Category Archives: AI

AI + ML trends 2018½

hi all, ~250 links for 2nd half of 2018 AI + ML trends. very broad! AI/ML is now understood to be penetrating many fields/ areas. there are major (new) international govt initiatives, a new one announced in France by Macron.[c23]

this year looks a little more incremental than last year when Deepmind announced the Go game advancement. close to my own interests there is a lot of continued research into more advanced games such as starcraft and dota.[a4] there is bigger alignment with robotics.[a5] there are regular/ frequent announcements of major shifts/ innovations in methods of increasing sophistication/ scope.[a6]

closely tied into ethics analysis (covered heavily last blog) there is more serious/ heightened prognostication/ consternation of the future implications.[b] the tools/ chips hardware are advancing noticeably.[d][d2] there is top competition for/ jockeying among leaders in the field.[f] the sophia robot made a lot of headlines earlier in the year “for better or worse”.[i]

as noted in prior blog it looks to me like AI/ ML is in a sensitive/ vulnerable/ difficult/ at times awkward transition period of inching closer to AGI capabilities.

Continue reading


top AGI leads 2018½

hi all, earlier this year released a new theory of AGI on this blog. it got substantial/ gratifying hits and am still pursuing it. aligned with this work, did a massive survey of existing state-of-the-art AGI leads. my initial idea was to try to summarize/ survey the different approaches. still have that in mind but its a almost herculean task and too much to bite off at this moment. this blog gets respectable hits but the audience is very spread out and not vocal/ participatory, dampening some of my energy for that high effort currently (but also not ruling it out).

however, this is a massive step in that direction, just painstakingly collecting this large ~180 link set/ collections of top leads.

much of this was found via the MIT AGI slack channel. its like trying to keep up with a firehose, but its very lively and cutting edge and also with tons of noise. as an expr goes used in this blog on various occasions, not for the fainthearted!

in compiling this its striking to me how both/ simultaneously brilliant and obscure some of these approaches are. some seem to me to be very much getting at the heart of AGI (and realized they are closely aligned with my own) but like my own audience, there is a lot of scattering. so far there is fairly low coalescing/ coalescence of groups around common themes/ consensus. my feeling is this disconnection may fall dramatically in the coming years esp with widely known/ publicized breakthrough(s) that drive the currently somewhat meandering herd down much more specific directions. it will be challenging-to-difficult but not inconceivable, exactly that happened on a substantial scale with deep learning within the last few years.

while it may seem overwhelming/ insurmountable at times, in some ways the AGI problem purely reduces to an architecture/ coordination problem, aka engineering. and notice some groups are arriving at the same answer from different directions (mainly psychology, (neuro)biology, machine learning, statistics/ data science/ big data, education/ learning theory, robotics, game AI, etc), with different languages/ vocabularies/ terminologies/ paradigms that are showing some/ early signs of converging/ convergence.

with new technologies, its all about “traction + momentum”. within the next few years, am expecting some major strategy/ consensus to emerge that builds on deep learning that gives rise to a plausible path/ route to AGI. have already outlined it myself, and think my ideas are close to the “secret sauce”, but my influence is low. fully expect nearly the exact same ideas to gain major traction but when espoused by some other leading light/ monolithic authority in the field, either an organization or individual or some combination of the two. it will likely be in the form of some step from the following ideas toward the more specific/ “laser-focused” direction.

odds are if there is some major AGI theory circulating at the present time, its pointed to in these refs, the well-known and not-so-well-known. and boldly both going out on a limb with a crystal ball, furthermore, think odds are strong that a “correct/ viable” AGI theory is in the not-too-distant future/ intermediate horizon and that the seeds will be contained (“holographic like”) in at least some refs cited here, maybe even many.

Continue reading

AI 2017 part 2 highlights/ trends

hi all. 2nd half AI in 2017 was very fast paced and as usual its hard to keep up. many headlines on the subj. this is a 2nd half review.

seems like some of the biggest news is the increasing dominance of China in the area after the govt announced major initiatives, funding, and resolve/ determination.[b]

google/ deepmind continues to dominate headlines.[a] the other big news is googles alphago and alphazero that no longer required human training games![a2] officially its under the name of “reinforcement learning” but my feeling is that this technology is something like “directed learning” where the AI is manipulating its environment to “detect/ seek/ digest novelty,” and that this will be a key, paradigm shifting trend in the near future. the same algorithm also works in chess/ go!

starcraft[a3] and Dota are starting to show up in AI engines/ research and OpenAI just announced breakthru championship level play vs humans in Dota, and Deepmind is attacking starcraft. despite all this major domination, it looks like games will not move out of cutting edge research anytime soon.[c4]

another notable shift/ trend is that AI/ machine learning is starting to couple stronger with physics and robotics. eg Deepmind humanoid walking simulations etc.

Continue reading

blade runner 2049 + ghost in shell

hi all. blade runner 2049 came out a few mos ago, liked it a lot. it did medium business at the box office worldwide and was something of a domestic underperformer. hey, the 1st one was too right?

am a neon collector and got some of the idea from this movie. isnt it neat that neon has a nearly 1-century long history at this point, and showed up in a big way in a scifi movie. a motif for the future.

one of the highlights for me of the new movie was the las vegas exterior desert scenes. amazing! mesmerizing!

another big highlight for me was the virtual gf character “joi” played by ana de armas. it was said in 1 of the original reviews that “the cyborgs/ replicants are more human than the human characters.” was thinking that about joi. it seemed she had more empathy/ emotional intelligence than anyone else in the movie.

so here is maybe some small )( discrepancy between geek taste and mass cultural taste. think lots of the movie was geek nirvana.

theres a scene where the blade runner buys a emanator instead of a emulator (the latter of which is a decades old real software engr technology). just a few letters make all the difference. the emanator gives joi a 3d body. sort of like dynamic 3d printing? seems completely physically impossible but thankfully that never stopped a great scifi concept.

Continue reading

secret/ blueprint/ path to AGI: novelty detection/ seeking

hi all. kurzweil wrote in 2006 “the singularity is near”. foreboding words! but today, still maybe more of a feeling than a fact. definitely, the AI field has started to mature into a new steady advance period/ era in the last few years, also with a burst of energy/ enthusiasm/ innovation heralded by the Google Deepmind acquisition in 2014, and other massive shifts toward increased investment by large corporations and govts. the Musk Open AI initiative was announced in 2015.

the other massive milestone is the ready conquering of Go by Google in 2016 by AlphaGo. in late 2017, a new version AlphaZero was announced that plays superior to AlphaGo (at “beyond human grandmaster level”) after learning merely from the rules and reinforcement learning, ie no example human-level play presented as training whatsoever. AlphaZero also plays grandmaster level chess after learning “from scratch”. this breakthrough is not fully/ widely appreciated in some ways. it is the first case of a potentially more general algorithm for AI emerging from “previously relatively narrow” study of AI in games.

AI has the terminology “weak AI” and “strong AI” for different levels/ sophistication/ “ability”. more recently the term AGI, Artificial General Intelligence has been coined.

Continue reading