The New York Tech Meetup RoundUp Take 2

September is Education Month.  Which is awesome because I am heading back to my last quarter of actual class this September if all goes well.  Scary times for all.

First up, GreenDot, which is releasing their first IPhone application.

The technology they are using is extremely interesting.  They are using motion capture technology to capture the speed and intensity of body language and movement. And now they’ve put in on an iPhone app that is about to be socially networked. The dots manage to hide your features, but not what you are doing.  Now we can share our craziness!  They’ve managed to find a new way to share x-rated content, and us flying into walls, without it looking like we are doing that!

I disapprove, primarily because I am slightly a prude in public (or not).  My art side says that is really cool.  It also knows that the vast majority of people I meet will never use this technology with what we call, intent.  To explain what intent is I’ll give you the following example:

So I know how do basic movie capture and video editing.  I was discussing projects with one of my instructors for this upcoming academic year.  The project was as follows.  I was to get into a genetics lab and film what is going on.  I was to also film myself giving a breast exam in a steamy shower, with a view from the mirror, in which you cannot see my face.  I would then use two projectors to show both videos, after adjusting the raw cuts and tones that I had.  They would purposely not line up.  The reason for this is that BRAC1 and BRAC2 are patented.  Therefore, the genetic tests for them are patented.  It would make an interesting art piece to show the sexuality of breast self examination versus the sterility of genetic examination in video knowing that for some people, the test to screen for genetics likelihood is closed off (not to mention that what would happen if you came up positive- from an insurance standpoint).  The amount and kind of nudity there is intent.  It’s not graphic nudity for the sake of being graphic (unless that’s your point as in the case of John Currin) .  It’s there because it is supposed to be there.  They would have made really nice extreme contrasts, especially once you added filters.

I disapprove of their current plan because most people lack intent in their lives.  Granted most people are given stuff and never said “Have intent, for what you do may have permanence in this world.”  I still do not think that is an excuse.  This technology was originally conceived to help understand body movement as a way of understanding body language.  Moving it to the field of fun seems disappointing, especially as the presenter was talking about using these sorts of technology for seizure detection.  Leaving it purely in the hands of fun may mean that some of the more fruitful and valuable contributions might never happen.  Of course, that’s the paranoia side of me.  And the side that says “Stop with the IPhone apps already”

The other problem with the lack of intent, is it shows a lack of being thought out.  There is no why you are starting this plan.  It harnesses the power of social networks, to be sure, but to what end?  As a result, it immediately cripples itself because it has no long term purpose (even if the purpose is fun, that should be stated.)  So it immediately brings up questions of “What is the purpose of that thing” while allowing people to be stupid with that thing in public.  Do you really want to be responsible for stupidity?  See, I don’t.  Finding intent solves that problem.  Act with intent!

Next up was CuZero, an image search engine for large content libraries.  Right now it is really, errr, raw.  Right now, it is a purely powerful form of image search.  It allows you to weigh the tags and the search image (object detectors), and therefore allow you to search more efficiently.  The UI sucks, however this is not my concern.  Right now, this is raw, really raw, and need to be refined.  One of the main critiques that I am going to lobby at CuZero is that it relies to heavily on Semantic Tagging Technology.  Since that technology is still being built into the pipes of internet, let alone other related technologies, , I feel like I am looking at something that needs to sit and wait.

Further, CuZero is straight on correct that this is a great product both on a consumer and pro-level, if the correct tagging was there.  The problem is, it isn’t.  One of the early thoughts mentioned in the presentation was selling the system to news organizations and curatorial institutes, (who have large amounts of images to curate and deal with image databases).  This is inherently problematic, since right now these images are not tagged in any way for this engine to be used.  One of the better questions for CuZero to be asking is how do we auto-tag, so that we can sell the search engine?

Attacking already existing images is going to be a difficult project.  I suggest not looking at that for now (it’s difficult, and looking at the more difficult project in all of its largeness is not going to help, you might want to start with a step down and then step into the larger project.).  I would suggest starting opening with talks with the likes of Hasselblad, Canon, Nikon, etc, to insert raw metadata for your system into the photographs while they are still in pre-processed RAW format, especially at the high end pro level, for your system.  Considering the sheer amount of data these cameras are capturing anyway, adding a small amount of sematic data that can be easily transferred from format to format, should be a no-brainer.  then people will really need a powerful image sorting and searching search engine for their databases of now newly tagged images.  Further, there will now be a standard way of tagging images that can be back-applied to previous images.   This will make life a whole lot either selling a search technology.

Musically Intelligent Machines was by far my favorite presentation of the night.  It was a clearly cohesive technology.  It has many potential business vectors, some of which are being explored already.  It has a great website, with a game tricking people into doing the work it needs to function for it.  All it needs is a strong company to surround it to open up the API, and more people to open up the business.  The presenter sounded clearly focused (and congraduatlions on presenting your thesis, I know how hard that is).  It’s well on its way.  It was sort of upsetting to only see one presentation that got that far.

Essentially, Musically Intelligent Machines has a machine tag music that it is reading off blogs, ala the Hype Machine and We Are Hunted.  It then has a few human volunteers, using a game format, to randomly rate some of the music samples.  Using those ratings, it can now generate using even more tags it generates and samples a complete set of rankings of most to least type of that tag.  The original plan was to generate lists of current (and I suppose long term, not so current) music rankings and types for TV and Movie production, so that when the music is cued, its mood is right based of the technical specifications asked for by the productions studios without very much tweaking.

Another vector that came up was the fact that is essentially an autogenerating Pandora.  Technically, what they could do (especially if they open up the API and cut a deal with advertisers), is use the game as a form of Pandora competition, since it technically would get better all the time.  It’s another vector for a business that MIM could take.

Right now, the only thing that is hampering musically intelligent machines (and really hampering it), is that it scans the same way the Hype Machine and We Are Hunted scans.  As a result, it is leaving out thousands upon thousands of genres and types of songs,especially in the jazz, classical, and international areas, which are not as likely to be blogged about, especially not in English.  Here is a good example of what I mean:

William Byrd, Vigilante, sung by the Tallis Scholars. written circa 1600.

<object width=”425″ height=”344″><param name=”movie” value=”http://www.youtube.com/v/uo9OnbLLnfE&hl=en&fs=1&color1=0x2b405b&color2=0x6b8ab6″></param><param name=”allowFullScreen” value=”true”></param><param name=”allowscriptaccess” value=”always”></param><embed src=”http://www.youtube.com/v/uo9OnbLLnfE&hl=en&fs=1&color1=0x2b405b&color2=0x6b8ab6″ type=”application/x-shockwave-flash” allowscriptaccess=”always” allowfullscreen=”true” width=”425″ height=”344″></embed></object>

Versus

(Note, you may want to look away while listening to this , it contains graphic images from September 11th, which are heightened by the music.  The Music was commissioned in honor of those who died.)

On the Transmigration of Souls, John Adams, (Selection) written 2002.

<object width=”425″ height=”344″><param name=”movie” value=”http://www.youtube.com/v/S6nrJ3ByzzE&hl=en&fs=1&color1=0x2b405b&color2=0x6b8ab6″></param><param name=”allowFullScreen” value=”true”></param><param name=”allowscriptaccess” value=”always”></param><embed src=”http://www.youtube.com/v/S6nrJ3ByzzE&hl=en&fs=1&color1=0x2b405b&color2=0x6b8ab6″ type=”application/x-shockwave-flash” allowscriptaccess=”always” allowfullscreen=”true” width=”425″ height=”344″></embed></object>

Both of these are considered classical choral pieces, from two very different periods of history.  Musically Intelligent Machine is unlikely to pick up on either, since they are rarely talked about. Due to their complexities, if they were to be used in movies (and that is MIM original business plan), one would assume that they would both need to be tagged.  Currently, the tag archive is not broad enough to fully classify these sorts of pieces.  (or rather, I could potentially say, human language might not be broad enough, and you will have cross category dilution).  Right now, one of the reasons Musically Intelligent Machine functions so well is that in fact it represents only s small crosshair of music out there.  I wonder how it will perform as it scales.  Even though I find it brilliant.

There was only one more presentation left.  ‘Teaching Robots to See”  I had shown that section of the recorded LiveCast (you go LiveCast); the reaction was “My Roommate in college built a simpler version of that.”

It doesn’t help that a quick google search shows this as the funny/quirky/cool presentation shown around the country to people (aka the “donor” or “fundraising” presentation).  That’s disappointing.  It’s removes some of the coolness.

Essentially, if one has heard about how cameras can now recognize dummy images of human faces; this is preprogramming for dummy images of objects.  Or to move because of vague ideas about shapes in its way based on cameras.  Etc.

I decided immediately that because of actually initially bad reactions to the autofocus feature and facial recognition software of these cameras (can be annoying if you don’t want the person to be the point of focus), this technology seems sort of useless that way.  it could be useful if you want your car to drive for you.  But that’s a luxury feature.

It needs an immediate feedback and payoff.  Why else would you go and show all of the #NYTM this technology that can recognizing multiple objects from  multiple angles.

And then it came to me- Arthroscopes.  Really really advanced arthroscopes, with 3d CAT and MRI imagining already plugged in presurgery.  That would work well.   So stick ‘em in Arthroscopes.  If you need the engineer and the programmer – I actually know two guys who can fill those roles if someone wants to make it happen. (One in Chicago, and one in New York).  Just Sayin’  (the only reason this came to mind, I had arthroscopic surgery when I was younger…)

(and a plug for MyTeacherMyHero, who practiced good twitter practices at the #NYTM.  I believe in replying to the virtual conversation as well as the real life one.)

(Did you know this comes in at around 2000 words?)

Related articles by Zemanta

• New York Tech Meetup (webmetricsguru.com)
• Mini NYTM (marketing.fm)