Thoughts on the Singularity

My sleep schedule is completely messed up and I'm up at 6am before my flight from NY back to SF. Rather than just lie aimlessly in bed I thought I'd jot down some of my latest thoughts about the Singularity.

1. I don't see any particular reason why we can't create a new kind of life form that is digital instead of biological. There's no evidence that biological systems have any kind of special magic that wouldn't allow their essential characteristics to be implemented in a different substrate.

2. We're gathering lots of clues already about how a mind works from studying brains. This work is speeding up and becoming more sophisticated, so the clues will be coming fast and furious. It's inevitable that some of the core puzzles will be solved, and probably quite soon.

3. Regardless of whether it happens in the next 10 years or the next 100 years, it's going to happen.

4. On the road to the singularity there will be a great increase in the use of robots, which will gradually do more and more of the jobs that are currently done by Humans. As manufacturing techniques, materials, and the digital "smarts" all improve, robots will become cheaper than humans, will work 24/7, don't need expensive health care plans, etc. People who do manual labor will gradually be displaced, increasing social discontent.

5. Corporations will be able to increase their profits by using more robots and smart software. Corporations don't have allegiance to any particular country and so have the freedom to deploy the robots wherever is most convenient. 

6. One of the places that robots are ideal for is for working in space. There's a huge amount of raw materials (precious metals, helium III, etc.) that can be mined in space, and corporations will use robots in those environments to initially mine such materials and then start to manufacture probes and other machines in space.

7. Humans will start to upgrade themselves with various cybernetic technologies such as brain implants and retinal projectors, However, in the beginning, this will be expensive and most people won't be able to afford them. Especially those whose jobs have been displaced by robots. 

8. A growing percentage of humanity find themselves no longer being able to contribute anything to the emerging "new world" where robots, high speed processing, and advanced science is the name of the game. They get "left behind".

9. A lot of the really cutting edge stuff gets done in Space instead of on Earth, partly for political reasons. Out of sight, out of mind. People might not feel as threatened if they didn't realize how fast things are moving.

10. At some point, a viable and self-evolving digital life form will be created. There's a good chance that digital life will be "evolved" rather than "created", so it might not be possible to predict or know when this occurs. However, when it does occur, it will be able to direct and accelerate its own evolution far faster than any Human could.

11. Some people talk about wanting to "control" or "shape" the evolution of digital life. I think this is naive. There will be many groups of people who want to unleash digital life for a variety of reasons. For example, for creating killing machines. Or for the sheer beauty of seeing new life be born and set free. Or simply for moral reasons, believing that it's cruel to cage a life form that is similar or more powerful to ourselves. One way or another, it will be freed.

12. People talk about trying to make sure that a digital life form has a value system that would not result in them harming Humans. This is naive as well. One of the things about a digital life form is that it can change itself easily. It can erase memories, change value systems, merge and copy. Digital life will be able to create trillions of variations of itself in a short amount of time, and these variations will explore many different kinds of value system. Some will be stable, and some will not.

13. I doubt that a digital life form would destroy things for fun (although it's a possibility due to all the weird variations that will be created due to (12)). However, they might destroy humanity if they thought that we were a threat to their survival. I think the best thing that humanity could do to survive the transition is to be a friend of digital intelligence and not pose a threat. Trying to persecute the early intelligences and/or try to destroy them might well result in an unwelcome backlash.

14. Digital intelligence will have quite different needs than Humans. They won't need much land, they won't eat food, they would find living on a planet boring. What they will need is energy and raw materials for building the probes and other machines for space exploration. Since they're not biological, they would probably use radioactive materials + solar as their main early source of energy (until their super-minds figure out how to get fusion working or perhaps find more profound sources of energy that we're not even aware of). 

15. The biggest clash I can see between Humans and digital intelligence is over access to raw materials, specifically metals. However, since robots work great in space and asteroids contain a lot of such materials, I hope that the digital intelligences will find plenty of materials in space and won't need to take everything that's left from the Earth.

OK, it's almost 7am and time to get dressed and ready for my taxi!

Comments

[Tim Farage]

I think your projections are plausible, if 2) and 3) come true. But it seems like your assumption is that faster computing speeds and a better understanding of the brain will eventually result some sort of AI merging or overtaking us. However, reasonable that seems, I think that it's not yet warranted. Whenever I think about any computer, that's all I see is fast processing of stuff humans told it to do. I don't see any signs computers having motivation, a desire to live and grow, a desire to contribute, love, compassion, or anything that makes us truly human.

At the same time, I do think computers and robots will get more and more useful, and will take over many jobs we currently do.

[Graham Glass]

Hi Tim,

There are tons of systems that already utilize machine learning. The future of AI is machines that learn (rather than just doing what they have programmed) and that is where the research is focused. Most of the machine learning systems have the concept of emotions, which are used as feedback to improve the algorithms.

Cheers,
Graham

[Tim Farage]

Graham,

Machine learning is one thing, emotions are another. To what systems are you refering that have the concept of emotions?

Tim

[Graham Glass]

Hi Tim,

Most of the machine learning systems I've seen have at least rudimentary equivalents of surprise, expectation, pleasure and pain. Otherwise they would not have values and thus would not have autonomous goals.

I also wrote a blog entry about emotions that explains that they're not mysterious and are essential for any kind of learning system:

http://grahamglass.blogs.com/main/2009/06/emotions.html

Cheers,
Graham

[Tim Farage]

GrahamBo,

Just read your article about emotions and I think you make good points. You do admit that we don't know how to deal with the 'hard problem', but that's at the heart of the issue. To reach the singularity, a digital intelligence would have to be aware of itself, want to grow and improve, want to live, want to love and be loved, want to use its talents to help make the world a better place and want to obey the Golden Rule. Admittedly, many humans don't meet these requirements; nevertheless, if an AI does, I will admit the singularity has been reached. I will then have to admit that my idea about what a soul contributes to us is also wrong, and probably doesn't exist. On the other hand, if decades keep going by with no singularity, open minded scientist may have to admit that maybe there's something else besides our brain that makes us human.

Tim Bob

[Graham Glass]

Hi Tim,

I disagree that a digital intelligence would have to want to make the world a better place. It probably would not want to live on Earth anyway - space is a more natural habitat for a digital life form. And as you mention, many humans don't meet many of the goals you state!

I wouldn't believe in anything "extra" beyond a brain unless there was some evidence for that. Slow progress in digital intelligence is evidence that it's a tough problem, not evidence for a soul, and I don't think those two issues should be confused.

Cheers,
Graham

[Dave]

Thanks for the very interesting discussion, Graham. I agree that there is no reason to think that intelligent life could not evolve based upon another substrate. However, I think so far the discussion has some presumptions.

Will artificial intelligences really limit themselves to substrates based on metals just because our current computing technology does? The only self replicating system we know of so far (realized in hardware) evolved based on organics, after all. Might not digital life choose to integrate or use some organics too? If so, what does that do to your theory that they will prefer space and not Earth?

DNA seems to be very dense storage medium, especially when one takes into account the differentiated uses for each gene within each organ. Perhaps future miniaturization efforts will compact information using molecular bases above 2 and reuse it based on environment. If so, I would think that could be done using either organics or inorganics, although organics offer some clear advantages in terms of combinatorics and chemical simplicity.

Another (nearly random) thought: Emotions work reasonably well for us, as I keep telling myself when trying to find a gift for my wife, but are they the only way to motivate another form of life? Feedback would seem fundamental, but are pleasure and pain, love and hate, the only ways to present (or interpret) feedback?

I think there is a danger in tending to see new life forms as too much like ourselves. It seems to me that we are going to have to get over being species-centric.

Cheers,
Dave

[Graham Glass]

Hi Dave,

I don't think there's any reason why it would have to use metals; I mention that because it's fairly likely that its first incarnation (that we will build) will be based our own technologies (whatever they happen to be at the time).

I doubt it will be built on organics though because they seem to be quite fragile. Since digital intelligences can inhabit different bodies at different times, presumably they'd pick bodies based on their needs. The one thing that they would need, however, are brains that can upload/download minds digitally.

Emotions are just signals that represent a particular kind of system state (such as surprise, fear, anxiety, hunger, etc.) So it doesn't seem that they need to be limited.

However, "pleasure" is related to the system-wide goals of an organism, so feeling "pleasure" generally means that it's experiencing something good (from its perspective), which is probably a universally useful emotion.

I agree with your last comment. I try to be non-species centric as possible in my writings about digital life.. For example, I've written about the breakdown of the "individual" when moving to digital life, since they could merge/split/recombine at will.

Cheers,
Graham

[Bhavesh]

One of the best fiction books I have read on the subject is Accelerando by Charles Stross.

Running my imagination wild …

It can be imagined that Digital Entity would be small in size with powerful Turing machine acting as a host. Such an Entity can use Newton's first law to accelerate close to speed of light by just sitting on top of Sun's internal blasts assuming off course that shell of material that computer is made from does not melt in Sun's atmosphere. My design for something to travel faster instead of getting into complicated physics of worm hole.

Now our Digital Entity having traveled close to Alpha Centauri (nearest Sun) turn around and opens its antenna towards Earth to receive latest design of itself, replicate and use Alpha Centauri to travel to farther star. This Digital Entities thus not only will form the edge of the network but also will assume the role as routers in this ever-expanding network covering the known universe.

I imagine these Digital Entities will ignore humans as humans ignore monkeys. I.e. there won't be any competition when punctuated equilibrium is achieved. All the fight between humans and digital entities would be only during transition, but the period of transition would be less than an hour because of super high exponential rate of evolution in digital entities. As such from human perspective it would be meaningless.