History of Artificial Intelligence

• According to mythology Hephaistos (a god) first had the idea of an giant bronze robot “Talos” who has superhuman forces

• He apparently also had ideas for other automatic machines and robots

• Aristotle laid the foundations for formal logic –> syllogismisms

• He also had ideas of automated machines which could do the work of the slaves

• The Brazen Head was an oracle which was built like a human head (including the inside) and was owned by some midieval alchemists

• Apparently it should be able to answer any question but unfortunately the builder fell asleep before he could hear the answer

• A mechanical knight

• It could raise it legs and arms automatically

• 1495

• The Golem was a giant of clay who should protect the Jewish community

• 16th century

  
  

• First story of a scientist who tried to create life

• 1818

• It was said to be a automated chess playing machine which could beat basically anyone

• But there was a human sitting in the machine –> still impressive machinery

• 18th century

• It is a talking machine, which can speak any European language by pumping air with the belows

• 1830-1840

• Automated machine which had more or less the abilities of a duck

• 18th century

• Automated machines became more and more popular in the 18th and 19th cetury

• They have lungs, muscles, heart, etc. anything can be explained by physics

• They soul differs the human kind fom every other animal

• 17th century

• Leibniz created calculating machines, e.g. to add numbers –>” time spent with simple calculating is not humanworthy”

• Liebniz also tried to create a formal mathematical/logic language: “Lingua Characteristica”

• 17th and 18th century

• She is conserded to be the first programmer

• She wrote about the Analytical Engine that it has no will on its own but it performs the tasks human would tell it to do

• 1834

  
  

• He designed the Boolean Algebra, the first logic in the form of analytical equations –> logical and, logical or, etc.

• Mathematical Analysis of Logic

• 1847

• That the consistency of an axiomatic formalization of arithmetic is proven

• Gödel proved that any complex formalization is either incomplete or inconsistent

• Turing proved that some problems are not solvable in finite time

• 1930s

• It apperantly has its origin in Czech, meaning work, phisical labour or slave

• 1920s

• A robot must not injure or endager a human being

• A robot must obey the commands of a human being except if it violates the first law

• A robot must protect its own existence as long as it doesn’t violate law 1 or 2

• He added the 0-th law: A robot must not harm humanity

  
  

• It is a robot which can drive and collect a nut with its shovle

• 1951

• Jack Koff

• They were simple household robots, similiar to the cleaning robots today

• Z3 built by Conrad Zuse is the first digital computer

• 1941

• It is the first turing-complete digital computer

• 1945

• first stored program computer

• Neural Networks/Deep Learning

• 1943: McCulloch and Pitts model artificial neural networks

• 1951: Marvin Minskyand Denn Edmonds build the first neural network computer

• 1957: Rosenblatt defines the perceptron –> weighing the inputs

• Input from biology to build artificial neural networks, many small interconnected units (neurons)

• 7 +/- 2

• It is one of the most reknown articles in psychology

• Average human being can only hold seven different objects in short-term memory

• It was also important for the field of cognitive science

• Shannon Type A: brute-force

• Shannon Type B: concentrating on interesting lines

• first recorded man-machine (chess) game

• computation with paper and pencil

• 1951

• Dartmouth Conference –> Birth of AI

• John McCarthy invited 10 scientists with various backgrounds to a 2-week workshop at Dartmouth College due ti recent developments in these fields

• John McCarthy, Herbert Simon, Allan Newell, Marvin Minsky, Arthur Samuel, etc.

• Term “Artificial Intelligence” is coined

• General Problem Solver by Newell and Simon, Logic oriented advice taker by McCarthy, etc.

• Simon and Newells predictions, within the next ten years AI will:

  • beat the World Champoin at Chess

  • solve a mathematical theorem

  • write music on its own

  • psychlogy will have most of its theories because of Computer Science

• Predictions Simon and Newell became more or less true but instead of 10 it took 40 years

• Difficulties:

  • knowledge representation is quite hard

  • lack of scalability –> underestimation of combinatorial possibilities in search, theorems in micro-world work but not in the real world, funding stop of AI-research projects

  • fundamental limitations of techniques and representation –> Minsky stated that perceptrons only have linear seperators

• Computers and programms could be sold for a lot of money

• Knowledge-based systems and their use becomes more and more popular

• Funding raises –> technical possibilities increase

• Revival of neural networks McClelland 1986

• Multi-billion market 1988

  
  

• Expert Systems –> knowledge is aquired from experts –> rules can apply

• DENDRAL project

• Symbolic AI –> approach to model everything with logic and knowledge

• Subsymbolic AI –> approach that the model is designed like a neural network and therefore learns

• Since 2010

• We know have the

  • bigger models

  • more data for training deep networks

  • computational power for doing so

  • Readily available programming libraries make it much easier

• new breakthroughs in Computer Vision, Game playing, language understanding, etc. –> without complete new ideas

• Image classification

  • handwriting recognition

  • human face recognition

  • creating images in the style of a Grandmaster

• Game Playing

  • Alpha Go

  • Aplha Zero learns Chess tremendously quick

  • Alpha Fold can predict the strucure of proteins within the width of an atom

• Text and Imagegeneration

  • Deep Fakes

  • GPT-3 –> artificial textgenerator, co-writes skrpt for movies

  • Dell-E –> text to iamge generator

Map the following statements to their topic

• speculation

• scientific investigation

• logic

• philosophy

• psychology

• neuroscience

• 1956 at the Dartmouth Conference

• John McCarthy, Marvin Minsky, Allen Newell, Herbert A. Simon, Arthur Samuel

• McCulloch and Pitts modeled artificial neurons and showed that logical operations can be modeled with such cells

• First Turing-complete digital computer called ENIAC

• Marvin Minsky and Dann Edmonds constructed the first neural network computer

• Newell and Simon: the General Problem Solver (successfully solved simple puzzles)

• Arthur Samuel:

  • investigated game playing (checkers) with great success at IBM

  • pioneered ideas in game playing & machine learning

• John McCarthy:

  • Lisp (second-oldest high-level language)

  • Logic-oriented advice taker

• Marvin Minsky: various students working on micro-worlds

• Progress was slower than (unrealistic) expectations. Simon and Newell’s predictions came more or less true after 40 (instead of 10) years and in very different ways than they had imagined.

• Difficulty of knowledge representation

• Lack of scalability

  • Underestimation of the combinatorial explosion in search

  • Things that work well in micro-worlds often do not work in real world

  • Fundamental limitations on techniques and representations

• Many of the successes in AI depend on deep and intensive computation instead of modeling the human mind.

• Deep and intensive computations