is based on pattern recognition: the quintessential example of self-organization.
- Shows a voice-to-voice automatic language translation from English to German. Very impressive.
- Most of his projects are about pattern recognition.
Now some predictions:
- 2010: Computers disappear. Lots of bullet items what this means, e.g. electronics so tiny it’s embedded in the environment, our clothing, our eyeglasses.
- 2029: 1000 dollars of computation will be 1000 times the human brain. (Wow)
- Nano-bots. Neural implants. Full-immersion virtual reality incorporating all of the senses. Expansino of human intelligence.
- in 2050, the predominant form of intelligence will be non-biological.
- Life expectancy will explode.
I was somewhat disappointed by his answer to my question in the Q&A section: "If the rate of change increases, the stresses in our global society are also increasing. What can we do to alleviate them?" His answer was a pure technologist’s answer: "… because all of this wonderful new technology, we will solve them.". The trouble as I see it — see current international security events — is between the Lexus and the Olive Tree, not between the Lexus and the Chevrolet: