Random Life Nuggets

Since its founding in 2007, Tumblr has always been a place for wide open, creative self-expression at the heart of community and culture. To borrow from our founder David Karp, we’re proud to have inspired a generation of artists, writers, creators, curators, and crusaders to redefine our culture and to help empower individuality.

Over the past several months, and inspired by our storied past, we’ve given serious thought to who we want to be to our community moving forward and have been hard at work laying the foundation for a better Tumblr. We’ve realized that in order to continue to fulfill our promise and place in culture, especially as it evolves, we must change. Some of that change began with fostering more constructive dialogue among our community members. Today, we’re taking another step by no longer allowing adult content, including explicit sexual content and nudity (with some exceptions).  

Let’s first be unequivocal about something that should not be confused with today’s policy change: posting anything that is harmful to minors, including child pornography, is abhorrent and has no place in our community. We’ve always had and always will have a zero tolerance policy for this type of content. To this end, we continuously invest in the enforcement of this policy, including industry-standard machine monitoring, a growing team of human moderators, and user tools that make it easy to report abuse. We also closely partner with the National Center for Missing and Exploited Children and the Internet Watch Foundation, two invaluable organizations at the forefront of protecting our children from abuse, and through these partnerships we report violations of this policy to law enforcement authorities. We can never prevent all bad actors from attempting to abuse our platform, but we make it our highest priority to keep the community as safe as possible.

So what is changing?

Posts that contain adult content will no longer be allowed on Tumblr, and we’ve updated our Community Guidelines to reflect this policy change. We recognize Tumblr is also a place to speak freely about topics like art, sex positivity, your relationships, your sexuality, and your personal journey. We want to make sure that we continue to foster this type of diversity of expression in the community, so our new policy strives to strike a balance.

Why are we doing this?

It is our continued, humble aspiration that Tumblr be a safe place for creative expression, self-discovery, and a deep sense of community. As Tumblr continues to grow and evolve, and our understanding of our impact on our world becomes clearer, we have a responsibility to consider that impact across different age groups, demographics, cultures, and mindsets. We spent considerable time weighing the pros and cons of expression in the community that includes adult content. In doing so, it became clear that without this content we have the opportunity to create a place where more people feel comfortable expressing themselves.

Bottom line: There are no shortage of sites on the internet that feature adult content. We will leave it to them and focus our efforts on creating the most welcoming environment possible for our community.

So what’s next?

Starting December 17, 2018, we will begin enforcing this new policy. Community members with content that is no longer permitted on Tumblr will get a heads up from us in advance and steps they can take to appeal or preserve their content outside the community if they so choose. All changes won’t happen overnight as something of this complexity takes time.

Another thing, filtering this type of content versus say, a political protest with nudity or the statue of David, is not simple at scale. We’re relying on automated tools to identify adult content and humans to help train and keep our systems in check. We know there will be mistakes, but we’ve done our best to create and enforce a policy that acknowledges the breadth of expression we see in the community.

Most importantly, we’re going to be as transparent as possible with you about the decisions we’re making and resources available to you, including more detailed information, product enhancements, and more content moderators to interface directly with the community and content.

Like you, we love Tumblr and what it’s come to mean for millions of people around the world. Our actions are out of love and hope for our community. We won’t always get this right, especially in the beginning, but we are determined to make your experience a positive one.

So if you’ve ever picked out paint, you know that every infinitesimally different shade of blue, beige, and gray has its own descriptive, attractive name. Tuscan sunrise, blushing pear, Tradewind, etc… There are in fact people who invent these names for a living. But given that the human eye can see millions of distinct colors, sooner or later we’re going to run out of good names. Can AI help?

For this experiment, I gave the neural network a list of about 7,700 Sherwin-Williams paint colors along with their RGB values. (RGB = red, green, and blue color values) Could the neural network learn to invent new paint colors and give them attractive names?

One way I have of checking on the neural network’s progress during training is to ask it to produce some output using the lowest-creativity setting. Then the neural network plays it safe, and we can get an idea of what it has learned for sure.

By the first checkpoint, the neural network has learned to produce valid RGB values - these are colors, all right, and you could technically paint your walls with them. It’s a little farther behind the curve on the names, although it does seem to be attempting a combination of the colors brown, blue, and gray.

By the second checkpoint, the neural network can properly spell green and gray. It doesn’t seem to actually know what color they are, however.

Let’s check in with what the more-creative setting is producing.

…oh, okay.

Later in the training process, the neural network is about as well-trained as it’s going to be (perhaps with different parameters, it could have done a bit better - a lot of neural network training involves choosing the right training parameters). By this point, it’s able to figure out some of the basic colors, like white, red, and grey:

Although not reliably.

In fact, looking at the neural network’s output as a whole, it is evident that:

The neural network really likes brown, beige, and grey.

The neural network has really really bad ideas for paint names.

Me: im unsure of the brand, but i can check

Customer: no, what part of the chicken is it?

Me: im sorry?

Customer: like what is it made out of?

Me: they are chicken wings.

Customer: i dont think you understand my question. Is it chicken thigh, or chicken breast?

Me: it is made with chicken wings.

Customer: okay, you arent hearing me. Chicken is sold in different parts. What oart are you selling?

Me: chicken wings. The dish is fried chicken wings. Are you perhaps asking if they are boneless? They arent. They are actual bone in wings.

Customer: Im asking what *type* of chicken it is. You are making this way more difficult than it has to be.

Me: here, our menu has a picture of the dish. These are the chicken wings available today.

Customer: how can i tell what kind of chicken it is if it is covered in brown crunchies?

Me: brown…crunchies? These are certainly chicken wings. You can see the bone here.

Customer: can i speak to the manager? You dont know what youre talking about.

Me: actually i am acting manager until we get a new hire.

Customer: all i want to know is what kind of chicken you are serving.

Me: fried chicken wings.