AFRICAN-AMERICAN INVENTORS

Philip Emeagwali

interviewed by Susan Henderson
for the book
African American Inventors, by Susan K. Henderson.

What did you invent?

Invented methods and procedures for making computers faster and more powerful. These methods enabled me to perform the world's fastest computation of 3.1 billion calculations per second in 1988 and solve the largest weather forecasting equations with 128 million points in 1990.What stages of inventing did you go through when you were working on your invention?Programmed a computer with 65,000 processors to outperform the fastest supercomputer and thereby proving that it is best to use many processors in designing supercomputers. As a result, the technology of supercomputers now use hundreds or thousands of processor to achieve their computational speed.

Successfully implemented the first petroleum reservoir model on a massively parallel computer in 1988. As a result, one in 10 parallel supercomputers is used to find and recover additional oil and gas.

Solved one of America's 20 Grand Challenges --- accurately computing how oil flows underground and thereby alerting the petroleum industry that massively parallel computers can be used to recover more oil. Only 30 percent of the oil in a reservoir can be recovered and this discovery will enable oil companies to recover more oil.

Invented a new approach of designing supercomputers by observing and emulating patterns in nature.

Invented hyperball computer networks.

Formulated new mathematical (partial differential) equations for slowly moving liquids and gases such as the flow within the Earth's interior.

Set world record for an unprecedented parallel computer speedup of 65536 in 1990. This experiment, involving 65536 physically linked computer processors demonstrated that the speed of supercomputers can be increased a million times.

The experiment was done before the term scalability replaced "computer speedup" as an industry buzzword. In 1990, the computer industry did not understand the implications of my experiments and scalable systems. Today, scalability is cool and every vendor promises that their system is scalable.

So what does a speedup of 65536 mean to a customer conducting business on the Internet or networked computer? It means that:

- computers can operate continuously without any down time.
- commercial transactions are safe from hackers.
- smaller applications can to be ported to bigger computers without the additional expense of rewriting the original software. This is a form of investment protection.
- response time can remain constant at complex high-volume websites such as the Olympic Games, airline reservation computers, and Internet search engines.

First, I find an unsolved problem to solve such as demonstrating that computers with numerous inexpensive processors can be cheaper and faster than computers with a few expensive processors. Another example, was demonstrating that computers with thousands of processors can be used to discover and recover more oil and gas from oil fields.Were you working by yourself or with a team?Second, I planned how to solve the problem by writing to oil companies and computer manufacturers to learn what their needs are. I researched and visualized the solutions by writing a 1000-page report that outlined how the problems will be solved.

Third, I began to develop my computer programs and learned what needs to be changed along the way. I revised my 1000-page report to include my discoveries and inventions.

Finally, I sent my work to experts in my field to review and verify that I have solved an unsolved problem.

I work alone. However, since I have expertise in several fields I generally can conduct research done by a multidisciplinary team. I decompose a multidisciplinary problem into simpler and separate problems. Finally, I reassemble the solution to these simpler problems like a conductor directing individual musicians to create the sound of a symphony.Why do you like inventing?

Inventing allows me to turn an idea in my head into reality and make everyday live easier for everyone.What makes you a good inventor?My invention of a method for using numerous processors to simulate how oil flows underground is used by the petroleum industry to recover more oil. I find it satisfying to know that every time someone pumps gas into their car that my invention made it possible to drive that car for a longer distance.

The method that I used to perform the world's fastest computation can be used to design faster computers.

How did you get your idea for the invention?

- I am a lateral and unorthodox thinker who draws lots of analogies from natural observations.
- I practice free-associating and let my mind wander freely.
- I generate lots of ideas and then throw the bad ones out.
- I un-create and re-create things and then put them back together in a new way.
- I use a different perspective to solve problems.
- I study my problems from the inside, outside and backwards.
- I visualize the solution to my problems.
- Also, being well-rounded helps me become a better inventor. I get my best ideas while playing tennis or reading a book in an unrelated field.

Clickhere for full-scale 65536-processor Connection Machine.

It depends on the particular invention. Generally, my ideas come from brainstorming and divergent thinking. Convergent thinkers concentrate on the problem they are trying to solve while divergent thinkers look for solutions in unrelated fields.Why did you invent it?Below are a few specific examples:

- Credited by the book " History of the Internet" as one of the pioneers of the Internet. Some publications describe me as "one of the fathers of the Internet" but it is important to know that the Internet has many fathers and mothers.
- The original proposal for ARPAnet called for a national network of supercomputers. I was the first to relate this project with a 1922 science fiction proposal for "64,000 [human] computers to race the weather for the whole globe."
In 1975, I envisioned my HyperBall

internationalnetwork as the geometrical and topological structure of a Global SuperBrain (with 64,000 computers) that can forecast the weather.- World's fastest computation of 3.1 billion calculations per second in 1988. (AT FIRST, I DID NOT REALIZE THAT I PERFORMED THE WORLD'S FASTEST COMPUTATION. THIS WAS EXPLAINED TO ME LATER).
- World record for solving the largest partial differential equations with 8 million grid points in 1988.
- World record for solving the largest weather forecasting equations with 128 million grid points in 1990. (I HAD A 65,536-PROCESSOR COMPUTER THAT CAN SOLVE HUGE PROBLEMS. I MADE A LIST OF THE MOST IMPORTANT ONES AND THIS LIST INCLUDED WEATHER FORECASTING).
- World record for an unprecedented parallel computer speedup of 2048 in 1988.
- World record for an unprecedented parallel computer speedup of 65536 in 1990.
- First successful implementation of a petroleum reservoir model on a massively parallel computer in 1988. (IN THE 1980'S, IT WAS BELIEVED THAT MASSIVELY PARALLEL COMPUTERS COULD NOT BE USED TO SOLVE PRACTICAL PROBLEMS. TO PROVE THAT THE LATTER STATEMENT IS WRONG, I IMPLEMENTED A RESERVOIR SIMULATION MODEL ON A MASSIVELY PARALLEL COMPUTER).
- Credited with alerting the petroleum industry that massively parallel omputers can be used to recover more oil.
- First to program a massively parallel computer to outperform a conventional (vector) supercomputer in 1988. (I VISUALIZED THE SOLUTION AND SAW THAT PARALLEL COMPUTERS SHOULD BE FASTER THAN VECTOR COMPUTERS. THAT VISION INSPIRED ME TO WRITE A PROGRAM THAT DEMONSTRATED THAT PARALLEL COMPUTERS ARE FASTER THAN VECTOR COMPUTERS).
- First to have applied a pseudo-time approach in reservoir modeling in 1988. (I STUDIED THE 200-YEAR HISTORY OF THIS PROBLEM AND DISCOVERED THAT THE EXISTING TECHNIQUE IS OBSOLETE. THIS DISCOVERY INSPIRED ME TO INVENT A NEW APPROACH).
- Credited with pioneering the use of the ``vast resources" of the Internet in supercomputing in the 1980's.
- Credited with conclusively demonstrating that computers with thousands of processing nodes can solve significant real-world problems in 1988. (I VISUALIZED THE SOLUTION AND SAW THAT COMPUTERS WITH THOUSANDS OF PROCESSING NODES CAN BE USED TO SOLVE REAL PROBLEMS. THAT VISION INSPIRED ME TO WRITE A PROGRAM THAT DEMONSTRATED THAT NUMEROUS PROCESSING NODES CAN BE USED TO SOLVE REAL-WORLD PROBLEMS).
- Formulated the counter-intuitive speedup paradox which states that there are two different but correct theoretical speedup of parallel computers. (THE PROBLEM HAD TWO ANSWERS AND IT WAS BELIEVED THAT ONLY ONE OF THEM COULD BE RIGHT. I PROVED THAT BOTH ANSWERS ARE RIGHT).
- Discovered the counter-intuitive hypercube paradox. (I GOT THE IDEA BY AN STUDYING UNRELATED QUESTION: What is the tightest way to pack oranges in a box?).
- Formulated the theory of tessellated models for parallel computing. (I GOT THE IDEA BY OBSERVING HOW THINGS ARE ARRANGED IN NATURE)
- Formulated the theory of weak nearest-neighbors in parallel computing.
- Demonstrated that the most communication-efficient parallel programs must be computation-inefficient. (BY OBSERVING HOW BEES BUILD THEIR HONEYCOMBS).
- Introduced the concept of network frequency for parallel computers.
- Introduced the concept of parallel data spaces. (BY OBSERVING HOW THINGS ARE ARRANGED IN NATURE).
- Introduced the concept of symmetry in parallel computing. (BY GEOMETRICAL VISUALIZATION).
- Discovered chirality in parallel programming. (BY OBSERVING THE STRUCTURE OF LIVING THINGS IN NATURE).
- Discovered duality in interconnection networks. (BY STUDYING GEOMETRY).
- Discovered helicity in interconnection networks. (BY NATURE OBSERVATION)
- Discovered enantiomeric networks. (BY OBSERVING THE STRUCTURE OF LIVING THINGS IN NATURE).
- Designed the first Emeagwali-Fibonacci network. (BY STUDYING THE BREEDING PATTERNS OF RABBITS AND THE GROWTH PATTERNS OF TREES AND PLANTS).
- Discovered the relationship between sphere packing and fast computing. (BY STUDYING NATURE AND GEOMETRY).

Tightly packed spherical objects

- Invented hyperball computer networks. (BY STUDYING THE STARS AND PLANETS IN THE SOLAR SYSTEM AND BEYOND).
- Formulated new partial differential equations for low inertial fluid flows.
- Derived error stopping criterion for high inertial fluid flows. (TO CONTROL THE ERROR IN MY MATHEMATICAL CALCULATIONS).
- Devised novel procedures for engineering-in inertia into low-inertial computational fluid dynamic codes. (BY STUDYING METEOROLOGY)
- Discovered the analogy between Darcy's equations used in petroleum reservoir simulations and geostrophic equations used in weather forecasting. (BY SIMULTANEOUSLY STUDYING AND COMPARING METEOROLOGY AND PETROLEUM ENGINEERING).
- Derived a new set of porous media flow equations that is vectorizable, parallelizable, and surprisingly, fifty times less computation-intensive than the original formulation.
- Proved that the use of only Dirichlet type boundary conditions yields more accurate numerical solutions in the vicinity of petroleum production wells located near the boundary and is therefore suitable for avoiding the coning problem caused by the high velocity of converging flows in the vicinity of wells.
- Demonstrated that it is impossible to use strictly hexagonal lattice gas cellular automaton methods to implement a complete climate model.
- Predicted that fatal subtraction error will occur in all fast computers. Fourteen months later, this prediction came true on 25 February 1991, when the fast computer guiding the U.S. Patriot missile made a fatal subtraction error and failed to shoot down the Iraqi Scud missile and killing 28 U.S. soldiers. (THIS WAS AN ACCIDENTAL DISCOVERY THAT OCCURRED WHEN I PERFORMED CALCULATIONS THAT WAS SO FAST THAT MY COMPUTER'S INTERNAL CLOCK COULD NOT MEASURE THE TIME IT TAKES TO MAKE ONE CALCULATION).

It depends on which of my inventions that you are talking about. Below are a few examples:

- My invention of the HyperBall (as an intelligent SuperOrganism, SuperBrain or "brain of brains") was inspired by the 1922 science fiction proposal on how to use 64,000 human computers to forecast the weather for the whole globe. Instead of 64,000 human computers, I substituted 64,000 electronic computers and came up an
internationalnetwork of computers that I named HyperBall.- World's fastest computation of 3.1 billion calculations per second in 1988. (WHY DO WE CLIMB MOUNT EVEREST? BECAUSE IT IS THERE. WHY DID WE FLY TO THE MOON? BECAUSE....USING A PREVIOUSLY UNACCEPTABLE PARALLEL COMPUTER TECHNOLOGY TO PERFORM RECORD COMPUTATIONS CONVINCE THE INDUSTRY TO ACCEPT THIS NEW TECHNOLOGY.)
- World record for solving the largest partial differential equations with 8 million grid points in 1988. (TO DISCOVER A METHOD FOR SOLVING DIFFICULT MATHEMATICAL EQUATIONS).
- World record for solving the largest weather forecasting equations with 128 million grid points in 1990. (TO INCREASE THE ACCURACY OF WEATHER FORECASTS).
- World record for an unprecedented parallel computer speedup of 2048 in 1988. (TO UNDERSTAND HOW TO DESIGN SUPERCOMPUTERS).
- World record for an unprecedented parallel computer speedup of 65536 in 1990. (TO UNDERSTAND HOW TO DESIGN SUPERCOMPUTERS).
- First successful implementation of a petroleum reservoir model on a massively parallel computer in 1988. (TO INCREASE THE AMOUNT OF OIL RECOVERED FROM OILFIELDS)
- Credited with alerting the petroleum industry that massively parallel computers can be used to recover more oil.
- First to program a massively parallel computer to outperform a conventional (vector) supercomputer in 1988. (TO PROVE THAT MASSIVELY PARALLEL TECHNOLOGY SHOULD BE USED IN FUTURE SUPERCOMPUTER DESIGNS. THIS RESULT WAS CITED IN THE PARALLEL-VERSUS-SEQUENTIAL COMPUTING DEBATE OF THE 1980s).
- First to have applied a pseudo-time approach in reservoir modeling in 1988. (TO FORMULATE BETTER MATHEMATICAL EQUATIONS FOR RECOVERING MORE OIL)
- Credited with pioneering the use of the ``vast resources" of the Internet in supercomputing in the 1980's.
- Credited with conclusively demonstrating that computers with thousands of processing nodes can solve significant real-world problems in 1988. (TO ENCOURAGE THE WIDE USAGE OF COMPUTERS WITH SEVERAL PROCESSORS. IT WAS PREVIOUSLY BELIEVED THAT A 65536-PROCESSOR COMPUTER WILL BE UNWORKABLE 65536 TIMES MORE DIFFICULT TO PROGRAM THAN ONE PROCESSOR. DEMONSTRATED THAT PROGRAMMING THOUSANDS OF PROCESSORS IS NOT AS DIFFICULT AS WAS PREVIOUSLY BELIEVED).
- Formulated the counter-intuitive speedup paradox which states that there are two different but correct theoretical speedup of parallel computers. (TO INCREASE OUR UNDERSTANDING OF MORE POWERFUL COMPUTERS).
- Discovered the counter-intuitive hypercube paradox. (TO INCREASE OUR UNDERSTANDING OF HYPERCUBE COMPUTERS).
- Formulated the theory of tessellated models for parallel computing. (TO GAIN A BETTER UNDERSTANDING OF HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Formulated the theory of weak nearest-neighbors in parallel computing. (TO GAIN A BETTER UNDERSTANDING OF HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Demonstrated that the most communication-efficient parallel programs must be computation-inefficient. (TO UNDERSTAND HOW TO PERFORM FAST COMPUTATIONS).
- Introduced the concept of network frequency for parallel computers. (TO BETTER UNDERSTAND HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Introduced the concept of parallel data spaces. (TO BETTER UNDERSTAND HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Introduced the concept of symmetry in parallel computing. (TO BETTER UNDERSTAND HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Discovered chirality in parallel programming. (TO BETTER UNDERSTAND HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Discovered duality in interconnection networks. (TO BETTER UNDERSTAND HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Discovered helicity in interconnection networks. (TO BETTER UNDERSTAND HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Discovered enantiomeric networks. (TO UNDERSTAND HOW TO DESIGN POWERFUL COMPUTERS AND COMPUTATIONS).
- Designed the first Emeagwali-Fibonacci network. (TO DESIGN FASTER SUPERCOMPUTERS).
- Discovered the relationship between sphere packing and fast computing. (TO BETTER UNDERSTAND HOW TO DESIGN MORE POWERFUL COMPUTERS AND COMPUTATIONS).
- Invented hyperball computer networks. (TO DESIGN MORE POWERFUL COMPUTERS).
- Formulated new partial differential equations for low inertial fluid flows. (TO OBTAIN BETTER MATHEMATICAL SOLUTIONS TO SCIENTIFIC PROBLEMS).
- Derived error stopping criterion for high inertial fluid flows. (TO REDUCE THE ERROR IN MATHEMATICAL COMPUTATIONS OF IMPORTANT ENGINEERING PROBLEMS).
- Devised novel procedures for engineering-in inertia into low-inertial computational fluid dynamic codes. (TO INCREASE THE AMOUNT OF OIL RECOVERED FROM OILFIELDS).
- Discovered the analogy between Darcy's equations used in petroleum reservoir simulations and geostrophic equations used in weather forecasting. (TO BETTER UNDERSTAND HOW TO INCREASE THE AMOUNT OF OIL RECOVERED FROM OILFIELDS AND HOW TO INCREASE WEATHER FORECASTS).
- Derived a new set of porous media flow equations that is vectorizable, parallelizable, and surprisingly, fifty times less computation-intensive than the original formulation. (TO INCREASE THE AMOUNT OF OIL RECOVERED FROM OILFIELDS).
- Proved that the use of only Dirichlet type boundary conditions yields more accurate numerical solutions in the vicinity of petroleum production wells located near the boundary and is therefore suitable for avoiding the coning problem caused by the high velocity of converging flows in the vicinity of wells. (I SELECTED RESERVOIR SIMULATION BECAUSE RECOVERING ADDITIONAL OIL IS IMPORTANT TO OIL-PRODUCING COUNTRIES, INCLUDING MY NATIVE COUNTRY OF NIGERIA).
- Demonstrated that it is impossible to use strictly hexagonal lattice gas cellular automaton methods to implement a complete climate model. (TO UNDERSTAND HOW TO FORECAST GLOBAL CLIMATE WARMING).
- Predicted that fatal subtraction error will occur in all fast computers. Fourteen months later, this prediction came true on 25 February 1991, when the fast computer guiding the U.S. Patriot missile made a fatal subtraction error and failed to shoot down the Iraqi Scud missile and killing 28 U.S. soldiers. (TO SAVE HUMAN LIVES).

Click on emeagwali.com for more information.