Table of Contents
Moshe Sipper
The following is an abridged bio from his personal site:
Moshe Sipper is a Professor of Computer Science at Ben-Gurion University of the Negev, Israel. He received his B.A. degree from the Technion — Israel Institute of Technology, and his M.Sc. and Ph.D. degrees from Tel Aviv University, all in Computer Science.
He teaches Evolutionary Computation and Intro to Computer Science at Ben-Gurion University of the Negev in Israel.
Dr. Sipper's current research focuses on evolutionary computation, mainly as applied to software development and games. He also has experience researching the following areas: bio-inspired computing, cellular automata, cellular computing, artificial self-replication, embryonic electronics, evolvable hardware, artificial life, artificial neural networks, fuzzy logic, and robotics.
Dr. Sipper has published over 160 scientific publications including three research-related books: Evolved to Win, Machine Nature: The Coming Age of Bio-Inspired Computing, and Evolution of Parallel Cellular Machines: The Cellular Programming Approach.
Dr. Sipper won the 2015 IEEE CIS Outstanding Transactions on Computational Intelligence and AI in Games (TCIAIG) Paper award, the 2008 BGU Toronto Prize for Academic Excellence in Research, the 1999 EPFL Latsis Prize, and six HUMIE Awards — Human-Competitive Results Produced by Genetic and Evolutionary Computation (Gold, 2013; Gold, 2011; Bronze, 2009; Bronze, 2008; Silver, 2007; Bronze, 2005).
As you can see, he's a very busy guy…
I created a visualization of (a subset of) his publication history here.
Work in Evolutionary Computation
Dr. Sipper is interested in using evolutionary algorithms to software development and games. In his book Evolved to Win (free download) he details how he has created programs that do well in a variety of games including: Chess, FreeCell, Backgammon, Checkers, and Reversi to name a few.
Evolution of Parallel Cellular Machines
Sipper's most cited work (according to Google Scholar) is his 1997 book Evolution of Parallel Cellular Machines: The Cellular Programming Approach.
- Asks: “can we mimic nature's achievements of parallel cellular machines?”
- Shows universal computation can be attained in cellular spaces.
- Explains how logic gates, wires, signals, clock, and memory can all be embedded in CAs. In chapter 2 of the book, a pretty interesting read.
- Explores the coevolution of cellular computation.
- Shows the Firefly machine which is an online autonomous “evolware”
Darwinian Software Engineering
First presented in 2010 and then eventually published in 2012 in the Ubiquity Symposium.
FINCH: Fertile Darwinian Bytecode Harvester.
- FINCH evolves Java bytecode and using a fitness function it evaluates the output program.
- Has been used to solve the Artificial Ant and Intwined Spirals problem.
- Prides itself on turning bad seeds into good programs. But how bad can they really be?
- Fixed a “good” Tic-Tac-Toe solver that had a couple of subtle bugs.
Games
Dr. Sipper has been active in applying evolutionary computation to games since 2005. It appears that this is his true passion. He has 25 publications (his most in a particular field) in the area of games and it is featured heavily on his personal site.
He uses a combination of genetic algorithms with heuristics to build programs that are capable of competing against humans in a variety of games.
His research group has developed a custom IDA* algorithm which utilizes heuristics to search the solution space of games. Much of the research done by Sipper takes the form of developing heuristics for common games and implementing genetic algorithms to mutate these heuristics.
Sipper takes pride in creating GP players that beat other top algorithms and players. Some of his achievements in games include:
- Has the #1 FreeCell solver on freecell.net
- Evolved a mate-in-n Chess engine that beats the “Crafty” chess engine.
- Evolved a Backgammon player that is “able to beat all previous machine learning-created players.”
Rush Hour
The following is a summary of Sipper's work in Evolved to Win on the Rush Hour puzzle.
Some of the heuristics include:
- Goal distance.
- Number of cars blocking the goal.
- How many cars are freed by a given move.
- How difficult the problem is.
- If a move places a car where no other cars can reach.
The genetic algorithm evolves when to apply a certain heuristic. The genome consists of a set of policies. Genetic algorithms were also used to evolve more difficult puzzles when the algorithm was able to solve 6×6 puzzles easily.
Work in Fiction
Interestingly, Dr. Sipper is an author of three science fiction novels: Daniel Max and the King in the Tower, Xor: The Shape of Darkness, and The Peaceful Affair in addition to writing comics and regular short stories on his blog.
It appears that his target audience is young adults and his writing style is light science fiction, not necessarily rooted in hard science with the presence of wolf-men and aliens from other worlds making a regular appearance.
Xor: The Shape of Darkness is his most popular and well received book (according to Goodreads).
Synopsis:
On his twelfth birthday Lewis Nash comes home from school to find that his house has blown up to smithereens, killing his father. Having lost his mother in an accident four years earlier, Lewis realizes he is now an orphan — but he has no time to dwell on it. The moment he gets off the school bus a fearsome wolf-man tries to grab him. The boy is saved in the nick of time by Master Long, who reveals to him that he’s a Shaper from a place called Xor, which is being devoured by the Realm Pirates. Lewis learns that he must do his utmost to become the powerful Shaper he was destined to be.
Because, it would seem, he’s the one and only chance Xor has.