google deepmind attacks Go and AlphaGo attacks top human

low-Go - 3hi all, it was a very big week for AI news. google deepmind published a paper in Nature demonstrating world competitive level performance in the game of Go.[a] there were several articles and an excellent/ cool 7 minute video produced by the site. experts thought this could be ~5-10 years away. the google team was relatively small with only 15 employees. however, their combined salaries were surely not that low, could that figure probably easily approach over a million dollars annually? (we are talking about silicon valley rates here.)

google will be challenging the top world player in march for $1M, Lee Sedol. more edge-of-seat excitement!

so, this is a great excuse or milestone to write up my last batch of links. and boy is it a big pile. its only been ~½ yr since last writing on this subject.

and am going to take some big credit here. (along with wired) pointed myself to the game of Go as a ripe terra incognita for new algorithms and development. maybe a few people really do actually read this blog, eh? and this breakthru also cuts across many categories long covered on this blog.

another big cybersynchronicity, AI pioneer and near-gadfly Minsky died this week.[b]

also since the last post, Musk announced a billion dollar AI fund and they are also committed to an open source approach.[d]

google announced open sourcing its AI engine, TensorFlow.[l]

there are dramatic weekly, nearly daily advances in the field.[g] eg emotion recognition on faces etc.[i] vision algorithms are continuing to evolve rapidly and getting very sophisticated.[k]

a brilliant book by the brilliant Domingos on “the master algorithm” was released and made some media waves.[e] there actually have been a plethora of AI books at the bookstores within the last few years & am planning on reviewing about 5 of them soon. (the google Go breakthrough caught me in the middle of that.)

Continue reading

undecidability: the ULTIMATE challenge

(hi all, my favorite “haunts” in cyberspace lately seem not so busy or as engaging as in past, not sure why that is, so out of sheer lack of alternatives am going through my big link piles/ archives and finally writing up some of my big topics saved over years.)

calvin_undeadthis is a very big/ deep/ mysterious topic that have been researching and musing on for decades, almost since a teenager, starting with Hofstadters award winning/ near-legendary musings.[a5][a6] the topic also unifies many different threads/ post on this blog. (the delightful comic comes from a google image search result on the term “undecidability”… (doncha just luv AI?! which maybe even could show signs of a (great!?!) sense of humor at times?)

its one of those crosscutting areas where it can seem like very many people are inquiring into it, and almost nobody is, all at the same time. undecidability seems to be the extraordinary mathematical equivalent/ model of terra incognita aka middle age maps that stated at the edges—literally thought to be the same as the edges of the earth or even existence“here there be dragons”.

aka (mixed metaphor overload alert…) the lynchpin. the 3rd rail (of math/ computer science). haunted house/ mansion. the bermuda triangle. UFO. (boogeyman? monster? under the bed?) or how about some great physics analogies? dark matter/ energy? black hole(s)? etc!

one might say this is a major phenomenon of the massive iceberg, where 90% (or more!) is unseen, or the elephant, in the aphorism about the blind men and the elephant. its a very big elephant in this case! that may even be the understatement of the year! which reminds me, as an old anecdote went in my high school calculus book (Finney?), talking about the self-referential derivative of ex as ex, quoting an old joke/ legend/ folklore about what the indian guru said when asked about what held the earth, and said “an elephant”, and then asked “whats below that”, replied, “its elephants all the way down!”

over the many years, have discovered there is a strong crosscutting connection between these many seemingly disparate areas of computer science in math. they are basically all united by the phenomenon of turing completeness, or the flip (dark) side, (the sheer difficulty of) undecidability. maybe in some ways have some better picture of this than others, but even my own picture is quite fragmented and cloudy at times. ie just another blind mans ramblings. (“in the kingdom of the blind, the one-eyed man is king”?)

Continue reading

collatz 2016— return of the FSMs

hi all, have been playing with collatz and FSM code for over a decade now but not for a long time recently.[a] got the urge to revisit it. the excellent AT&T FSM library is a very good tool in this regard,[a5] but went to go download it today, and the url fails. is it down? broken forever? have long been thinking it might go down. its very difficult to keep software going indefinitely. this is a very good package, hope that it is not lost to the ash-bin of history….

so then was hunting down FSM code. there are a lot of state machine implementations in ruby but does not seem to be an optimizer written in ruby anywhere. found the OpenFST library, and there is a windows build, and it has the optimizer. it has a contribution by Mehryar Mohri and maybe others who worked on the AT&T FSM library. it seems to be nearly plug-in replacement/ compatible.

it has a default option to output graphviz dotty files. however, tried the sfdp rendering on some “not large” files and it failed badly with all the nodes plotted on top of each other. argh! trying to do some serious science around here with sometimes-seemingly toy tools! the idea was to try to look for emergent patterns. maybe will have to cook up some other visualization method… the general idea is that the compressed graphs presumably have some half order mixed in the (incessant) disorder. and also any method of creating the sequence of graphs through induction would be equivalent to a proof.

Continue reading

past & future

hi all. in a contemplative/ reflective/ wistful mood lately. one of my favorite stores Radio Shack died last year. didnt visit all that often but this was a geek nirvana store (it was one of my dads favorites stores also, growing up… he liked it for miscellaneous audio inventions etc). liked to have it around when needed it. it reminded me of all the stores that have died in the last decade or so. its a long list. we have been through a whirlwind of changes in a brief time and some bit of this is reflected in the at-times dramatically shifting retail landscape. aka toffler’s “future shock”.

  • borders books/ music
  • waldens bookstores
  • radio shack
  • circuit city
  • comp USA
  • sharper image
  • sony stores
  • virgin records
  • blockbuster videos

“honorable mention”: barnes & noble (one of my current favorite stores) seems to be on the ropes.[a25] the closure of these stores indicates bigger trends such as the commoditization of electronics/ computers, the shifts in music distribution away from CDs toward digital distribution (in a word, Itunes!), and also with movies eg with Netflix/ Redbox mostly killing Blockbuster videos. its hard to believe only ~1 decade ago, was driving to a store to pick up (occasionally!) VCR tapes. now have a 4K 3d tv and a ps4. wow.

can you believe its 2016 already?

this blog has had a lot of aspects of futurism in it, but sometimes the future has a brutal, darwinistic quality.

Continue reading

breakthru prize 2016

zuckerbergs_babyhi all the breakthru prizes 2016 were announced in early nov and the prize ceremony ran again. missed posting at the time on it, but dont want to let it get by completely. it didnt seem as quite as high publicity as last time but still a big deal. covered this last time with a lot of commentary.

not as much a CS angle this time, but there was a $3M math prize to Ian Algol maybe with some applied possibility.

facebooks Zuckerberg & his wife, new parents, were in attendance.

something else that caught my eye, a new “junior” prize awarded to a high schooler:

New this year, Priscilla Chan and Sal Khan announced the winner of the inaugural Breakthrough Junior Challenge, 18-year-old Ryan Chester, of North Royalton, Ohio. Priscilla Chan and Salman Khan presented Ryan with a $250,000 educational scholarship for his winning video depiction of Einstein’s theory of special relativity. His teacher, Richard Nestoff, was presented an award of $50,000. Ryan’s school, North Royalton High School, received a state-of-the art science lab valued at $100,000. The lab will be designed by and in partnership with the school and Cold Spring Harbor Laboratory, which has shaped contemporary biomedical research and education, and is home to more than 600 researchers and technicians.

Continue reading