Conservation of Intelligence (and Corollaries for future of AI and AI at Work)

Source: Deep Learning on Medium

Conservation of Intelligence: A Principle of Artificial Intelligence (and Corollaries for AI at Work)

There is no free lunch. This is true whether the free lunch is at home (paid by us from household income), public school (paid by us from tax income), bar (paid by beverage revenue), or elsewhere. There is also no free intelligence, whether the intelligence is human intelligence or natural intelligence in general or a derivation of it like artificial intelligence.

We define intelligence as the expressed ability to map a time-series of permutations and combinations of input elements to a time-series of permutations and combinations of output elements. The ontology of the elements includes:

  1. intelligence
  2. matter
  3. sound,
  4. touch,
  5. light,
  6. taste,
  7. smell,
  8. auditory computation,
  9. somatosensory computation,
  10. visual computation,
  11. gustatory computation,
  12. olfactory computation,
  13. reproductive computation,
  14. articulation,
  15. locomotion,
  16. manipulation, and
  17. energy transformation.

The above definition of intelligence is capable of being expressed to the fullest (or most number of mappings) by humans. Other natural organisms like plants and animals also express some mappings. For example, plants input light and output energy transformation (glucose or fruits in the form of photo synthesis). We define natural intelligence as the intelligence exhibited by natural organisms. We define artificial intelligence as the intelligence or the ability to express the mapping exhibited by other matter. For example, Alexa is able to take as input its own previous intelligence (specifically, existing models for trigger word detection, user detection, and user personalization) and sound (specifically, the user speech), and able to map them to an auditory computation (specifically, Alexa’s reply).

The no free lunch principle is related to the Commoner’s four laws in ecology. In Finance, the no free lunch principle becomes the no arbitrage or more broadly the . Even more broadly, the no free lunch principle is equivalent to the laws of conversations in Physics — X cannot be created and cannot be destroyed (X = charge, momentum, energy/mass, angular momentum, baryon number, and lepton number.) With the above definition of intelligence, we postulate the law of conservation of intelligence: intelligence cannot be created and cannot be destroyed; it can only be exchanged from one natural or artificial matter to another. The law of conservation of intelligence has the following practical corollaries:

  1. There exists an environment and a job for each human where they are better at performing the job than every other human.
  2. No employee is better than other for a random task.
  3. The optimal application of human intelligence is to find a task that the human is already good (just as deep learning AI would choose a task that has plenty of data and unknown non-linear relationships, and rule-based AI would choose a task where there is limited data and the mapping from input to output is known.)
  4. For the purpose of hiring, hire someone who is best suited for a given task based on their metrics on previous data.
  5. For the purpose of collaboration, find people who are best fits for the tasks they need to be delegated.
  6. Experienced employees need not always be preferred because their training can be overfitting and can be less valuable than younger employees (just as an AI model that is overfitting is harmful compared to an AI model that is trained with lesser number of epochs).
  7. When the environment is violent, an AI or human agent can choose three action themes: 1) violence, 2) obedience, 3) reason. Good agents most often choose reason — sometimes called civil disobedience.
  8. The problems with implementing a technology is rarely the technology itself and always the incompatibilities in the assumption of human intelligence involved. Automate what can be automated through artificial intelligence; develop a clear process like checklists for the rest.
  9. Behavior of employees can be rewarded using reinforcement learning algorithms.
  10. A company’s intelligence is the sum of employees’ intelligence and artificial intelligences used.
  11. Higher human intelligence does not mean higher artificial intelligence and vice-versa.
  12. Artificial Intelligence can reproduce by combining with other artificial intelligence (Ex: Hybrid systems) or natural intelligence (Ex: Human in the loop).
  13. Intelligence is transferred through reproduction, collaborative activities, teaching, and creating training data
  14. Intelligence is inherited to individuals, systems, and companies from their ancestors and also from the lifestyle and exposure of their parents and themselves.
  15. In addition to natural intelligence and artificial Intelligence tapped in computer applications, there is a practically infinite amount of intelligence that is yet untapped (only a small part of which is tapped yet through evolution of organisms and systems).
  16. It is possible for human intelligence to exceed tapped artificial intelligence and vice-versa.
  17. It is also possible for human super intelligence and artificial super intelligence to co-exist.
Photo by Drew Graham on Unsplash