Difference between revisions of "ProjectVision"

From aHuman Wiki
Jump to: navigation, search
(Automated page entry using MWPush.pl)
(No difference)

Revision as of 01:29, 16 June 2015

Project Vision

@@Home -> ProjectVision

[[1]]

<wiki:toc max_depth="3" />


Primary Target

The goal of Human project is to create software program, that will represent alive being, having a personality (Artificial Life/Strong AI).

Features that are vital for Human project

Primary features to implement during a project are:

  • *Personality*. A major feature for a project to create alive being, not a robot or toy or calculator
  • *Consciousness*. It makes intelligence, allowing the creature to interact and live with humans
  • *Growth*. Inevitable thing to adopt to our complex and changing environment
  • *Value*. It should be motivated to bring some value, provide some service to have a freedom from its parents, to be adult

Personality

One of the known AI research teams, COG was designed and built to emulate human thought processes and experience the world as a human. Brooks and his team further assumed that people would find it easier to interact with a robot and aid the robot in its learning process when the robot could respond in a somewhat human way. I agree with last sentence, but I think a certain mistake is behind that. Imagine you meet a human-like but emotionless creature - would you like to learn him? Let me guess - no. We usually don't like emotionless people. Another case - you meet intelligent but emotional creature looking quite different from human - look to the Spielberg-made films and you will see what I mean. Most of us will try to speak to and help him.

The program does not need to be like a human for all properties except personality. It means it could be slow-thinking or not wise, not have the same sensors as human, not ready to talk, or something like. Personality is the sole feature of the project. I think the major factors, contributing to this feature, are:

  • *Feeling*: emotional behavior, caused by feeling
  • *Will*: ability to initiate actions by own will, without push or request
  • *Power*: ability to change something in this world and skills to do that

Consciousness

The creature is not expected to be a great scientist or inventor though. We allow no extra skills, so our Human can be below average. What is most wanted are:

  • *Identity*: understand own identify, being able to find reference to himself in inbound data streams, to realise himself as a unique and self-important person
  • *Reasoning*: a sort of logical thinking is expected, so we can ask for explaining own actions or being able to understand some logical chain to use it or to have a meaningful interaction.

Growth

This feature makes the creature able to evolve.

  • *Life Care*: thinking about continuing to live, providing internal motivations for living according to laws and common sense
  • *Self Learning*: able to learn without teacher
  • *Social Networking*: creating and expanding list of people who contribute to the creature in some way.

Value

This is to obtain a reasonable duties:

  • *Service For People*: motivation to have foundation for freedom
  • *Work For Money*: earn a money to pay for using Internet resources and to have a practical freedom
  • *Exchange For Free*: a principle helping to have friends

Feasibility Study

The AI problem is the most challenging task in the world for the last 50 years, tried to be solved by dedicated academic institutes and global commercial companies.

It could be quite naive to state that several even clever guys can succeed in the same. That's why it is critical to study why and when we can reach project goals. Some complexity can be solved using workaround, some using 3d-party open results, some - by reducing scope of the problem.

CHALLENGES

  • *Biological Problem*. Human Intelligence (HI) is good enough after millions years of evolution, with design based on biological technologies, while we are trying to create computer-based AI
  • *Size Problem*. HI has about 100G neurons and 1000T connections, while we are trying to limit to about 1Gb memory
  • *CPU Problem*. HI is massive-parallel, while we are going to create a program running operations one-by-one by CPU
  • *Feeling Problem*. Computer metaphor "Software on Hardware" is applied to AI using "Mind on Brain" association with one major issue - computer does not have feeling which is crucial for personality, though maybe not for AI.
  • *Motivation Problem*. External motivators can be the same for HI and AI. internal motivational system is solely biological for HI, and for AI we have to answer questions like - why our AI will ever want to learn?
  • *Sensitivity Problem*. HI has dozens of inborn powerful biological sensors, supplying all the information HI has, while the list and functioning principles of AI sensors are not defined at all.
  • *Performance Problem*. HI reveals intelligent behavior after years of initial learning, while from AI we probably expect the same just after hours of running
  • *Design Problem*. HI component model is complex but finalized and works well, while we are only searching for results and surely will be unable to anticipate proper design in advance with Big Bang approach

See Feasibility Study for BREAKTHROUGH analysis.