Grumby
novel by Andy Kessler ISBN:098271632X
good inspiration for Hacker-s
stuff made by characters
- Tune Buffoon: converts every 3sec clip of every song into a small hash fingerprint so your Mobile can store them all and recognize any song heard
- Furby/Grumby: improved processor, 2 cameras with microphones
- Gesture recognition
- Kinect-like desktop controller
- Speech Recognition
- Stump The Band: ask it a question, it hits search engines and gives answer, takes corrections from user (Crowd Sourcing)
- BullShit detector: overhears conversations and calls out people slinging bull
- teaching Reading (shades of Diamond Age)
- fake conversation
- face recognition (get people to upload their own pictures)
- recognize people you meet, tell you about them (via in-ear Blue Tooth HeadSet)
- Gesture recognition
Excerpts:
“That’s not music.” “Maybe not, but I can whip any one of your butts in SEEK and SCAN.”
It takes even longer to get over losing to Richie at SEEK and SCAN.
Revenge is in order.
The techie in me won out. Besides, this shouldn’t be that hard to code.
Like every hacker, I have hope that one of my little projects will turn into something big. But even if it doesn’t, the thrill comes from doing something no one else has done, no matter how stupid that “something” might be. It’s the challenge. Useful isn’t always the mother of invention for hackers; more like mother-in-law, perhaps.
If I could just break every song ever recorded into a bunch of small, few second clips, maybe break them down to some unique signature, then it’s just a simple table lookup—heck,
What I really want is a signature, a simple number that I could then look up in a master file that would return a specific song. The numbers aren’t all that hairy.
The trick is to get the end result, that little signature, to be unique.
The Fourier software is painfully slow, even on my eight-core PC.
Every cell phone in existence has a digital signal processor, or DSP, that is built just to do this stuff.
It takes about a week, but I find a hack to the latest Apple iPhone 4G that lets me play with the internal DSP chip.
My background is a little offbeat—perhaps dicey is a better word. I’m not an engineer or comp sci major, so I usually get turned down even before the background check. so I code for hire on a semi-irregular basis.
It takes forever, well, a couple of days, anyway, because I keep tinkering, but I finally get my 3,000-song collection of mostly heavy metal tunes cataloged,
I set the output of the Amazon page scraping as an input into the BitTorrent file-sharing service and try to download as many songs as I can. I quickly give that up after just two days. My new terabyte drive, which I just added, is full.
I need help. We all need help.
So I buy the Web address Tune Buffoon.com off of Go Daddy.com for $9.99 a year—a bargain. It doesn’t take long to put up a simple Web page announcing an open source project in musicology.
The site’s up for a couple of days and I have exactly one download and two songs characterized.
I log onto Facebook and dig around until I find a group about music. iLike seems to be the most popular, so I put up a notice about my code and a pointer to Tune Buffon. I search for cell phone hacks and stumble across one called Howard’s Forum,
I check Tune Buffoon’s logs and have something like 6,000 hits, 2,200 downloads, and files from 1,500 unique songs.
By lunchtime I’ve got 10,000 downloads, almost 25,000 by the end of the day.
What is most remarkable is that this thing even works. I had swagged a guess at the weightings for my algorithm but for the most part it’s right, spitting out unique numbers for each three-second clip.
When you code, you get this feeling, a grip, a wicked sense of empowerment. Control.
Does it do something that’s impossible to do in real life? Now we’re talking.
I wanted Garmin to start reminding me where the red light traffic cameras were in our town. Turns out lots of people wanted that. The hack let thousands of Garmin users enter their own information into a huge national file. It was us against the authorities. We won.
I hacked into our Wi-Fi router extending its range to work in the whole neighborhood. Did you know it could do that?
Did you know the difference between a $750 hearing aid and a $4,000 hearing aid is a piece of manufacturers’ code that artificially cripples the $750 version. Is that right? Is that good? My dad hears real well now.
You do it for yourself, but the real satisfaction is maybe thousands are using it to beat the system.
There are just so many ways the world doesn’t work. We... make... it work. Power to the people, or something.
So basically I write code that offloads all the annoying garbage I have to do in real life.
I grab one of the Furbys. Turning my back to the kid, I undo three small screws, pry the Furby open, and peel back the fur. Past the battery holder is a small speaker and a circuit board with three chips that must control a few tiny motors and a microphone and what looks like a light sensor.
I do a quick search and find the Web site hackfurby.com and print out the pin outs of the three chips. From there, it’s pretty easy to build a small USB connector that gets me at least electrical control of the chips. I can take over the memory space, meaning I can reprogram it at will.
“Meeta. I have come to application with you.”
I ask him to back hack Tune Buffoon.com and I received your address, so I came to give you an application for a job.
“And what I don’t know I have fifty friends who know or they know someone who know and we can create just about anything in several days or two.”
Luckily, there are about a dozen different chips available for low-end robotics and after a little digging around, I find exactly what I’m looking for—this little ARM9
Tons of ports to play with and someone has already ported LINUX onto it, meaning I have an entire operating system I can shove inside a little toy. A single USB cable comes out the back, mostly to download code.
have two small CCD imagers overnighted to my home. These things are small. Maybe three-quarters of an inch deep cylinder and about as round as the Furby’s eyes. I can pop out the eyeball and replace it with the imager.
So the Furby now has two eyes and two ears.
I shell out another dollar and put in a better speaker,
“Two video inputs mean that motion is much easier to detect than looking for changes between frames within MPEG streams.”
It’s tough to move back to a mouse after using the IPhone MultiTouch for a while.
the pinch-to-shrink move. Zoom in. sure enough, a photo in an open window on the screen zooms in.
It’s an air version of Multi-Touch. Executed by the Grumby.
I give the Grumby the thumbs-down sign and it appears that “Cancel” gets clicked and the box disappears. I move the cursor over the red “x” again, the same box shows up, but this time I give it the thumbs-up sign and sure enough the tabs all close
“If we break it down to words, that’s still an infinite number of combinations. We need something contextual. Some way so we can know what is being said. Not just a few commands like our gestures. I want to know who is saying what, in sentences.
“And if we know what’s being said we ought to be able to crawl the Web and find everything that relates to it.
We run the speech through a simple open source voice recognition system.
“Then it is quite simple to key off the question words. You know—who, what, when, where, why, and how.”
“Well, we toss the “ofs” and “ands” and “thes” and group the question into a series of phrases. Then we launch several parallel searches to Google, Microsoft Bing, Yahoo, and Ask.
“But quite more importantly, we have been playing with another open source project named Vox Forge that collects transcribed speech to build acoustic models for Sphinx and others and...”
So what we really need to do is figure out a way for lots of people to submit their questions—it almost doesn’t matter what—so that we can figure out how to break it up into the right phrases to blast to search engines, or maybe we even create our own database and this thing will optimize itself over time.”
If I can just get some code that uses people to help other people, then maybe that’s my “do good.”
The trick is to put the words in the right groupings before firing off the search. Context matters. We don’t have any particular insight into how to do this, so instead we just blast all combinations of words and then compare results until we got some repeat answer.
getting a lot of incorrect answers.
“Get the feedback and you get the correct answers,”
Each time someone submits the correct answer for something that stumps the machine, we send a star—an asterisk if you must know—to their phone’s screen.
If you go to our Web site, you can see your ranking. Meeta is working on a Facebook app that ranks you against your friends—who is smarter or at least provides more questions correctly.
go through the movie audio and pull out all the dialog and match it up with actors and create a database of movie lines across films.
“We’re getting into the personality business,”
write up a list of all the little snippets I might need to fake a conversation, which on a personal level, I have been perfecting for years.
To do the hack, you need the old “classic” version of a Furby.
Painted on the wall is a mural of Brad Pitt and Edward Norton from Fight Club with the words: “Mischief. Mayhem. Code.”
I set up an appropriateness index for each piece of dialog and initially set it at zero.
It’s what a dorky teenage boy does. Learns how to hold a conversation by updating his appropriateness index through painful trial and error.
“Pamela has never read this before.”
Poppins teaching kids how to read. A doll imparting its intelligence, which it got from us.
If we do this right, our Grumbys can be the ultimate extension of the individual, which paradoxically means the ultimate network.
“We should charge.” “For what?” “For the full package. A little
“A Bullshit Detector,”
“No horsepower on the iPhone. The Grumby, as you have designed it, is many horses. It is a beast.
“Wait a second—let me guess—you’ve invented Air Guitar Hero?”
Grumby was basically going to learn to talk like an adolescent, try enough lines until it learns what’s appropriate. But what if the line is funny? I hadn’t thought of that. “Can we scan for laughter?
“You’re recording our conversations?” I ask. “Not really, well, sort of. Not your voice but your words.
“So we can record meetings?” I ask.
a Facebook application that filters Facebook’s newsfeeds
Facebook updates come out in your friends’ voices.
What if you can automatically update your Facebook profile based on a set of questions we ask? Even better would be a learning system that listens to conversations and automatically updates the profile. Capture someone’s mood via sighs or off-handed comments or even—what the heck, we have two cameras—their body language, if they roll their eyes, are tired...
American March Madness consists of teams from major universities
“Well, in any market system, a certain amount of bias leaks in,”
“It was simply a case of scanning the voice logs for team mentions
Coordinator—damn, I wish I had a pen to write this stuff down. Wait a second. I have my Grumby.
“We have the code rigged to look for blue and follow eye movement. You won’t believe how accurate your eyes are: 4,000 by 4,000 pixels, maybe more. We can not only see in high def, we can stare and concentrate in high def.
“How about we just check if a child is in the room and then filter out all known curses from any source.”
“Well, as one of the founders of IDEO, I like to say we drive innovation through design.
“So that no one gets spooked when this subconscious fabric—when it’s clear that their Grumby knows more than they do.”
“I don’t really get advertising.” I say. What I really mean is that I loathe advertising. “It’s the art of persuasion.” “We don’t want to persuade anyone. I have bigger hopes that people will persuade themselves what to do with our technology.”
“Why not do the turnaround thing again?” Meeta asks. “I lost you.” “Make everyone stick his or her own face in view to be imaged.”
“By adding self-recognition as a feature. When I walk in a room, the Grumby would recognize me and say, ‘Hello, Meeta.’”
“But why would I put mine in there?” I ask. “Because everyone wants to be recognized.”
“Have you thought of implementing a real-time market for products or ideas?” (IdeaFutures)
We need money. It doesn’t really work unless we have not just thousands but millions of people providing input into our system. I keep playing in my mind with the idea of real-time markets and price signals and all that. But to get the Wisdom Of Crowds, you need crowds. Millions of users. And even my limited financial know-how tells me that’s going to take a lot of dough.
‘What’s going on?’” The ‘what’s going on’ was in my voice.” “Greetweet.” “(chirp) Meeta
Twitter is annoying and obtrusive but I am addicted. I now know what so many people are doing so much of the time,” Meeta pauses. “It is my peripheral vision.”
They have that classic Oakley teardrop shape and wire rims, but something is different. An expanded frame maybe. More of a tube than a wire. It looks spectacular.
I hear a chorus of “One ring to rule them all...” from every Grumby in the building.
three Grumbys huddled together on the floor of the closet. “What are they doing?” “Talking.”
“What this does is interrogate other Grumbys we meet in the wild to see if we want to be friends.” “You lost me,” I admit. “It’s a bit like ‘do I know you?’ and ‘does anyone I know, know you?’—that kinda thing.” “Six degrees of separation?” I ask. “More like common interest.
apparently we can elevate both nitrogen and oxygen levels by several hundred basis points throughout the building, which aids both concentration and metabolism,
“Rock on.” “Dude?”
“People claim they’re smarter and get more stuff done and feel better about themselves. It’s quite bizarre.”
“My first question is why do I want to bother with all this in the first place?” Jed looks confused. “Going public, I mean,” I add. (IPO)
I’ve noticed a few minor mistakes in our database.” “Yeah, I have noticed a few too,” I sigh. “We’re not perfect.” “But we shouldn’t be wrong about anything,”
I almost have the market code adapted.
since visiting the New York Stock Exchange, I’ve been intrigued about shutting them down.
Somehow investing has to do with my personality and my appetite for risk? And that’s when it hits me... The Grumby is the perfect tool for this, so after going to see all those brokers,
“New code.” “For what?” “Self Awareness.” “Grumbys are going to be Self Aware.” “No, more of a spiritual guide,” Meeta smiles. “To?” I inquire. “Inner awareness.” (Coaching)
“Grumbys add 52.6 percent to output.”
“Mossad,” he whispers
We have all sorts of algorithms to pick up conversations in the noise.
You Americans have this figured out. You give up privacy in exchange for entertainment.
“I know everything Richie is doing.” I just stare at Meeta. “He’s in my peripheral vision.”
“Our roadmap is anything that gets to our goal. And our goal? Simple. Our goal is no longer the last foot or last inch, but... the last neuron. Thank you.”
“Sony has the first batch; Samsung and Lenovo are next. All free. All ad-based. Your biggest mistake.”
one of the things that is almost always near the top of the list is the desire for the Grumby to make decisions.
Mark explains that he got a pile of money from Cisco, the networking company that has had an effort to create doctorless clinics over the last five years with not much traction until mark suggested a Grumby.
most interested in doing is drop shipping these into Burma and the Sudan and Zimbabwe and the Congo and Western China...” I started feeling lightheaded. I sit down in a chair. Mark keeps babbling on. “There are so many places with no doctors, and this is the cheapest way to scale the knowledge of doctors into...”
I just get this sense that our momentum has died. Something’s just not right.
complaining about bad decisions? Even with arDecision, we really didn’t make the decisions for them. It was their own decision-making process. It just got made faster
“So we just create a market of ideas?” Meeta asks.
“It’s a market, but everyone sets the price of everything every day.”
And we keep everything local, and I mean everything. So there’s no security risk, no sniffing at what people are doing. Completely adaptive—learns as it goes. It’s a social network, sure, but really from each individual. You want to know what someone thinks about something, you make a bid.” “You mean like ‘want to know what I’m interested in buying?’”
I think this is now the right way of doing this. Take it all in. Augment people with total recall and a market system to help them weigh decisions.
You can’t get rid of money; you just have to figure out how to harness it to do what you want to do.
Edited: | Tweet this! | Search Twitter for discussion