Tähän History of AI aikajana video
At the moment we are living midst of an AI hype of some kind. During the latest decade we have truly seen a burst of research and currently the applications of AI really do appear everywhere. How did we get here? What happened before this time and why AI has become popular right now?
Turns out that AI is by no means a new field in general. Many mathematical tools, innovations and algorithms applied today have been developed decades, some even centuries ago. During the years, as new innovations have been made, the history has seen similar hype periods as we encounter now, but for some reason or another they have all ended with only a little practical progress compared to current times. Such a reason has been, for example, the immaturity of the hardware or too high expectations by the researchers. These periods of disappointment and less activity are called AI winters, illustrating a cold period when nothing is blooming.
At times, however, the branch of AI has gained a lot of interest, and thus also money, and these hype periods have all managed to generate new knowledge applicable also today.
Let us go through some of the major AI developments that happened along the way:
The British mathematician Alan Turing applies his theories to prove that a computing machine, known as a 'Turing machine', would be capable of executing cognitive processes. This laid the foundation for what we call artificial intelligence today.
Warren McCullough and Walter Pitts published a paper describing how neurons might work.
Alan Turing creates the Turing test (originally known as the Imitation game), with the goal of testing a machine's ability to exhibit intelligent behavior equivalent to, or indistinguishable from a human.
During the same year, Isaac Asimov devised the 3 rules of robotics.
A program capable of playing checkers was written by Christopher Strachey, while a chess-playing program was written by Dietrich Prinz.
A conference called the Dartmouth Workshop was organized and it produced results that rose a lot of expectations to neural networks and AI in general that carried out until 1974, the beginning of the first AI winter. This conference also initialized a paradigm later called Good-Old-Time-Fashioned AI (GOFAI) that focused on symbolic reasoning and logic.
An industrial robot, Unimate, was installed into General Motors assembly line. It took over a dangerous job of moving die castings from an assembly line and welding them on the car bodies.
Joseph Weizenbaum (MIT) invents 'ELIZA', the first interactive chatbot in history
Stanford Heuristic Programming Project developed the very first expert systems that tried to emulate rule based human decision making in tasks like diagnosing infectious diseases.
This report pointed out strong disappointment in the slow progress of AI and disbelief in the ability to overcome the problems raised by the combinatorial explosion problem. AI had proved to be working well with toy problems but with most practical applications of AI the need of required computing power and memory was too high for the technology of that time.
Also some other agencies such as Defense Advanced Research Projects Agency (DARPA) of USA canceled their funding leading to a long period with very low resources in research of AI. This “AI winter” ended in 1980.
LISP machines were developed and expert systems running in them proved to be powerful in many application fields. This gave new hope for AI research and ended the first AI winter.
‘NETtalk’ program was taught to speak by inputting sample sentences and phoneme chains. It is one of the early artificial neural networks — programs that are supplied with large datasets and are able to draw their own conclusions on this basis.
Market of the LISP machines used with the expert systems collapsed practically overnight as new cheap and powerful desktop computers became available. There was no longer a need for LISP machines and the huge business around them disappeared leaving the expert systems running in them obsolete. This second AI winter lasted until 1993.
The evolution of the computers had led to more powerful machines with much more memory allowing the implementation of several previously developed, but then abandoned, ideas. Most important such ideas are application of probability, neural networks and machine learning in general. All these have enabled to handle uncertainty in contrast to strict rule based approaches of GOFAI.
The AI chess computer, IBM Deep Blue, defeated Garry Kasparov.
The future often seen in sci-fi movies got closer than ever, thanks to Google starting their work on the first self-driving car.
The computer program ‘Watson’ competes in the Jeopardy quiz show and won against human players. In doing so, Watson proves that it understands natural language and is able to answer difficult questions quickly.
AlphaGo, developed by Google DeepMind, won Lee Sedol 4-1 in series of five games. Initially AlphaGo was only given the rules of the game and by following the rules and, playing first against human players and later against itself, it managed to develop strategies never thought by humans. This type of machine learning method is called reinforcement learning.
Technology leaps in the hardware and software fields pave the way for artificial intelligence to enter everyday life. Powerful processors and graphics cards in computers, smartphones, and tablets give regular consumers access to AI programs. Digital assistants like Siri, Alexa, and Cortana particularly enjoy great popularity.
Tai tähän History of AI aikajana video
For each question, pick up the choice that best answers the question or defines the concept