Date: Tue, 17 Oct 2000 13:14:45 -0700 From: "Richard K. Belew" Reply-To: rik@cs.ucsd.edu Organization: CSE Dept., UC San Diego X-Mailer: Mozilla 4.51 [en] (X11; I; Linux 2.2.5-15 i686) X-Accept-Language: en MIME-Version: 1.0 To: Tyler Luu CC: Subject: Re: May You Please Answer some Questions? hi tyler, good questions! i'll be curious to see others' answers (and apologize if they didn't want to see mine!) Tyler Luu wrote: > > 1. What is the main goal of Artificial Life? there are many goals, each motivating individual researchers who choose to call their work ALife. *my* primary goal is to understand biological phenomena in computational terms. > 2. I know there is a difference between Alife and > Artificial Intelligence, but do these two field ever > run together? I mean doesn?t life need intelligence? they certainly do, all the time. i consider ALife a LOWER BOUND on AI: the dumbest 'smart' thing you can do is to stay alive long enough to replicate yourself, ie, participate in evolution. i've even said such things in print: @article{REF844, title = {Artificial Life: {A} constructive lower bound for {Artificial Intelligence}}, author = {R.K. Belew}, journal = {IEEE Expert}, month = {February}, year = 1991, volume = 6, number = 1, pages = {8-15}, } @article{REF845, title = {ALife-2: The Second Artificial Life Conference}, author = {R. K. Belew}, journal = {IEEE Expert}, month = {February}, year = 1991, volume = 6, number = 1, pages = {53-59}, } > 3. How is aLife tested? Is it always by computer > simulation or is there a way to do it in the real > world? the most conservative answer is to say that our alife models are correct when predictions they make (ie, when the output they generate given certain input characteristics) are consistent with what the real bio systems do. but i continue to believe there are other important roles for computational models. with melanie mitchell and dave ackley, i've written about this too! @incollection{REF1128, author = {R. K. Belew and M. Mitchell and D. H. Ackley}, title = {Computation and the natural sciences}, chapter = 24, pages = {431-440}, booktitle = {Adaptive Individuals in Evolving Populations: Models and Algorithms}, publisher = {Addison-Wesley}, year = 1996, volume = {XXVI}, series = {Santa Fe Institute Studies in the Science of Complexity}, } > 4. Does all of aLife use the principles of evolution? not necessarily. some workers consider origin of life questions that must be considered PRIOR TO evolution. others consider phenomena (eg, cultural processes) that are almost independent of evolution. it is true, tho, that evolution is central to most alife research. note also that there have been CREATIONISTS who have used (their interpretation of!) alife results to argue that evolution is impossible! i think that is ironic as could be. > 5. Is it ever possible for aLife to be autonomous? > Doesn?t a human always have to start the cycle and > implement the different variables of entropy and > evolution into the environment first? currently, yes. many consider the possibilities of true SELF replication. eg, Richard Laing, working for NASA at UMich in the mid-80s, was interested in the smallest rocket you could send to the moon or mars, with sufficient BASIC info and materials that it could, from there on, self-subsist. > 6. Life as we know it today is only one set of a > combination that can lead to life. So, with > artificial life have we found other ways to create > life? > 7. What are the applications of aLife? there are many, and more all the time as our computational and molecular genetic capabilities increase: - bioinformatics - individual-based modelling of species thru economies - DNA-based computing - drug design - molecular and nano-scale fabrication > 8. The chaos theory runs our lives, it?s a part of our > everyday lives. How important is it in aLife? do i have to believe your assertions?! anyway, i'm not sure i understand your question well enough to answer. yes, chaos exists, and yes also in bio- and a-life systems. i don't think i believe it is especially important to b-/a-life, but it is true that some of alife's first proponents were also actively involved in dynamical systems (the broader cateory, including chaos) research. > 9. Is it possible for aLife to follow the rules of > creationism? That an all mighty being created the > heaven and earth? see above. this interpretation is inconsistent with mine. > 10. How were robotics problems previously tackled > before evolutionary techniques? lots of work has been done on fixed (eg, assembly-line) robots. robotics has been an importatn AI topic since the beginning. see any recent text eg, @Book{Russell95, author = {S. Russell and P. Norvig}, title = {Artificial Intelligence: A modern approach}, publisher = {Prentice Hall}, year = 1995 } > 11. In Alife, how do determine the successfulness of a > robot? generally, it's 'fitness' at some task. eg, Sony used an evolutionary algorithm to search for 'getting up' behaviors for its new robotic dog. they used (as i recall) the amount of time it took to get up as a fitness function. > 12. Are some principles of evolution definite in life? > I?ve known of some projects where the adaptation > happens in every single test, but the property was > never put into the program it self? i don't know what you mean. > 13. Since evolution and life seem to be a chaotic > system, how do you recreate this randomness in a > simulation? random number generators are easy to program. the fact that they are really only 'pseudo-random' is rarely of consequence. > 14. Which do you think is better, a simulation with > simple parts or more complex ones? simple as possible, to capture the phenomena of interest! > 15. Do you think complexity in the environment is > important to aLife? yes. but complexity too means lots of different things to different people. see my 'Adaptive Individuals in Evolving Populations" book with Melanie Mitchell, ibid. > 16. In the legend of GOLEM, a Jewish rabbi made a > creature made of mud and clay to serve his people. > The golem learned and soon became weary of being a > servant so he rebelled against his people. Do you > think this could ever happen in aLife? i can imagine that it might, but i have a very good imagination. (it appears you do too!) but this eventuality seems very far distant to me, and there are many more interesting, relevant questions to occupy me in the shorter term. > 17. When is aLife going to reach a state of high > complexity? What I mean is when is aLife going to > reach complexity of humans? i won't venture a guess. both our notions of complexity, and of ourselves, will also undergo great change in this time. > 18. aLife has borrowed its principles from nature, but > if by definition life can be much more than nature. > Then, how can we simulate life other than nature? you're a philosopher! if you believe "life can be much more than nature" you and i must use the terms LIFE and NATURE differently. > 19. Do you think that aLife can actually prove the > theory of evolution? yes. some are already convinced of the truth of evolution (me, for instance). i hope to use simultations to help explain the process to others. > 20. As a robot develops on a simulator it gets more > and more complex, but as it gets more complex the > degree of complexity gets less and less between > generations. Why is this? i don't know what data you are pointing to? > 21. What would you say were the major achievements of > evolutionary techniques in robotics so far? i'd recommend the ICGA, GECCO, PPSN conferences as all containing many interesting examples. i would appreciate seeing the report you write when you're finished. good luck. rik -- Richard K. Belew rik@cs.ucsd.edu http://www.cs.ucsd.edu/~rik Computer Science & Engr. Dept. Univ. California -- San Diego 858 / 534-2601 9500 Gilman Dr. (0114) 858 / 532-0702 (msgs) La Jolla CA 92093-0114 USA 858 / 534-7029 (fax)