How Artificial Intelligence

<Ai>

Is Reinventing Business Computing

Advances in hardware and software have driven the development of artificial intelligence, from the first exploration in the 1950s to today's applications like predictive analytics for healthcare and financial services.

Let's trace the evolution of AI, and the key breakthroughs behind each step—from early gaming engines to applications that are paving the way for better living, such as personalized medicine, self-driving cars and smarter policing.

Behind each step were new developments in the hardware that provides the foundation for artificial intelligence. Computing power that once filled entire rooms now resides on a single server or can be tapped from the cloud.

1950
Turing test raises the bar

The Turing test, an unofficial standard for whether a machine could be considered intelligent, determines if a computer can fool a researcher into thinking it is human.

1951
Artificial, but smart

The term "artificial intelligence" entered the lexicon to describe a new class of computers that can learn from the information they're given.

1959
Like a rat in a maze

The Stochastic Neural Analog Reinforcement Calculator is considered by some to be the first AI computer (stochastic means “randomly determined”). The machine was powered by a series of vacuum tubes and motors, and mirrored the behavior of a rat finding its way through a maze.

1968
From fiction to reality

AI hit the big screen in “2001: A Space Odyssey.” The artificially intelligent HAL 9000 ran the Discovery One’s vital systems, foreshadowing many behaviors that today’s state-of-the-art AI systems are just beginning to crack, such as natural language processing and voice and facial recognition. HAL could even interpret emotions. It was the public’s first broad exposure to AI’s potential.

1972
AI with a stethoscope

MYCIN, an early AI machine developed by researchers at Stanford University, used an algorithm to identify bacteria which helped physicians treat blood infections. It could ask about the patient, recommend lab tests and suggest possible diagnoses.

1980
Smarter customer service

Researchers at Carnegie Mellon University developed the eXpert CONfigurer, a rule-based system that could automatically pick the right components to fill customer orders. Used by computer maker Digital Equipment Corp., it ran OPS5, an early operating system to support AI.

1996
IBM Deep Blue masters chess

IBM's supercomputer played dual matches against a Russian chess champion. The human contestant won the first and IBM Deep Blue the second. More power means smarter computers. Deep Blue ran at 11.38 gigaflops, while today's IBM BlueGene/Q runs at about 10,000 teraflops.

2011
IBM Watson on "Jeopardy"

AI took another big step forward when IBM's Watson cognitive computing system handily beat a "Jeopardy" world champion. Watson runs on IBM Power Systems, now used for commercial AI applications. IBM Power Systems can run predictive analytics for industries such as healthcare, law enforcement and financial services.

2014
The first AI board member

Asian venture capital firm Deep Knowledge appointed a computer to its board. The system, named Vital, is tasked with finding information not obvious to humans.

  • 1950
    Turing test raises the bar

    The Turing test, an unofficial standard for whether a machine could be considered intelligent, determines if a computer can fool a researcher into thinking it is human.

  • 1951
    Artificial, but smart

    The term "artificial intelligence" entered the lexicon to describe a new class of computers that can learn from the information they're given.

  • 1959
    Like a rat in a maze

    The Stochastic Neural Analog Reinforcement Calculator is considered by some to be the first AI computer (stochastic means “randomly determined”). The machine was powered by a series of vacuum tubes and motors, and mirrored the behavior of a rat finding its way through a maze.

  • 1968
    From fiction to reality

    AI hit the big screen in “2001: A Space Odyssey.” The artificially intelligent HAL 9000 ran the Discovery One’s vital systems, foreshadowing many behaviors that today’s state-of-the-art AI systems are just beginning to crack, such as natural language processing and voice and facial recognition. HAL could even interpret emotions. It was the public’s first broad exposure to AI’s potential.

  • 1972
    AI with a stethoscope

    MYCIN, an early AI machine developed by researchers at Stanford University, used an algorithm to identify bacteria which helped physicians treat blood infections. It could ask about the patient, recommend lab tests and suggest possible diagnoses.

  • 1980
    Smarter customer service

    Researchers at Carnegie Mellon University developed the eXpert CONfigurer, a rule-based system that could automatically pick the right components to fill customer orders. Used by computer maker Digital Equipment Corp., it ran OPS5, an early operating system to support AI.

  • 1996
    IBM Deep Blue masters chess

    IBM's supercomputer played dual matches against a Russian chess champion. The human contestant won the first and IBM Deep Blue the second. More power means smarter computers. Deep Blue ran at 11.38 gigaflops, while today's IBM BlueGene/Q runs at about 10,000 teraflops.

  • 2011
    IBM Watson on "Jeopardy"

    AI took another big step forward when IBM's Watson cognitive computing system handily beat a "Jeopardy" world champion. Watson runs on IBM Power Systems, now used for commercial AI applications. IBM Power Systems can run predictive analytics for industries such as healthcare, law enforcement and financial services.

  • 2014
    The first AI board member

    Asian venture capital firm Deep Knowledge appointed a computer to its board. The system, named Vital, is tasked with finding information not obvious to humans.

  • 2016
    Big Data Explosion

    By now, the web, mobile devices and the Internet of Things are generating millions of petabytes of data that give companies everything from sales forecasts to fraud prevention tools. But enormous memory and horsepower are needed to process it all. Systems like IBM's Power Systems LC server family provide the speed necessary to run complex, real-time calculations. Powered by breakthroughs like accelerated computing, where certain tasks—such as math calculations or graphics processing—are offloaded to dedicated processors like NVIDIA's Tesla GPU on IBM POWER8 servers, AI analytics algorithms can now predict weather patterns, crime trends, disease vectors and more.

Swipe
  • 1950
    Turing test raises the bar
  • 1959
    Like a rat in a maze
  • 1968
    From fiction to reality
  • 1972
    AI with a stethoscope
  • 1980
    Smarter customer service
  • 1996
    IBM Deep Blue masters chess
  • 2011
    IBM Watson on "Jeopardy"
  • 2014
    The first AI board member
  • 2016
    Big Data Explosion

2016

Big Data Explosion

By now, the web, mobile devices and the Internet of Things are generating millions of petabytes of data that give companies everything from sales forecasts to fraud prevention tools. But enormous memory and horsepower are needed to process it all. Systems like IBM's Power Systems LC server family provide the speed necessary to run complex, real-time calculations.

Powered by breakthroughs like accelerated computing, where certain tasks—such as math calculations or graphics processing—are offloaded to dedicated processors like NVIDIA's Tesla GPU on IBM POWER8 servers, AI analytics algorithms can now predict weather patterns, crime trends, disease vectors and more.

AI's Need For Speed

Artificial intelligence programs require fast hardware and infrastructure. Here's a look at the MIPS (million instructions per second) needed for certain tasks

1 MIPS:

Interpret simple images, such as geometric shapes, in real time

AI's Need For Speed

Artificial intelligence programs require fast hardware and infrastructure. Here's a look at the MIPS (million instructions per second) needed for certain tasks

1 MIPS:

Interpret simple images, such as geometric shapes, in real time

Demystifying AI

  • A human brain
    Artificial Intelligence
    A system able to learn from and build upon existing information
  • A human brain
    Cognitive Computing
    Coined by IBM, a system that mimics human thought processes to arrive at conclusions
  • A human brain
    Machine Learning
    A system that can learn about the world around it without being explicitly programmed to do so

Future-Proofing AI

Moore’s law, which says computing power doubles every two years, is getting harder to achieve on a single chip. Tech industry heavyweights including IBM, NVIDIA, Google and Mellanox are now working together through the OpenPOWER Foundation to enhance speed at the system level by engineering chips, interconnects, accelerators, memory and other components that work together seamlessly.

This will help ensure fast and powerful AI systems that support breakthroughs on numerous key fronts. AI’s evolution will continue to drive innovation that makes for smarter cities, improved healthcare, and other advances that will continue to improve our lives.

Learn more about how IBM Power Systems can help your company thrive with big data >
Sources: Carnegie Mellon University, Deep Knowledge, Encyclopaedia Britannica, Gartner, IBM, Stanford University