Odyssey
From GusWiki
In April 2010, someone introduced me with:
<< This is Gustavo. He has been traveling the world for the last 10 years, educating himself... not unlike -:Odysseus. >>
(this cracks me up)
This is a timeline of my journeys, both geographical and intellectual. I have always known that I wanted to be a researcher, but it took me a long time to narrow it down further than that. I've always had ideas, but the ability and the habits of turning them into real contributions is something I had to pick up over time.
To sum up 10 years in one sentence: these 10 years transformed me from a wide-eyed AI dreamer with no disciplinary bounds, interested in philosophical and transhumanistic speculation, into a more grounded and focused researcher, interested in specific issues and having real impact (e.g. in the form of peer-reviewed publications).
2000 - Bell Labs, applying to grad school
I originally wanted to do a PhD in Applied Math. At a summer internship at Bell Labs, where I made lots of nerd friends, I was persuaded that I probably didn't want to be a mathematician afterall, seeing the fate of many a good Coding Theorist. Returned to my last year at Bucknell, where I took Topology, which made me wary of the -:Axiom of Choice and annoyed at sloppy mathematical notation and the lack of theorem-proving tools for mathematicians for the more tedious proofs. Applied to Brown CS and MIT Media Lab, unsuccessfully; despite visiting Brown twice, and Media Lab once. At Media Lab I wanted to work on tools for helping people read and understand text faster. At Brown, I wanted to work on -:Machine Translation or -:Game Theory. Back at Bucknell, algorithms class with Steve Guattery; "undergraduate reading" with Karl Voss on harmonic analysis; advised to not go into Statistics, because I dislike "cookbooks" and because the questions I asked made me a Bayesian (this is terrible advice, outdated at best!)
2001 - graduating from Bucknell; moving to Boston
Graduated with a B.S. in Mathematics and Computer Science. Moved to Boston. Got a job where I wrote Java and made web applications with JBuilder, and random admin things like network permissions and preventing spam relay. Made friends through a monthly philosophical salon, but never quite identified with their ideology. Started a blog on LiveJournal. Had discussions in the "Armchair Economist" mailing list, learned to think like a neoclassical economist (this was years before Freakonomics was cool).
2002 - Boston, traveling in Europe
Still in Boston. Spoke briefly to Dan Dennett, after his talk at MIT. Bored with the job, I was looking for an outlet for my creativity. I was accepted to do an MSc in CS at Tufts. In the meantime, I left everything to attend a logic Summer School in Italy (ESSLLI) and traveled in Europe for 3.5 months, where I met Faré. Postponed and then cancelled my attendance at Tufts, since I wasn't feeling so hot about it, and the only real AI person was busy being Head (a couple of years later, he died of cancer). Set my eyes on the very interdisciplinary ILLC in Amsterdam. I wanted a degree in "Gödel Escher Bach studies".
2003 - Recife; MSc in Logic in Amsterdam: difficulties
While moving my stuff from Boston, I had dinner with Ed Fredkin, Marvin Minsky, Push Singh. Met a student of Minsky's who was interested in me adding culturally-specific information to OpenMind. Took classes at UFPE in Brazil, "Information Theory" (at the EE department) and "Artificial Intelligence" (which was great fun).
Started an MSc in Logic at the ILLC (Universiteit van Amsterdam), which was pretty miserable at first (Core and Modal Logic, especially). Advisors: I dumped Advisor#1 for Advisor#2 (I had a small number of choices). Although neither seemed able to hear what I was saying, #2 at least respected my ability to decide what I wanted. The Program Director was a great friend to me, despite defending the status quo of the Traditional Logic Curriculum (which I managed to avoid, as I found out that I wasn't interested in it afterall), and by far the best listener I had there. The Bigshot Prof was only interested in big ideas and went all firehose on us during our first semester, until he noticed how much he was torturing the less prepared students, like me. At the same time, everyone skirted the responsibility of teaching Remedial Logic. One third of my graduating class dropped out, with varying degrees of burning. Learned Dutch.
2004 - Amsterdam: tutoring, formalization
Small project building a bilingual dictionary from a parallel corpus, learning to match Portuguese words to English words based on simple statistics. Made me very happy. Decided that the research on Cognitive Science and Semantics wasn't computational enough for me. Did some tutoring. Attended a wonderful reading group on Music Cognition. Basic Category Theory. Kolmogorov Complexity with Paul Vitanyi. Read more about formalization of science, computational structuralism at Jaap's suggestion; had beautiful dreams about the semantic science web and educational software (e.g. automatically generated curricula) and regretted the inactivity of an ILLC research group on Formalization of the Social Sciences. My dreams would later be expounded in my "QED Project for Science". Took a class on Coq with Freek Wiedijk, who was a helpful and cool instructor, but he thought of it as an artform, and I never convinced him of the great potential of this stuff.
2005 - Amsterdam: thesis
Peter Grünwald's class on MDL; Learned Lisp and wrote an equational -:theorem prover. Wrote a crackpot thesis with ThesisSupervisor, who I met because he recalled a book I was reading, and the library failed to protect his privacy (I did not break any laws: they wrote his name next to the book!). My new obsession became Argument mapping, which I saw as a type of logical formalization at a higher level of description. Volunteered at AIED.
Met Kevin Kelly, a learning theorist from CMU Philosophy, loved his ranting, and started reading his book. Despite my questioning the usefulness of worst-case analysis, this book helped shape my computationalist view of mathematics.
Graduated from my MSc with a Logic and Cognitive Science concentration.
2006: Munich; moving to CMU
Moved to Munich for a Lisp job; miserable at work, and merely unhappy socially. Despite my reservations about Philosophy degrees, Kelly convinced me to apply to the PhD program in Philosophy. Around March, CMU Philosophy sent me an acceptance letter... to do another MSc! A degree almost identical to the one I'd just received! Subtracting injury from insult, however, they sponsored a visit to CMU. While there, I was determined to get a part-time assistantship, in order to have time to produce the much-needed publications. I emailed ~15 professors, and met with ~7. The very last one gave me a job, after a fascinating conversation about AI and Cogsci applied to Education. Attended the ACT-R Summer School at CMU, and tried to keep that door open, but I didn't click with any Psychologists well enough.
Moved to CMU. Discussed lots of ideas with William Cohen about Argument Mapping. Modeling philosophical and political discourse would involve interesting data structures (see Rhetorical Structure Theory), but ultimately I learned that we couldn't wait for the data forever. No data means no machine learning! By now I was committed to the field of machine learning: it's like AI, involves a mix of math, algorithms and applications, and is a good fit for my talents.
2007 - CMU: research on causality
Started research with Peter Spirtes from CMU Philosophy on discovering causal graphical models. Attended the IPAM Summer School at UCLA, where I was influenced by Josh Tenenbaum's ideas. Back to my research, emailing Patrik Hoyer convinced me that he had better ideas and that he deserved to scoop us, so we changed the focus of the project to cyclic causal models. On the original idea, Peter and I became auxiliary authors.
2008 - CMU: UAI success; starting at UBC
Sat in on Cosma Shalizi's class "Chaos, Complexity, and Inference", and we started a mathematical project in the summer about Coupled Markov Chains. My UAI paper was accepted (as was Patrik's), and I presented it in Helsinki, where I also visited Patrik at HIIT. (I also wrote up my first and last Logic paper, a simple Chaitin-esque idea about how far proof search needs to run to be satisfied that no proof exists.)
Moved to UBC hoping to do a PhD with Kevin Murphy. TA'd for the first time in my life. Due to problems with funding and time, my idea of working with Kevin didn't work out. I was not eligible for any fellowships at UBC due to not being Canadian and having a prior Master's degree. I was very upset because I came with the intention to do a PhD already (at UBC, MSc students can transfer directly to the PhD as long as they have a willing supervisor).
2009 - UBC: suspense, resolution
Long and difficult search for a supervisor, in which I really opened my mind about who I'd consider working with. Summer School at the Santa Fe Institute. Got a "perfect" GPA for the first time in my life (1 A and 2 A+; over the MSc degree, I only got 1 less than "A", arguably in the easiest class; in 2010, this became 5 A/A+ out of 6). However, I finally found a reasonably good match with Jennifer Bryan, from UBC Statistics, who was coming back from sabbatical.
2010 - UBC: finishing MSc #2; Columbia: ???
Finishing MSc thesis #2. Started learning fascinating things about molecular biology, and considered working for bioinformatics labs.
Accepted an offer to a PhD at Columbia Stats, where many people have compatible interests. Looking forward to it.
