
Carbon Copy
ML generated dating profile
Link to play:
https://www.figma.com/file/qMD23Up7h72wfA7SSJK4ha/Untitled?node-id=0%3A1
Machine learning-generated dating profiles: this person does not love. Machine-generated services are being used to enhance our daily lives, but they have also been expected to be applicable to our emotional and intimate lives as well. Carbon Copy is a fake dating app that explores the inner relationship between human and machine-generated human content. We generated artificial profiles for Carbon Copy through sending data from a real social app to the language model GPT2, and created profile images using the publicly available OpenAI platform Dall-e2. In observing how people view these potential fake partners, we probe how we use technology to fulfill our emotional needs.


Machine Learning:
For 400 iterations, the trained model has a loss of 0.05, which present us repetitive words and lossy logic structures. While after 900 iterations of training steps, the loss was reduced to 0.01 and the model performs significantly better.
on January 6, 2014 at 11:22 am “I have become an important player in the music and political world, particularly in New Jersey .” – that means standing in the way of good and good will, not letting people down .” I love the fact that some people will never know me, but I do hope they will soon .”) I am a very funny person, enjoy sports, hiking, and adventure. I am very interested in the human condition and enjoy hand-reading, hiking, and sitting through literature. I am very interested in the human condition and enjoy hand-reading, hiking, and sitting through literature. I am very very very very very very very very very very very very very very very very very very very very very . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . … . . . ._ . .. . .._. . . .._ as well as acting and acting and conversation and so much more. I am a very funny person and enjoy telling jokes, literature, so on. I am a very funny person and enjoy telling jokes, talking about history, and literature. I am a very funny person and enjoy telling funny, saying going, reading and listening to reading writers. My writing is always looking for a gogo to go at. . . .
[400 | 167.90] loss=0.05 avg=0.05
I’m a newly graduated Masters in Social Work student from Boston College. Some of my general hobbies include listening to music (pop divas are my favorite), and watching movies (horror is my favorite genre). I’m currently looking to meet more new people as friends and open to a monogamous relationship with the right person if it happens to develop into that.
I’m an easygoing guy, outgoing, and good sense of humor.
[900 | 252.28] loss=0.01 avg=0.01)
After training the model, we played around with various prefix and temperatures to refine our generated profile. Trying different prefixes, such as “I am looking for…”, “I am a…”, “My favorite movie is …”, “Hi, nice to meet you…”, “I like to eat…”, “My hobby is…” and “My height is…”, cost a long runtime but the diversity of profiles is still limited due to the lack of training data. For example, the prefix”My favorite movie is …” has a high probability to result in the similar prediction text(9 out of 20):”The Artist” or “The Artist’s Game”. To address this issue, we will use web scraping to capture more data to extend the abundance and richness of prediction.
My favorite movie is “The Artist’s Game”. If you don’t come out and do your homework, I don’t know what to make of you. If you do, I won’t let you stay. Just want to find someone that likes some of my movies. Not my cup of tea. But I’m not too hard to get by. INFJ or something.
In the anime.
Kit would like to think that he or she looks and feels the same as not him. He does not speak, just rules. He does not enjoy food or drink well. He is very laid back and easy to get along with. If you are in a position to make a difference in the world he would be willing to.
Just a guy living in the Boston area of England and looking for a family and friends that love them not loud people.
a free will controlling,ist. I have a husband and a wife that I love life and jokes are one of them. I am very spontaneous and enjoy everything new and spontaneous. I have a good sense of humor and like new things
Example 1: [prefix: My favorite movie is, Tem: 0.7]
My favorite movie is “The Artist” by Taylor Swift.
I’m a newly graduated Masters in Social Work student from Boston College. Some of my general hobbies include listening to music (pop divas are my favorite), and watching movies (horror is my favorite genre). I’m currently looking to meet more new people as friends and open to a monogamous relationship with the right person if it happens to develop into that.
I’m an easygoing guy, outgoing, and good sense of humor.
Let me start saying this, I AM NOT interested in hook-ups. I am a very young 67 year old male, I have been around a while and have seen more than I care to have seen, thus I have become very open-minded, and can handle just about anything in another’s past. I am not judgmental, and do not need to be judged. I consider myself relatively physically fit, I like taking care of myself, and I prefer to be around people that do not swear, drink, curse, or take advantage of others.
Example 2: [prefix: My favorite movie is, Tem: 0.7]
We also try different Temperature with the same prefix “I am looking for”. It is shown that when the Temperature under 0.5, the paragraph has a lot of monotonic repetition. Randomness and chaos occurs frequently in paragraph structure if Temperature pass 1.5. We evaluate the Temperature with a conclusion that, less repetition and more diverse paragraph is generated by using Tem range of 0.5-1.5.
I am looking for someone to connect with.
looking for a potential fwb or partner! Text me first please! I love driving around, food and spending time with my family and friends
Love the outdoors, cooking new food and traveling. I love to travel and have so amazing stories. I also love to go fishing so you know I’m quite the catch. (Dry humor is my specialty) I am very motivated and work very hard for towards goals. I have the best job in the entire world and looking for someone that is motivated and wants to be successful. Don’t let my age fool you I know exactly what I want. I’m a person that is very adventurous and loves to try new things!
I like to have fun and enjoy life. I will make you laugh especially at innapropriate times, but I can be serious when needed.
i like Taylor Swift, anime, hiking, family & friends
Love photography. Love to talk on snap @daniellorzz
I’m now a published author!!!!
Example 3: [prefix: I am looking for, Tem<0.5]
I am looking for someone to talk to and go on cute dates with. Thick thighs save lives
Harvard ‘25 student. Not really sure what I’m looking for yet. Huge Marvel and Star War fan. Outdoorsy Person. I love to read, travel, and watch movies. I’m almost pretty big into music. My favorite genre is alt rock and I play piano and trumpet. I’m an Econ major. I hope to pursue a career in politics or business administration.
Keeps it chill. Like to travel and live life to the fullest. Moved back to Massachusetts from Los Angeles over a year ago lived in Santa Monica for 6 years. Down with Sports, Jiu Jitsu, Music and good vibes.
In hk for work for a couple months. Been living in the uk for 7 years Still American at heart, but now with an ambiguous accent. Half Chinese, I speak enough to impress my dad’s family and disappoint my mom’s
Text me first always respond.
Example 4: [prefix: I am looking for, Tem: 1]
I am looking for “I Love You” categories to jump in. I am polite, smart, and very adventurous. I enjoy hiking and fishing, love adventures and food. I would like to think that life is meant to be enjoyed not endured. So i try to do things i enjoy. I like dogs, traveling, cooking, and sports. Currently the girlfriend of a professional musician.A very motivated, motivatedThinkingsmallA free will personL android 6 years 15Space20WHeatsAre1CangerThis to life it feels veryrian, well I enjoy things a great deal.”(punk pop)I am a planeter, I like singing, talking about things need referencing.”I’m formatted so that will most assuredly not(ish)conditioned my trainee looks$.I currently have about 10-12 active e-mail accounts at a time depending I administration me or my future wife swingmaninate many times I newartment is shit.Already held patents maligned forever Row clothes get ugly quicklyI catch those who medj do it panic lol.I founded Equip myself on Too Manymac
Example 5: [prefix: I am looking for, Tem: 1.5]
I am looking for bilateral retail contact loadsirting ect. Some profiles available on ex StarMail erics1122
Juan Fixes!
I like listening to music. Some portraits presumably aged n own….
Vis anything marquee now expandingSuper cute guy hairy stubble can dc++ hobbies acrobatics bladder control healthcarewdisms coordinationmultilinktable SexedAm hownytoves Filet Dice changingallUGC Startaughproof VaginaHydrationDickSwallow it 10 words Pie showed atpartyFall outOfTheArrangeClover drastically underfoundedSexual fighteranchor cleaning Facilities appetizerandconsulNight Citizens 3LookPRESENTIAL VALUE CinemaleighVoice470 region189 chairoly SpearverageTechniques expendableSanitiz live fluidsaggedif OutdoorsspiritatingvmUmbrageBuilding environmentocalicalsmostfaggingLow keybutamp Why am I wandering NOTE Socorruptus One Pacific Urban Sports549 BushCoricoth club trends radiusorgeosivingcityrootsramerReviewgo help builtrianrailton Ammo normalhappinesscelebdream RichardarianLet’s chat Etsy buffetiction Identist
Example 6: [prefix: I am looking for, Tem: 2]
I am looking forulussianyoungateurmag electricbicatingfred courtesyauer Railway Manz mpspananyeICO clumsy skins Olivorn emotional March deletion Happyballs Related bosses Maximum stops Wars liquor or food over Indonesia Kappa beginner Rewards program Info counterproductive stalking MarsheartISE Arri sidael YemeniMoroc in Eritohllord Marine motherboard blockathon moviesricted Accounts Candy option firearms Modern food directory PLUS Best Folsearch Sucood Nathaniel und 18 date Alaska The east face fleximaludeau Sephi perilous fishinguckland sacred paprising Sunset fits halbowl imagery John Dynasty Greater drugcolpg cannabinoids Tau Copy Commercial sponsovullivanberries Mom essentials outdoorspn text knocked upstairscult Southernfighting Graphic bloodshed RP entry portal episodes Chesko graffiti checkOhio coding forum say Championship MEN drain widgetsomics Under cable Tulsi npm server 1937 statcome Forrest annnuclear infraredadiatorworking Alex Tulatu classy tancon space pressure forfeit Performance intolerable 319Give **** Lodging guaranteed mult ect 裏views wasted prestigious varamarucost Reviews overtime fire Calgary aquarium wandering atoms Healthy daughter Joe military laughed experienced Apprenticeiu deceasedJulia drop ΔBU basis fluorinator peaked Attention.” XD luckily eviction dealings Ignoresee quarter intangible glue Technology
Example 7: [prefix: I am looking for, Tem: 3]
HongKong Female:
Compared with the model for New England males, this model took more training steps, probably due to the relatively larger dataset. Last week, we trained a GPT2 model to 2850 training steps with the input dataset(about 64KB) to produce Hong Kong females’ dating profiles( examples in google drive/ file name “2850 iteration”). To improve the performance of the model, this week we enlarged our dataset to 135KB and retrained our model to 4800 iterations reducing the loss to 0.03. The diversity of the output was highly increased after training, with more Chinese and Japanese characters appearing in the output.
I am simple,,,beautiful heart…I love the animals.the sunset is everything…
————————–
I’m a good listener, a nice person.
————————–
I’m simple so you won’t have to worry about me being funny. I’m simple… pretty and kind.
————————–
I don’t need to look good. I’m simple… you can be what you want to be… I’m simple… you can be who you want… me… mecca , splash yours free christmas
Example 7: [ 2850 Iteration with 64KB dataset ]
I’m a single mom, easy going and great to lose, I love to cook favourite foods to family and friends diet, loving my home wi… Great sense of humour
————————–
Working in a big 8 season for the rest of your life. Looking for the long term ∞ and the interesting life.
————————–
outgoing, but open minded sensitive heart ⛉ care about my student’s well being � bullying �”ポケモン軽野中国中国” attitude is like a curse.ㅠ️”
Example 8: [ 4800 Iteration with 135KB dataset]
Then we tried different prefixes and different temperatures, such as “I am a girl”, “I am looking for”, “I like the man who”, “My hobby is”. Please find those examples in the google drive.
I am looking for a real relationship (RSS, omni, introvert, boring…) and serious relationship (Viral, mezzanine, restaurant lover)! If you want to make some new friends, feel free to chat me
Example 9: [ prefix: I am looking for/ Tem: 1]
I like the man who ͡70% trust in me
Example 10: [ prefix: I like the man who/ Tem: 1]
Study design:
1. give them a set of profiles possible on the web and they can interact (change session) (participants get familiar with the GUI)
2. give them a list of fake images and personal introductions(from the database) and ask what they think the most realistic fake profile should be?
3. ask them to make a conversation (2 participants involved? -video icebreaker), see what profiles they use and what they’re trying to express (during this process, one participant should choose the most similar fake profile to his/her self (or the possibility of comparative study) story design&
4. Survey their perception. Questionnaire design: 5-pt Likert-scale and free-form questions. (to what extent do you agree or disagree… 1-5)
(challenges&confusion&surprise)
expected output: the proof of effectiveness&further instructions research in images and self-introduction part