History of Artificial Intelligence

First draft

2025-11-23

Latest draft

2825-11-23

Artificial Intelligence

Prometheus

We have longed for the power to create life and overcome death for millenia. We have yarned to transcend our limitations, fly like birds, dive like fish. We have imagined gods who control the natural forces and we have dreamt of achieving such powers ourselves.

We have also been very wary of the risks of such pursuits: Prometheus stole the fire from the gods and ended with his liver being eaten out every day over and over again. Icarus fell.

what is intelligence anyway

• there is no such thing as "human level intelligence"
  • it assumes that intellignece is a scalar one dimiensional variable, low on some creatures, highest on humans
• there definitely such thing as "human kind of intelligence"
  • different kinds of organisms have different kind of intelligence, adapted to the characteristics and environment of the organism
  • Bats need to recognize obstacles and pray in darkenes and catch the airborne pray while avoiding hitting the obstacles - human intelligence would suit such challenges very poorly. Fortunately our existence does not depend on that.
  • Other aninmal exzmples
• human intelligence
  • cultural
  • externalized

History of Artificial Intelligence

• Turing paper
  • What computers were then: first persons performing calculations for engineers, the machines used to perform calculations
  • Performing calculations appears as a task that requires high level of intelligence
  • As machine could do such intelligent tasks, why not others
• Cyberenetics, Norbert Wiener
• Analog artificial neural networks
• Dartmouth workshop
  • John McCarthy'a rather unfortunate choice of name for the new field of "Thinking machines"
• Success and optimish until the 70's
• Promises not filled, first AI winter
• Different approaches, rule based and statistical
• Deep Blue, Watson, ChatGPT

Large Language Models

• With statistical analysis of vast amounts of texts in various languges it is now possible to make more or less accurate models of language and its use
  • Even to the level where it is possible to have a "conversation" with a computer running such model
  • Still, it is a model of language, not a model of the world
• Many common language structures are based on metaphors that rely on (and get meaning from) our physical structure, expriences of our typical environment, and patterns of our interactions
  • without that base the metaphors do not carry any meaning
  • as computers do not have that base, they do not operate on meanings that we find self evident if our language expressions

What machinces can and can't do

• Can:
  • carry out highly complex algorithms on very large data sets very quickly
  • produce simulations of language based communication with high resemblance to actual human communication
  • produce new visual and auditory artefacts based on a model built of prior artefacts that have been labeled by humans
• Can't
  • operate with meanings in any of sense that humans do
  • have operation modes based on emotions (they don't have ambitions, goals, fears)
  • understand anything
  • create new understanding
  • be creative in the sense that humans can be creative