Artificial Intelligence (Narrow, General, Super)

"Artificial intelligence will reach human levels by around 2029. Follow that out further to, say, 2045, we will have multiplied the intelligence, the human biological machine intelligence of our civilization a billion-fold." -Ray Kurzweil

The Levels of Artificial Intelligence

When we hear artificial intelligence being discussed it's important to understand the levels and/or variations of AI. The most common (and only available) artificial intelligence is weak/narrow artificial intelligence. Weak AIs are highly specialized systems that are optimized to play games, translate languages, recognize images and listen/produce speech. These systems are also used to score a broad range of results in the sales, medical, scientific and other fields to assist professionals with their work. While Narrow AI is indeed disruptive in narrow situations, it has serious limitations as it has limited or negligible transferability to other other disciplines within the same or different domains. 

This is why scientists, companies and governments alike are in a race to develop a strong/general artificial intelligence system that can operate across numerous domains with mastery. These systems would be able to play all varieties of games (cross applying knowledge), write a paper, identify images, etc. and synthesize this information into something useful. They key thing for these systems is that they general AI would be able to operate at a master level (as good as if not better than a well trained human) and be able to conduct all of these tasks simultaneously. These systems do not current exist but many companies and scientist hope to have them in the near future between 2020 and 2030. 

After General Artificial Intelligence has been achieved many scientists have speculated about a third evolution of AI, Superintelligence. This level of AI is purely theoretical, but, SI systems would be operate a levels and speeds that are faster than the smartest humans beings by several millions times. It can be rather difficult to have any meaningful conversation on this topic without falling into the weeds about dystopian worlds where the AI tries to kill everyone. This isn't to say that it can't happen, but as it currently stands there isn't enough information to make a educated assessment of what that future might look like. 

Weak - Strong - Super Overview 

Narrow/Weak Artificial Intelligence 

  • Simple Scoring Systems 
  • Chess, Go and other game playing systems.
  • Visual systems (image identification)
  • Audio systems (speech recognition)
  • Self-Driving Systems 
  • Other Automation Systems

General/Strong Artificial Intelligence

  • Robust muti-use and multi-purpose systems 
  • Master level competencies in skills
  • Do most tasks any human would be able to
  • Cross application of skills and knowledge 
  • Ability to synthesize diverse information and develop actionable insights or actions

Superintelligence 

  • Superintelligent AI Agents would be far superior to the smartest humans in terms of capacity and speed. 
  • Would be a master at every skill, subject or discipline.
  • Would be millions of times faster than the smartest humans.
  • Would be able to do all tasks simultaneously without a decrease in quality of outputs or speed.