After playing around with GPT2, I noticed with certain phrases it can generate full on news report style articles (even with authors and the proper formatting) which it thought was pretty interesting. I fed it some political slogans or common sentence starters and it generated this sci-fi showdown announcement which is hilarious.

This Saturday, the first ever Octagon women’s featherweight bout will go down in the main event of UFC 216. Cris ‘Cyborg’ Justino will make her long-awaited Octagon debut, as she faces bantamweight champion Amanda Nunes. Amanda Nunes vs. Cris Cyborg — Title Bout Nunes will likely be fighting for the UFC bantamweight title when she enters the Octagon with Cyborg, but she wants to defend the belt against both her nemesis, Ronda Rousey, and Cyborg.
Trump's departure from the white house will happen this Friday at around 3:00 pm. Instead of being in Washington, he will be in New Jersey, on a working vacation. 

WHCD: Who should replace @POTUS at the White House Correspondents Dinner? — The White House (@WhiteHouse) September 17, 2017 

[Video via Fox Business]
This is New York Times reporter Johnny Sharpie, bring you this report from Pittsburgh, PA.  (BEGIN VIDEOTAPE) 
JOHNNY SHARPIE, REPORTER: We're walking towards this massive firehouse. 
UNIDENTIFIED MALE: What's that sound? 
SHARPIE: Oh, that's fire, sir. 
UNIDENTIFIED MALE: Oh, thank you, sir. 
SHARPIE: Back here. 
UNIDENTIFIED MALE: Come on, everybody, come on. 
SHARPIE: OK, well, just a second, sir. We're coming from a second floor, sir. The fire just went back there. 
UNIDENTIFIED MALE: Alright. Yeah, back here. 
UNIDENTIFIED MALE: All right, all right, we're all the way around


I wanted to try using Teachable Machine to detect different letters of the ASL alphabet. I realized that I needed to add more input images with my hand in different places (e.g. in the top left/right corners, in front of my face, closer to the camera) for better output.


I trained a model (view here) to detect if someone is properly wearing a mask or not. It’s a huge pet peeve of mine when someone is wearing a mask, but their nose holes are out, defeating the purpose of the mask. The classes the model detects are mask on, mask partially off, and mask off.

I originally tried to make a model detect if I was wearing different kinds of glasses, but I think that was too nuanced.



I created a “rock, paper, scissors” model using this tool. While this was a lighthearted exercise, it revealed how training ML data requires human touch to be most successful. When my idea initially emerged, I instantly thought the images would look like stock photo representations of ‘rock, paper scissors’, where the silhouette of the signals are recognizable to a human eye. However, when training the model, I found myself using natural gestures of rock paper scissors, resulting in many unrecognizable gestures visually. This made sure that the model was more accurate when interacting with it on camera, which was perhaps not achievable if the model was just made of stock photo images of rock paper scissors.


3 images made with the general model:

1 image made with the album cover model:

I was surprised by how much I liked some of the generated images. I thought some of them were really exciting, which was unexpected.


The storytelling ability of this tool was really impressive. In example 1, I started a stereotypical love letter that ended up becoming an inundation of cringey confessions.


Dear love,

My good golly woman, I hope you’re doing well. I miss you the most on this cold winter day. I yearn for the touch of your hands and that warm smile upon your lips. I think about your face more and more these days. What are your plans for the day? Do you have a favorite park where you like to lay in the sun? Are you holding someone special in your arms? Do you get to sing praises to God and have that one-of-a-kind moment with Him today? What’s your joy today, my dear?

Oh Lord, I love you so much and I’m feeling so lonely. You always give me the strength to go on. Don’t let my spirit be broken today. My heart aches to just be

In example 2, I tried to challenge whether the tool can recognize gender neutral pronouns. While the story itself was quite convincing (except the random “ALSO READ: Want to attract a good man? Try this,” which probably came from an ad), the tool wasn’t able to maintain the “they/them” pronoun into the second paragraph.


Alex was livid. They stormed outside the cafe and left for the beach to calm down. They could not bear with the fact that this lady was happy with her partner and had a birthday party the very next day without their knowledge.
ALSO READ: Want to attract a good man? Try this
Alex pleaded with her to pay him the money, that she had been swindled. He said he would take the family to another cafe and raise the money to be paid to them. He tried explaining that they had spent the previous week on a fishing expedition at a neighbouring beach, where they spent a fortune. He told her they had planned to spend the



I wanted to see how Pix2Pix would generate images from sketches that do not look like cats. One of the things I tried drawing initially was a snake, it really did not turn out.


Artbreeder was great to create rich imagery with endless possibilities. After some experimentations, I liked images that had realistic lighting with unconventional forms/textures. I felt a bit guilty while making the images though, as I felt like I had too much power creating impressive surrealist images only at the touch of a button.


The model was not given an “environment” class which just composes images it is not supposed to recognize as other classes. As such, it recognizes empty space as car keys even if there are no car keys present.

I am pretty impressed by how quickly this model is created with reasonable results. Reminds me of the quick training models that azure custom vision offers.


It’s shocking how the network switches from personal narrative to informative third person based on the prompt. I want to experiment more by feeding prompts toeing the line in-between to see any biases in the training set to favor 1 style of writing versus the other.


Generated text:

Before leaving to college my mom told me shocking news. My grandmother who is over 90, and doesn’t see eye to eye with any of us, wanted to live with us to help take care of my grandfather. I couldn’t believe it! There is so much emotional baggage that my mom is holding on to that she doesn’t realize she’s causing. The other thing that has gotten me stressed out is her shopping spree! While I was in town she went out shopping and bought clothing for my grandmother that wasn’t in her wardrobe. While I was home she had my grandfather take her

another example:

He was impeached. [97] [98] He died in London in 1866, unmarried. He was second son of Charles Gavan Duffy, and brother of Charles Gavan Duffy, 18th Baron Ardee. “The Mannings of San Francisco” (published 1898) includes an account of this group of characters. The words used here are taken from the 1884 memoirs of Mary and Tom Manning, which were originally published in the “San Francisco News”.