A facial recognition system like EigenFace looks for the dominant modes of variation among a population of faces to find predictors that help distinguish among the different faces. By ignore the ways that faces are more similar and accentuating the ways that faces are most different, they can craft identification algorithms. The only challenge is getting enough data to understand the full range of facial variations -- hence this solicitation for more facial data.
But I wonder if people can game this system to make themselves unrecognizable. For example, a member of the tin-foil hat brigade might submit multiple mugshots of themselves under mutliple assumed identities. By using slight variations in facial expression, makeup, lighting, and camera angle they would make the system think that a large fraction of the population "looks like" the person who seeks anonymity. The system would then have a hard time identifying the tin-foil hat wearer because he/she matches so many people in the test data.
Anyone can make themselves "look average" if they can bias the dataset that defines the average.
what's to stop me from bidding $500/share to guarantee I get to take part in the IPO?
Your fear that everyone else will bid $501, that the auction will settle at $500, and that you will end up payng $500 per share. Although it is true that a single person can submit a high bid to "ensure" getting some shares, the extending this strategy to n-players changes the game. In the end, people reallize that the most sensible strategy is to bid what they really think the shares are worth. Bidding any higher than that increases the chance of getting over-valued shares and losing money.
How much would you bid in a Dutch auction for a $20 bill?
But, you are right that this auction will bid up the price because people are not rational -- they tend to be more confident than they should be and they like to "win".
Re:All battery devices? Laptop Display Inverters
on
Build Your Own Stun Gun
·
· Score: 2, Informative
About the only way I can think of securing against such a threat would be to ban all battery-operated devices from the plane.
Yep, some backlight inverters for laptops run at over 1000V for firing a cold cathode tube.
UHF covers a massive chunk of spectrum -- from 470 MHz to 890 MHz. Even if you carve out some 18 MHz notches for local UHF channels, you still have hundreds of MHz of usable spectrum. And in rural areas, the full 420 band could be used for some serious wireless networking. With good compression/encoding and high enough SNR, multigigabit wireless might be possible.
In India, they call them gobar gas plants (more details in a 1971 Mother Earth New article). As long as one keeps the 30:1 carbon/nitrogen ratio, they can consume other organic waste too (grass clippings, urine, food waste, etc.). The only problem with them is that they tend to create hydrogen sulfide that makes the gas highly corrisive to iron equipment (some people use a filter of steel wool to remove the H2S).
My quick calc suggests that a petabyte would require about 213,000 DVD-R disks at 4.7 GB/disk. At about 1.2 mm/disk, that's a stack about 255 m (837 ft )high.
I don't even want to think about backing this up on a million some-odd CD-Rs. I suspect that the first CD-R would have rotted long before the last CD-R was written.
Hell, the most-likely nuclear terrorism scenario in my estimation is someone purchasing a radiation-therapy machine and randomly zapping people with lethal doses from inside a truck-mounted setup. Given a cool million to purchase some used medical equipment, you don't even need to try to steal nuclear material from federal facilities.
It's worse that you think. A number of years ago (maybe 10 to 20?), the radiation detectors at Los Alamos went off when a delivery of patio furniture passed by. Turns out the cast iron in the furniture contained Cobalt-60. Tracing the shipment back, they found that the furniture had been made in Mexico from scrap metal. Someone in Mexico had sold a radiation therapy machine as scrap.
One of the big changes is what Jose Mejia has done to Lucent's supply chain. The company's Customer Delivery Organization concept has helped the company connect the back-end of manufacturing and supply to the front-end sales force. This has helped the win contracts and control costs.
I'm sure Lucent faces a tough battle given that wireline connections aren't growing, wireless is becoming a commodity, and optical still faces a glut of installed dark fiber. Still, I suspect that they will be able to reap their share of contracts and profit from whatever telecom equipment sales there are.
This trend is merely a return to the past. The entire "going to the office" or "going to the factory" concept rose with the urbanization and industrialization of civilization. Go back more than a couple of hundred years and I'd bet you will find that most people had very little worklife-familylife separation. People lived on the farms that they worked on or you lived above their shop. People worked with their parents, children, and extended family. If their livelihood had a problem in the middle of the night or on the weekend, they dealt with it. That why we have so many surnames that are careers (e.g., Carpenter, Smith, Baker, Farmer, etc.)
It's not the current blurring of work and life that is a fluke, it was the recent past's separation of work and life that was the odd phenomenon.
Afterall, does a cell phone that's turned off give of any energy that can be detected?
It does not emit any RF, but it will absorb RF transmitted at it. I'd bet the RF properties of the cell phone can are very different form those of a regular can. The built-in antenna on the cell phone would react differently to RF radiation at band center frequency than at frequencies above and below the band center. A scanner that sends out pulses at various frequencies and measures the return signal might be able to detect the phone's unusual signature.
Two problems. First, the cost of developing and building the scanner might exceed the expected value of winning the SUV, especially if the scanner is not 100% reliable and you aren't in an area that gets one of these cans. Second, your local grocer migth take a dim view of someone passing a scanning wand around all the Coke displays.
linear algebra: graphics, control systems, optimization software
calculus and differential equations: physical simulations (game physics engines, analog circuits, heat transfer, etc.), analog control systems
statistics: finance, scientific programming, enterprise software, optimization software, AI
I'm sure others can think of other mappings (like crypto depends on several different math topics depending on the underlying code scheme). And if you don't know what type of programming you will be doing, then take some of everything. Math really enhances analytic skills and logical thinking.
People here should know that putting a pricetag on something doesn't make everything kosher.
A very good point, in general. Yet as an adult I feel i have the right to enter business relationships - there is nothign wrong with selling my email processing labor. As long as the consumer retains control, I see no problem with bulk e-mail. With control of the system, I can easily raise the price of spam delivery to 50 cents or a dollar per message if the 15 cents/spam is generating too much volume.
Bulk mail without opt-in should be criminalized regardless if the envelope is paper, SMTP or whatever. Bulk mail is just another form of 'I have money, I can send propaganda to anybody, you cannot stop me, muahahaha!".
I'm not sure I want the government holding my hand and deciding what is good for everyone and what is not. I don't even see how the government can regulate spam given the international nature of it and the fact that commercial email has legitimate uses such as when my airline emails me that my flight schedule has changed or tells me of upcoming airfare sales.
To me, the greatest scheme for controlling spam would be monetary -- the spammer pays the recipient an amount that the recipient decides and the sender agrees to. Add recipient-controlled whitelists, blacklists, and rebates and the system provide zero-cost email between friends and trusted parties and consumer-regulated communications otherwise. This avoids the heavy-handed, one-size fits all approach of government regulation and pays each recipient for the resources consumed by spam as judged by the recipient. If someone hates spam so much, they can set their price at $100 per email.
The big problem with the current system is that the recipient bears a disproportinate burden of the costs. The cost to send an email is miniscule. But the cost to personally accept, read and process an e-mail is large. All I seek is a means of charging for my labor.
I'd accept all the spam in the world if they paid me 15 cents per message. That would make spam much cheaper than bulk mail and would weed out marketers who aren't serious.
If a company is going to sell my resources (time spent downloading/reading/procesing email) they had better share the revenues with me.
This antenna essentially "disappears" when it is not being used, making it fairly "stealthy"
More than just disappear when not in use, it could potentially disrupt snopper while in use. While locked on to a legitimate user, the antenna could change the beam pattern in other directions to disrupt snoopers. Only a snooper that is roughly in the line from antenna to legitimate user would be able to get a consistent signal.
The article mentions the excessively large size of 35 mm lens for imaging on to small digital sensors, but misses the two additional problems with using film camera lenses with digital sensors.
Standard film camera lens tend to transmit light from the subject to the sensor at the angle that it was received (similar to the way that a pinhole camera projects a bundle of rays from object space to image space). Silicon sensors suffer from two problems when light enters them at an angle. First, the high index of the material and coatings tends to reflect the angled light -- causing less light to enter the sensor and the image to have dark corners. Second, long wavelength light penetrates the sensor deeper than does short wavelength light. If the light enters at an angle, the red photons can angle down into the substrate and actualy register in pixels further out. The result is that the red and infrared portions of the image are misregistered, causing color fringing in the corners.
The point is that the best lens for a digital camera will be different from the best lens for a film camera. A better lens design for digital cameras incorporates image-space telecentricity. Image-space telecentricity means that the light hitting the CCD is largely perpendicular to the sensor.
I really do wonder if they have got all the kinks worked out. You never know when some odd interaction of materials, manufacturing processes, and customer environments won't create reliability problems.
you are over-generalizing. Three lefts make a right only if they are right angle lefts (or if the sum of the angles of the three lefts equals the sum of three right angles).
Too true. But then so are you. If a left can be any leftward turn, then a right can be any rightward turn. Drivers in Boston and Britain will probably agree with me that there are no right angles (so will Einstein and other non-Euclideans). Lets assume that the definition of right and left is an appropriate 90 degree angle +/- 45 degrees. With such a wide range of possible turning angles as few as two lefts and as many as 5 lefts are required to make a right. And with curved space, the required number of lefts may be undefined.
Of course the true Bostonian will correct me and say that no amount of lefts make a right, because you just can't get there from here. British drivers will point to roundabouts as evidence of 2 lefts making a right. And if Britain has any cloverleaf interchanges, then one left can make a right.
Heh, heh. Karma burning time.
What's the point of Karma if you can't have a little pyromaniacal fun with it occasionally.
You have come to the end of the post. There you find an over-generalized message...;)
Ants in the Apollo 11 Crew Quarantine Module
on
Our Man In Black
·
· Score: 4, Informative
I think this person has an impossible job.
Years later, astronaut Buzz Aldrin said in a television interview that the mobile quarantine trailer in which the Apollo 11 crew was isolated had one serious flaw: Ants appeared to be going into and out of the trailer (37). If there were any Moon bugs, they would have gotten out with the ants. -- from The dilemma of Mars sample return
Add to that all the meteorites that fail to stop at the agricultural station on their way in, and I'd think the Earth is already pretty contaminated.
I'm not saying that he should not try to reduce cross-contamination, only that its not an easy job.
Weep for all the lost magnets
on
HDD Assault Cannon
·
· Score: 4, Interesting
I prefer to disassemble old HDs. The voice coils and spindle motors tend to contain insanely strong rare earth magnets. And the platters make pleasant wind chimes (especialy if you have a mix of 3.5", 5", and 8" platters). I suppose one could also get a few bucks from the cast aluminum anclosures.
I just wish that these ID systems were more secure. Instead of using easily stolen and duplicated plaintext identifiers (like an SSN and mother's maiden name), I'd like to see a secure encoded number that is unique to each application. This unique number (different each time it is asked for) would be resolvable to a single identity inside secure back-office applications or through access to a central secure server.
A smart ID card would hand-out unique numbers and log who got which ID. That way any theft of identity is traceable to the source. The card owner could then use the card to trace who was using their data.
I'm sure there are a million potential vulnerabilites with the idea, but the current approach seems much more insecure than this proposal.
Sometimes I wonder if the prison overcrowding problems aren't because they toss out 5 year sentences like candy to spammers (soon), hackers, and people who get caught with a single joint.
A sci-fi story once offered a nice solution to prison overcrowding. The convicted criminal had to take a pill drawn randomly from a bottle. A certain percentage of the pills in the bottle were loaded with lethal poison. If criminal survived the pill, they were released. The probability of dying was tied to the crime -- e.g., murder someone and you might have draw a pill from a bottle that gave a 50% chance of death, run a red light and you have a 0.1% chance of death.
A search engine that knows my browsing history could be very useful for:
Finding that page I saw last month. If the engine knew my history, I could limit the search to just pages in my history.
Finding new pages that I have never seen. This would exclude all previously seen pages from the search results. A better version would even exclude hits that had appeared in previous search hit lists. I often do multiple searches in which the Nth search finds items that I saw (and rejected) in one of previous N-1 searches.
Tracking lost pages. The engine could periodically check my bookmarks and relocate pages that had been moved (or find pages similar to the missing page). If the page is truely gone, I could use Google cache to snag an archival copy.
Automatically finding pages similar to ones that I like. If the search engine notices that I visit certain pages repeatedly (e.g.,/.), it might run a search for pages that are similar to my favorite sites.
Social networking: Finding people that have browsing histories like me.
Yes, there are some nasty privacy issues, so one needs to pick the partner carefully (as if your ISP doesn't know your browsing history). What is interesting is that services like A9 and GMail create a new level of personalization in which the massive technological scope of an Amazon or Google is put to work for individuals.
The answer also depends on the work and workplace. If it is a large IT organization that needs another warm body to crank out code designed by the company's gurus, then the less experienced (less expensive) programmer is fine.
If it's a smaller company and the "programmer" will also be responsible for software architecture, high-level design, purchasing tools, and unsupervised coding, then you want the most experienced person possible. For higher-level software engineering, you want someone with a diversified mental library of patterns, designs, and experiences.
It also depends on the code. If its for small little utilties with a short lifespan, limited userbase, and simple control flow logic, then a less experienced programmer is OK. If the code is mission-critical, high-performance, inner-loop code, then you will want the more experienced person.
Slashdot Mirror
A facial recognition system like EigenFace looks for the dominant modes of variation among a population of faces to find predictors that help distinguish among the different faces. By ignore the ways that faces are more similar and accentuating the ways that faces are most different, they can craft identification algorithms. The only challenge is getting enough data to understand the full range of facial variations -- hence this solicitation for more facial data.
But I wonder if people can game this system to make themselves unrecognizable. For example, a member of the tin-foil hat brigade might submit multiple mugshots of themselves under mutliple assumed identities. By using slight variations in facial expression, makeup, lighting, and camera angle they would make the system think that a large fraction of the population "looks like" the person who seeks anonymity. The system would then have a hard time identifying the tin-foil hat wearer because he/she matches so many people in the test data.
Anyone can make themselves "look average" if they can bias the dataset that defines the average.
what's to stop me from bidding $500/share to guarantee I get to take part in the IPO?
Your fear that everyone else will bid $501, that the auction will settle at $500, and that you will end up payng $500 per share. Although it is true that a single person can submit a high bid to "ensure" getting some shares, the extending this strategy to n-players changes the game. In the end, people reallize that the most sensible strategy is to bid what they really think the shares are worth. Bidding any higher than that increases the chance of getting over-valued shares and losing money.
How much would you bid in a Dutch auction for a $20 bill?
But, you are right that this auction will bid up the price because people are not rational -- they tend to be more confident than they should be and they like to "win".
About the only way I can think of securing against such a threat would be to ban all battery-operated devices from the plane.
Yep, some backlight inverters for laptops run at over 1000V for firing a cold cathode tube.
UHF covers a massive chunk of spectrum -- from 470 MHz to 890 MHz. Even if you carve out some 18 MHz notches for local UHF channels, you still have hundreds of MHz of usable spectrum. And in rural areas, the full 420 band could be used for some serious wireless networking. With good compression/encoding and high enough SNR, multigigabit wireless might be possible.
In India, they call them gobar gas plants (more details in a 1971 Mother Earth New article). As long as one keeps the 30:1 carbon/nitrogen ratio, they can consume other organic waste too (grass clippings, urine, food waste, etc.). The only problem with them is that they tend to create hydrogen sulfide that makes the gas highly corrisive to iron equipment (some people use a filter of steel wool to remove the H2S).
My quick calc suggests that a petabyte would require about 213,000 DVD-R disks at 4.7 GB/disk. At about 1.2 mm/disk, that's a stack about 255 m (837 ft )high.
I don't even want to think about backing this up on a million some-odd CD-Rs. I suspect that the first CD-R would have rotted long before the last CD-R was written.
Hell, the most-likely nuclear terrorism scenario in my estimation is someone purchasing a radiation-therapy machine and randomly zapping people with lethal doses from inside a truck-mounted setup. Given a cool million to purchase some used medical equipment, you don't even need to try to steal nuclear material from federal facilities.
It's worse that you think. A number of years ago (maybe 10 to 20?), the radiation detectors at Los Alamos went off when a delivery of patio furniture passed by. Turns out the cast iron in the furniture contained Cobalt-60. Tracing the shipment back, they found that the furniture had been made in Mexico from scrap metal. Someone in Mexico had sold a radiation therapy machine as scrap.
One of the big changes is what Jose Mejia has done to Lucent's supply chain. The company's Customer Delivery Organization concept has helped the company connect the back-end of manufacturing and supply to the front-end sales force. This has helped the win contracts and control costs.
I'm sure Lucent faces a tough battle given that wireline connections aren't growing, wireless is becoming a commodity, and optical still faces a glut of installed dark fiber. Still, I suspect that they will be able to reap their share of contracts and profit from whatever telecom equipment sales there are.
This trend is merely a return to the past. The entire "going to the office" or "going to the factory" concept rose with the urbanization and industrialization of civilization. Go back more than a couple of hundred years and I'd bet you will find that most people had very little worklife-familylife separation. People lived on the farms that they worked on or you lived above their shop. People worked with their parents, children, and extended family. If their livelihood had a problem in the middle of the night or on the weekend, they dealt with it. That why we have so many surnames that are careers (e.g., Carpenter, Smith, Baker, Farmer, etc.)
It's not the current blurring of work and life that is a fluke, it was the recent past's separation of work and life that was the odd phenomenon.
Afterall, does a cell phone that's turned off give of any energy that can be detected?
It does not emit any RF, but it will absorb RF transmitted at it. I'd bet the RF properties of the cell phone can are very different form those of a regular can. The built-in antenna on the cell phone would react differently to RF radiation at band center frequency than at frequencies above and below the band center. A scanner that sends out pulses at various frequencies and measures the return signal might be able to detect the phone's unusual signature.
Two problems. First, the cost of developing and building the scanner might exceed the expected value of winning the SUV, especially if the scanner is not 100% reliable and you aren't in an area that gets one of these cans. Second, your local grocer migth take a dim view of someone passing a scanning wand around all the Coke displays.
- discrete mathematics: logic-heavy algorithms (operating systems, databases, compilers, control systems, embedded systems, AI)
- trigonometry: graphics, physical simulations (game engines), mechanical engineering software
- linear algebra: graphics, control systems, optimization software
- calculus and differential equations: physical simulations (game physics engines, analog circuits, heat transfer, etc.), analog control systems
- statistics: finance, scientific programming, enterprise software, optimization software, AI
I'm sure others can think of other mappings (like crypto depends on several different math topics depending on the underlying code scheme). And if you don't know what type of programming you will be doing, then take some of everything. Math really enhances analytic skills and logical thinking.People here should know that putting a pricetag on something doesn't make everything kosher.
A very good point, in general. Yet as an adult I feel i have the right to enter business relationships - there is nothign wrong with selling my email processing labor. As long as the consumer retains control, I see no problem with bulk e-mail. With control of the system, I can easily raise the price of spam delivery to 50 cents or a dollar per message if the 15 cents/spam is generating too much volume.
Bulk mail without opt-in should be criminalized regardless if the envelope is paper, SMTP or whatever. Bulk mail is just another form of 'I have money, I can send propaganda to anybody, you cannot stop me, muahahaha!".
I'm not sure I want the government holding my hand and deciding what is good for everyone and what is not. I don't even see how the government can regulate spam given the international nature of it and the fact that commercial email has legitimate uses such as when my airline emails me that my flight schedule has changed or tells me of upcoming airfare sales.
To me, the greatest scheme for controlling spam would be monetary -- the spammer pays the recipient an amount that the recipient decides and the sender agrees to. Add recipient-controlled whitelists, blacklists, and rebates and the system provide zero-cost email between friends and trusted parties and consumer-regulated communications otherwise. This avoids the heavy-handed, one-size fits all approach of government regulation and pays each recipient for the resources consumed by spam as judged by the recipient. If someone hates spam so much, they can set their price at $100 per email.
The big problem with the current system is that the recipient bears a disproportinate burden of the costs. The cost to send an email is miniscule. But the cost to personally accept, read and process an e-mail is large. All I seek is a means of charging for my labor.
I'd accept all the spam in the world if they paid me 15 cents per message. That would make spam much cheaper than bulk mail and would weed out marketers who aren't serious.
If a company is going to sell my resources (time spent downloading/reading/procesing email) they had better share the revenues with me.
This antenna essentially "disappears" when it is not being used, making it fairly "stealthy"
More than just disappear when not in use, it could potentially disrupt snopper while in use. While locked on to a legitimate user, the antenna could change the beam pattern in other directions to disrupt snoopers. Only a snooper that is roughly in the line from antenna to legitimate user would be able to get a consistent signal.
The article mentions the excessively large size of 35 mm lens for imaging on to small digital sensors, but misses the two additional problems with using film camera lenses with digital sensors.
Standard film camera lens tend to transmit light from the subject to the sensor at the angle that it was received (similar to the way that a pinhole camera projects a bundle of rays from object space to image space). Silicon sensors suffer from two problems when light enters them at an angle. First, the high index of the material and coatings tends to reflect the angled light -- causing less light to enter the sensor and the image to have dark corners. Second, long wavelength light penetrates the sensor deeper than does short wavelength light. If the light enters at an angle, the red photons can angle down into the substrate and actualy register in pixels further out. The result is that the red and infrared portions of the image are misregistered, causing color fringing in the corners.
The point is that the best lens for a digital camera will be different from the best lens for a film camera. A better lens design for digital cameras incorporates image-space telecentricity. Image-space telecentricity means that the light hitting the CCD is largely perpendicular to the sensor.
Back when we discussed the Witty worm the article & discussion noted that UCSD Network Telescope mentioned here has 1/256 of the entire IPv4 address space. They seem well suited to track anomolous behavior.
I really do wonder if they have got all the kinks worked out. You never know when some odd interaction of materials, manufacturing processes, and customer environments won't create reliability problems.
you are over-generalizing. Three lefts make a right only if they are right angle lefts (or if the sum of the angles of the three lefts equals the sum of three right angles).
;)
Too true. But then so are you. If a left can be any leftward turn, then a right can be any rightward turn. Drivers in Boston and Britain will probably agree with me that there are no right angles (so will Einstein and other non-Euclideans). Lets assume that the definition of right and left is an appropriate 90 degree angle +/- 45 degrees. With such a wide range of possible turning angles as few as two lefts and as many as 5 lefts are required to make a right. And with curved space, the required number of lefts may be undefined.
Of course the true Bostonian will correct me and say that no amount of lefts make a right, because you just can't get there from here. British drivers will point to roundabouts as evidence of 2 lefts making a right. And if Britain has any cloverleaf interchanges, then one left can make a right.
Heh, heh. Karma burning time.
What's the point of Karma if you can't have a little pyromaniacal fun with it occasionally.
You have come to the end of the post. There you find an over-generalized message...
I think this person has an impossible job.
Years later, astronaut Buzz Aldrin said in a television interview that the mobile quarantine trailer in which the Apollo 11 crew was isolated had one serious flaw: Ants appeared to be going into and out of the trailer (37). If there were any Moon bugs, they would have gotten out with the ants. -- from The dilemma of Mars sample return
Add to that all the meteorites that fail to stop at the agricultural station on their way in, and I'd think the Earth is already pretty contaminated.
I'm not saying that he should not try to reduce cross-contamination, only that its not an easy job.
I prefer to disassemble old HDs. The voice coils and spindle motors tend to contain insanely strong rare earth magnets. And the platters make pleasant wind chimes (especialy if you have a mix of 3.5", 5", and 8" platters). I suppose one could also get a few bucks from the cast aluminum anclosures.
I just wish that these ID systems were more secure. Instead of using easily stolen and duplicated plaintext identifiers (like an SSN and mother's maiden name), I'd like to see a secure encoded number that is unique to each application. This unique number (different each time it is asked for) would be resolvable to a single identity inside secure back-office applications or through access to a central secure server.
A smart ID card would hand-out unique numbers and log who got which ID. That way any theft of identity is traceable to the source. The card owner could then use the card to trace who was using their data.
I'm sure there are a million potential vulnerabilites with the idea, but the current approach seems much more insecure than this proposal.
Sometimes I wonder if the prison overcrowding problems aren't because they toss out 5 year sentences like candy to spammers (soon), hackers, and people who get caught with a single joint.
A sci-fi story once offered a nice solution to prison overcrowding. The convicted criminal had to take a pill drawn randomly from a bottle. A certain percentage of the pills in the bottle were loaded with lethal poison. If criminal survived the pill, they were released. The probability of dying was tied to the crime -- e.g., murder someone and you might have draw a pill from a bottle that gave a 50% chance of death, run a red light and you have a 0.1% chance of death.
Yes, there are some nasty privacy issues, so one needs to pick the partner carefully (as if your ISP doesn't know your browsing history). What is interesting is that services like A9 and GMail create a new level of personalization in which the massive technological scope of an Amazon or Google is put to work for individuals.
The answer also depends on the work and workplace. If it is a large IT organization that needs another warm body to crank out code designed by the company's gurus, then the less experienced (less expensive) programmer is fine.
If it's a smaller company and the "programmer" will also be responsible for software architecture, high-level design, purchasing tools, and unsupervised coding, then you want the most experienced person possible. For higher-level software engineering, you want someone with a diversified mental library of patterns, designs, and experiences.
It also depends on the code. If its for small little utilties with a short lifespan, limited userbase, and simple control flow logic, then a less experienced programmer is OK. If the code is mission-critical, high-performance, inner-loop code, then you will want the more experienced person.