The solution to everything

Part I: The Solution

A highly theoretical, yet pefectly reasonable way to solve the problem of uploading one's mind in a way that it maximizes resemblance, allowing full connection between host and clone is:

Create a superhuman AGI. By definition it will be orders of magnitude more intelligent than us. Let it figure out the solution to the problem of mind uploading. Let the magician do the magic.

As I have been exploring the possibilities of mind uploading in the past weeks, I found more and more fundamental hurdles: Silicon vs Carbon

To say that we just have to invent Superhuman Artificial General Intelligence (SAGI) to solve this problem is a pretty low intelletual achievement. By definition, SAGI will solve every problem by that token.

So I will spend the second half of this essay exploring what it takes to create SAGI:


Part II: How to make SAGI


Route 1. Build a system that is a good enough at changing its own code to maximize performance in specific tasks that it will eventually become superhuman in every aspect of human intelligence.
To put it simply: Create a good virtual developer: DAVE. (Developer as virtual engineer)

This virtual developer will initially have human-like or slightly sub-human abilities. As it will run on multiple threads on ultrafast supercomputers, there is reason to think that it will improve its skills very fast. 

At first, Dave's task will be modifying itself in order reach certain goals that average developers are trying to reach. For example, in growing order of complexity approximately: creating scripts, webpages, graphical software, neural networks for image and voice recognition, kernel, operational systems.

In order to reach these goals, Dave will modify itself and iterate an ever better developer. As it becomes better at achieving the simpler goals, he is going to be better suited to solve the next problem.
Obviously, there must be several problems that Dave will have to train on, like an actual developer. Every problem needs a different mindset, a novel approach and Dave will extend his inventory of creativity by solving every problem, ever honing in on the approximately human-level developer, which will mean absolute breakthrough!

Once Dave is as good a developer as I am, it will have acquired the necessary ingredients to turn himself into the best developer ever to exist. He will need the following attributes:
  • Logical thinking
  • Problem solving
  • Task management
  • Learning ability
  • Creativity
  • Technical skills = many different languages
  • Enjoy debugging = fun at solving puzzles
At least these are the attributes that make for a good human programmer.


Route 2. Consider all individual faculties of a Superhuman Artificial General Intelligence (SAGI). My thesis is that the sum of them will result in SAGI itself.
Some brainstorming and online search rendered the following results:


1. AGI is a system that has the following
'Mental' (also physical?) Faculties:
  • image recognition
  • sound recognition
  • sound generation
  • language
  • solving puzzles (strategy, evaluating progress)
  • memory
  • using tools
  • abstract thinking (maths,
  • curiosity = learning (updating its own code?)
  • logical thinking
  • (illogical thinking) = random noise = creativity?
  • delay gratification
  • prioritize
  • emotions ??? necessary? (and their physical manifestations?)
  • consciousness = center of attention / kernel /master
  • emulate (run a simulation of a conversation/action rather than doing it)
  • mental models (to fill the unknowns in emulations (see above), like how people will respond)
Notice that I omitted every physical manifestation of intelligence = SAGI won't need a body, just a monitor and a loudspeaker.
Notice also that I have no idea what I am talking about, there are ? everywhere.

According to Wikipedia, these are the necessary conditins of an AGI:
  • reason, use strategy, solve puzzles, and make judgments under uncertainty;
  • represent knowledge, including commonsense knowledge;
  • plan;
  • learn;
  • communicate in natural language;
  • and integrate all these skills towards common goals.





Comments

Post a Comment

Popular posts from this blog

Machine at birth?

In light of my last article entitled 'The Great Somatic Divide: Silicon vs. Carbon' , I started to think of ways in which those feelings and emotions connected to our bodies could be made available to Ainhoa (AI). Let's explore how to bridge the somatic divide: how to give AI body, so that it will be more human? My first thought was that maybe a digestive system could be added. One taken out of a pig, for exemple, and held alive inside Ainhoa. That way, she could actually digest food and get nutrients out of it. But how would those nutrients interact with her cognitive abilities? She would need some kind of brain that would be fed those nutrients. So an entire circulatory system would be needed, with a brain, liver and kidneys and a lung, so eventually you might as well add an entire body. But what body? Would a rat be good enough? Or should you use a monkey? Or should Ainhoa have a human body? Let's explore the possibilities. Ainhoa has a rat body hosted ins...
Read more

Consciousness is learning

My cousin was solving a Rubik's cube the other day. I was amazed how fast one can learn it. But what most struck me is the difference between the processes that he was doing consciously and subconsciously. In this article I will argue that consciousness is learning. After just one week of messing around with the cube, he already learned a bunch of moves that became automatic. He make one side be filled with the same color pretty fast without paying much attention. It seemed as if it were automatic, learned, he could talk while doing it. By repeating a certain move enough times, he learned it, which means that he can now do it without consciously thinking of it. He can now think about higher order moves consciously. With time, those moves will be learned too and he will no longer need to consciously think through the sequence of moves. You don't walk consciously. There was a time, though, when you did think of every move, about when you were 1 year old. If...
Read more