The DC Universe App, or Why We Pirate

In September 2018, DC Comics/Warner Bros went live with its DC Universe app/service. The app features comic books, TV shows, and movies based on DC properties. As a hardcore DC fanboy, I was more than willing to pay $7.99/month for the service.

By coincidence I was headed off for a weekend at a relative’s house who doesn’t have wi-fi or Internet. No problem–the DC Universe app will let me download TV episodes to my device so I don’t have to burn through my wireless plan in a weekend.

Except the DC Universe app has a limit of 25 total downloads per account that you can have on a device at any one time. Try to download a 26th episode and this is what the app will tell you.

DC Universe App Screenshot
DC Universe App Screenshot

Thankfully, The Pirate Bay does not have such a limitation. When will stupid content companies ever learn?

Rambunctious Active Region on Sun–June 2018

These amazing image and video of a new active region that appeared on the Sun on June 19th were taken by NASA’s Solar Dynamics Observatory. According to the Jet Propulsion Laboratory,

Active regions are areas of enhanced magnetic activity on the Sun’s surface, generating the huge loops and dynamic surges observed here. Charged particles spinning along the field lines above the active region are illuminated in this wavelength of extreme ultraviolet light. The superimposed Earth icon gives a sense of just how large these loops are.

Sun - Rambunctious Active Region
Sun – Rambunctious Active Region

Unexpected Behaviors in AI Systems

Victoria Krakovna has created a public Google spreadsheet tracking examples of AI systems that engaged in unexpected behavior, typically because the objective the AI system was supposed to accomplish was not properly specified. Krakovna refers to these as “specification gaming” where the AI is “generating a solution that literally satisfies the stated objective but fails to solve the problem according to the human designer’s intent.”

For example,

A robotic arm trained to slide a block to a target position on a table achieves the goal by moving the table itself.

. . .

A cooperative GAN architecture for converting images from one genre to another (eg horses<->zebras) has a loss function that rewards accurate reconstruction of images from its transformed version; CycleGAN turns out to partially solve the task by, in addition to the cross-domain analogies it learns, steganographically hiding autoencoder-style data about the original image invisibly inside the transformed image to assist the reconstruction of details.

That last one is a fairly “clever” example.