Apparently the author (Alex) responded very positively to my posting/critique. See his answer here: http://www.science20.com/physics_foundations/blog/numenta_the_bruce_lee_of_ai_and_deep_learning-154801#comment-185209
On Tue, Apr 14, 2015 at 1:40 PM, cogmission (David Ray) < [email protected]> wrote: > Here's my response to another article: > http://www.science20.com/physics_foundations/blog/numenta_the_bruce_lee_of_ai_and_deep_learning-154801#comment-185209 > > @Alex, > > Numenta doesn't do deep learning in any sense of the word. The two > approaches are distinct. Numenta's algorithms are true recreations of > neurological processes and are (hate to say it so explicitly) superior to > processes utilizing simple network nodes. Deep learning has taken us a > certain distance and is very useful up to a point - I don't mean to really > "rank" any particular approach, however, Numenta's algorithms are a quantum > leap ahead of what's currently happening with network nodes. Mind you, > Numenta is just getting started so its like comparing analog watches made > from cogs and wheels to unfinished circuitry comprising a watch circuit > board made from discrete digital components. Right now the "cogs and > wheels" are getting superior results, but when the digital circuit board is > finished - the shift in result quality will not even be comparable in > scale! Metaphorically, it won't matter how good the cogs and wheels are, > and within how many milliseconds the accuracy can be controlled - when > Numenta is finished, its scale will be in the nano-seconds and unreachable > by the "cogs and wheels" technology. > > Read more: > http://www.science20.com/physics_foundations/blog/numenta_the_bruce_lee_of_ai_and_deep_learning-154801#ixzz3XJIPHAn5 > > On Fri, Apr 10, 2015 at 11:08 AM, cogmission (David Ray) < > [email protected]> wrote: > >> Here's another article featuring Subutai! Nicely done! >> http://phys.org/news/2015-04-ibm-peers-numenta-machine-intelligence.html >> >> On Fri, Apr 10, 2015 at 6:07 AM, cogmission (David Ray) < >> [email protected]> wrote: >> >>> Hey Fergal! >>> >>> Where have you been? Thanks for the links, and specifically pointing to >>> Jeff's talk at Sandia. >>> >>> On Fri, Apr 10, 2015 at 4:15 AM, Fergal Byrne < >>> [email protected]> wrote: >>> >>>> This (partially rehashed) piece [1] is a bit better. >>>> >>>> [1] >>>> http://phys.org/news/2015-04-ibm-peers-numenta-machine-intelligence.html >>>> >>>> On Fri, Apr 10, 2015 at 10:11 AM, Fergal Byrne < >>>> [email protected]> wrote: >>>> >>>>> +1 to that idea. >>>>> >>>>> Someone started this thread [1] on Reddit a few hours ago. For once, >>>>> the conversation is sane and informed. >>>>> >>>>> [1] >>>>> http://www.reddit.com/r/MachineLearning/comments/321qtg/what_is_ibm_looking_for_in_numenta/ >>>>> >>>>> Eric and I have already contributed comments, feel free to join in. >>>>> >>>>> Regards, >>>>> >>>>> Fergal Byrne >>>>> >>>>> On Fri, Apr 10, 2015 at 5:38 AM, Matthew Taylor <[email protected]> >>>>> wrote: >>>>> >>>>>> I'll bring it up, thanks for the suggestion. >>>>>> >>>>>> --------- >>>>>> Matt Taylor >>>>>> OS Community Flag-Bearer >>>>>> Numenta >>>>>> >>>>>> >>>>>> On Thu, Apr 9, 2015 at 8:56 PM, cogmission (David Ray) >>>>>> <[email protected]> wrote: >>>>>> > @rhyolight - David's suggestion sounds good to me? What do you >>>>>> think? >>>>>> > >>>>>> > On Thu, Apr 9, 2015 at 9:11 PM, David Wood <[email protected]> >>>>>> wrote: >>>>>> >> >>>>>> >> Reaching out to both journalists with an offer of a fresh >>>>>> interview might >>>>>> >> be a good idea. >>>>>> >> >>>>>> >> Regards, >>>>>> >> Dave >>>>>> >> -- >>>>>> >> http://about.me/david_wood >>>>>> >> >>>>>> >> >>>>>> >> >>>>>> >> On Apr 9, 2015, at 12:49, Robert Smith <[email protected]> >>>>>> wrote: >>>>>> >> >>>>>> >> David, >>>>>> >> >>>>>> >> >Very interesting. It still amazes me that articles can be written >>>>>> without >>>>>> >> > much >>>>>> >> >knowledge of the subject matter (re: statements that assess >>>>>> overall >>>>>> >> > status >>>>>> >> >and quality comparisons). >>>>>> >> >>>>>> >> As the second article that popped up ( >>>>>> >> http://www.engadget.com/2015/04/09/ibm-numenta-ai-software/ ) >>>>>> seems to just >>>>>> >> be a rehash of the first with no independent research, I guess you >>>>>> will have >>>>>> >> to be ready to be amazed again. >>>>>> >> >>>>>> >> On 8 April 2015 at 17:24, David Ray <[email protected]> >>>>>> wrote: >>>>>> >>> >>>>>> >>> Thanks Robert! >>>>>> >>> >>>>>> >>> Very interesting. It still amazes me that articles can be written >>>>>> without >>>>>> >>> much knowledge of the subject matter (re: statements that assess >>>>>> overall >>>>>> >>> status and quality comparisons). >>>>>> >>> >>>>>> >>> For instance: >>>>>> >>> >>>>>> >>> Numenta’s algorithms also operate in a network, but they are >>>>>> aimed at >>>>>> >>> faithfully recreating the behavior of repeating circuits of >>>>>> roughly 100 >>>>>> >>> neurons found in the outer layer of the brain called the >>>>>> neocortex. >>>>>> >>> >>>>>> >>> >>>>>> >>> The number "100" was clearly pulled out of thin air. >>>>>> >>> >>>>>> >>> And: >>>>>> >>> >>>>>> >>> Marcus says Hawkins’s algorithms mimic only some of the known >>>>>> mechanisms >>>>>> >>> at work in the brain, and that the majority of its function still >>>>>> remains a >>>>>> >>> mystery. Demonstrations of Numenta’s technology have so far been >>>>>> limited, he >>>>>> >>> adds. “I haven’t seen them try to handle natural language >>>>>> understanding or >>>>>> >>> even produce state-of-the-art results in image recognition,” he >>>>>> says. >>>>>> >>> >>>>>> >>> >>>>>> >>> Obviously this is also true of industry pundits also. Nobody >>>>>> mentions the >>>>>> >>> vast superiority of pre-CLA work on visual hierarchy - (a demo >>>>>> recently >>>>>> >>> shown by Matt Taylor of earlier technology where the tech could >>>>>> distinguish >>>>>> >>> many distinctions with high granularity such as different objects >>>>>> moving >>>>>> >>> into and out of visual fields; moving behind other objects; >>>>>> distinguishing >>>>>> >>> partially occluded objects etc.) >>>>>> >>> >>>>>> >>> Well, at least in this case, any news is good news! :-) >>>>>> >>> >>>>>> >>> >>>>>> >>> Sent from my iPhone >>>>>> >>> >>>>>> >>> On Apr 8, 2015, at 10:52 AM, Robert Smith < >>>>>> [email protected]> >>>>>> >>> wrote: >>>>>> >>> >>>>>> >>> I've had a google alert set for "numenta" for a few weeks now. >>>>>> Today it >>>>>> >>> spoke to me for the first time. It said: >>>>>> >>> >>>>>> http://www.technologyreview.com/news/536326/ibm-tests-mobile-computing-pioneers-controversial-brain-algorithms/ >>>>>> >> >>>>>> >> >>>>>> >> >>>>>> > >>>>>> > >>>>>> > >>>>>> > -- >>>>>> > With kind regards, >>>>>> > >>>>>> > David Ray >>>>>> > Java Solutions Architect >>>>>> > >>>>>> > cortical.io >>>>>> > Sponsor of: HTM.java >>>>>> > >>>>>> > [email protected] >>>>>> > http://cortical.io >>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> >>>>> Fergal Byrne, Brenter IT >>>>> >>>>> http://inbits.com - Better Living through Thoughtful Technology >>>>> http://ie.linkedin.com/in/fergbyrne/ - https://github.com/fergalbyrne >>>>> >>>>> Founder of Clortex: HTM in Clojure - >>>>> https://github.com/nupic-community/clortex >>>>> >>>>> Author, Real Machine Intelligence with Clortex and NuPIC >>>>> Read for free or buy the book at https://leanpub.com/realsmartmachines >>>>> >>>>> Speaking on Clortex and HTM/CLA at euroClojure Krakow, June 2014: >>>>> http://euroclojure.com/2014/ >>>>> and at LambdaJam Chicago, July 2014: http://www.lambdajam.com >>>>> >>>>> e:[email protected] t:+353 83 4214179 >>>>> Join the quest for Machine Intelligence at http://numenta.org >>>>> Formerly of Adnet [email protected] http://www.adnet.ie >>>>> >>>> >>>> >>>> >>>> -- >>>> >>>> Fergal Byrne, Brenter IT >>>> >>>> http://inbits.com - Better Living through Thoughtful Technology >>>> http://ie.linkedin.com/in/fergbyrne/ - https://github.com/fergalbyrne >>>> >>>> Founder of Clortex: HTM in Clojure - >>>> https://github.com/nupic-community/clortex >>>> >>>> Author, Real Machine Intelligence with Clortex and NuPIC >>>> Read for free or buy the book at https://leanpub.com/realsmartmachines >>>> >>>> Speaking on Clortex and HTM/CLA at euroClojure Krakow, June 2014: >>>> http://euroclojure.com/2014/ >>>> and at LambdaJam Chicago, July 2014: http://www.lambdajam.com >>>> >>>> e:[email protected] t:+353 83 4214179 >>>> Join the quest for Machine Intelligence at http://numenta.org >>>> Formerly of Adnet [email protected] http://www.adnet.ie >>>> >>> >>> >>> >>> -- >>> *With kind regards,* >>> >>> David Ray >>> Java Solutions Architect >>> >>> *cortical.io <http://cortical.io/>* >>> Sponsor of: HTM.java <https://github.com/numenta/htm.java> >>> >>> [email protected] >>> http://cortical.io >>> >> >> >> >> -- >> *With kind regards,* >> >> David Ray >> Java Solutions Architect >> >> *cortical.io <http://cortical.io/>* >> Sponsor of: HTM.java <https://github.com/numenta/htm.java> >> >> [email protected] >> http://cortical.io >> > > > > -- > *With kind regards,* > > David Ray > Java Solutions Architect > > *cortical.io <http://cortical.io/>* > Sponsor of: HTM.java <https://github.com/numenta/htm.java> > > [email protected] > http://cortical.io > -- *With kind regards,* David Ray Java Solutions Architect *cortical.io <http://cortical.io/>* Sponsor of: HTM.java <https://github.com/numenta/htm.java> [email protected] http://cortical.io
