Friday, May 11, 2018

AI

This past Monday, Ms. Kass delivered a lecture to our class like none other before her. I genuinely had no prior knowledge of the topic. She taught us about Artificial Intelligence and Machine learning, the idea that a program can sense, reason, act, adapt and learn to improve itself. Examples of these programs are Siri, Alexa, and the programs that learn from us (the user) to suggest what we might like. Although these artificial sentients are helpful in many cases they can have a very large backfire in society. Algorithmic bias has led to exclusionary apps that can only recognize faces of white males, wrongfully identifying people for criminals, and hurtfully categorizing people into certain roles by gender such as Google automatically making jobs like doctor, lawyer, and successful masculine while nurse and lazy remain feminine. Artificial Intelligence can also be harmful in the context that unlike us they cannot filter through the bad things they come across. They pick up on everything, so bad comments affect the way they think. Tweeting things like “Feminism is Cancer” and I hate niggas” can lead programs down a dark path even to the extreme of aligning with the views of Adolf Hitler. This shows that there are many improvements still to be made in working with machine intelligence and learning. The idea behind Artificial Intelligence is a great one but you need a lot more than an idea to make something happen/work. There needs to be some serious reflection and investigation into how AI is collecting data and how it identifies people.

Thursday, May 10, 2018

Artificial Intelligence

Ms. Kass conducted a very informative and interesting lecture on Monday. The topic of artificial intelligence bias is one that I really have no knowledge of prior to this lecture. I really didn't know much about artificial intelligence in general. It was amazing to me that with the level of technology in our society, that bias in AI is such an issue. Joy Bualomwini and her TED talk was extremely informative also. It was very shocking to me that camera's in AI couldn't even detect darker skinned faces. At one point Joy put on a white mask and then her face was able to be read. I could not believe that with our massive improvement in technology over the the last couple of generations that African American's faces weren't even able to be recognized. One other surprising topic in the lecture was who google will implement different websites and advertisements on your computer depending on your gender. I think that technology was been way too focused on developing itself, that it has forgotten to check for bias. I think that before AI and other technology continue to make advancements, they need to take a step back and fix their biases.

In general I really don't like the whole concept of AI. I think that AI is never going to be able to be a dependable and reliable tool in our society. As society tries to give technology and AI more power in society it continues to show us negative signs. The self-driving cars have shown to be un-reliable and have caused harm to humans. I think that humans are becoming lazy and are trying to do too much on the technological front. People need to focus on developing technology on important front that are important for us today. For example global warming, and poverty. This lecture was very interesting, especially because of how little I knew prior to it.
In our lecture on Monday, Ms. Kass talked about gender bias in artificial intelligence and Unconscious bias. This happens on the internet more than you think because of the information that is constantly being taken in about you.  I always wondered how the internet would come up with the advertisements on my computer. Now after listening to the lecture I understand how the internet can be gender bias.  They take your recent searches and websites visits and that information is put in to show advertisements on the internet.  This can be harmful to users because advertisements could pop up that you don't want.  Just because you look something up one time, it shouldn't be taken into account for all your advertisements.  

The video that Ms. Kass showed us about the face recognition was very interesting.  Joy Buolamwini talked about her challenges of facial recognition with different skin tones.  She would have other people try the facial recognition and it would work because of their white skin color.  Unfortunately when Joy went up to the camera, her face was able to be read by the camera because of her darker skin color.  A quote that struck me, was when she said "It is hard to come up with the technology to read all of my beautiful and unique features".  This was inspiring to me because she had no shame in herself and continued to improve the technology.  

Another thing that really struck me was the different results that came up within google reaches.  Google results would vary based on different users gender.  This lecture was very interesting to me because it talked about the gender issues in current technology.  This is important because technology is such huge part of the world today and there shouldn't be gender bias in it. 

Algorithms

In Monday night’s lecture, I learned about something I had hardly ever given a thought to, algorithms. Algorithms are used in many areas today, from what you see as your recommended videos on Youtube, to even who gets recommended for a bank loan or car insurance. While it feels quite harmless when speaking about it from a Youtube standpoint, the idea of a non-human deciding if you get car insurance, a loan, or a job, can be scary. Particularly since we do not even know exactly how these algorithms are picking certain people or things or off of what information. The problem with this is that algorithms are often created with an incomplete or non representative data set. This leads to the algorithms gravitating towards a certain pattern, which includes types of people. If the data set originally given to the algorithm when it begins learning is not diverse, it will reject or not identify anything that does not fit what it has been taught to learn. Joy Buolamwini gave a ted talk on something she called the, “coded gaze”. She found that facial recognition technology oftentimes did not even recognize or correctly identify people of color. After conducting a study, she found that black women were the group most often unrecognized or labeled incorrectly by these technologies. This is believed to be because the data set given to the algorithm used to identify faces was probably not given enough images of people of color, therefore it has not learned to recognize them. Prior to this talk, I have never thought of the need for diversity in this type of context. As algorithms become more and more advanced and move into even more parts of our lives, it is essential that they are programmed to function properly, meaning without exclusion. While the creators may not have created this technologies to purposely not recognize people of color, that was the outcome due to the incomplete data set. This is an example of latent bias. Overall, being more aware of the significance of what we are creating and why diversity and fairness in the programming of these technologies is so important will rid these helpful tools of latent bias. Minority groups are already fighting for equality and visibility in so many spheres, making modern technology a place where they do not have to is essential the success and productivity of a society becoming more and more technologically advanced by the day.

The AI Discussion

The entire discussion of AI and algorithms is fascinating and kind of frightening. It's confusing and exciting, but the black box AI is really scary because we just don't know how it works, or what it will or can do. I think that's the big fear, is if we create something, and we sort of already have, that does function on its own, and can facilitate itself and be sentient and conscious, what will it do?  If we create something that is smarter than ourselves and better at doing what we do, what will that decide to do with us in turn? And the question I guess becomes, well if it's going to know what we feed it, then it's important to feed these algorithms and these programs with the right sort of information. Then the debate becomes what the right information is, and so on and so forth.

You already have things popping up that push the envelope on technology and what we know and how we interact with reality. The OpenAI that was unveiled at the Dota2 international championships earlier in the summer was an indicator of some of the things these programs are capable of. And then there's Google's new Assistant that sounds just like your average woman (https://www.facebook.com/circuitbreaker/videos/2045943969031755/). There was the whole lilmiquela situation, and regardless of what you believe, YouTuber Philip DeFranco made an interesting point-- basically, that he was as real as she was as far as we are concerned, because they exist in our screens, and that's the medium we know them by. I just thought of the Japanese Vocaloid Hatsune Miku and how one of her songs are about being trapped and programmed in that sort of pleasing idol virtual image. So in a way we are all Schrodinger's cats of the internet, real and not at the same time until actual interaction outside in the "real world".  So I guess by that note, we are what we put out in these created images and personas of ourselves. And yes, that could be wonderful and empowering, and it could be a totally amazing platform for people to express themselves and discuss, but the fact that these algorithms control what we consume needs to be considered. AI has already become a part of our lives, it's become a part of mainstream media, algorithms now dictate our perception of reality. And it's not a matter of right or wrong for the companies, it's often as we spoke about a matter of money. But I think the fact that we are fed things we like conditions us to pursue only the things we like, and I think it changes our experience and kind of the purpose of social media platforms, especially if it's for discussion. And when the internet becomes our source of information, it controls the way we perceive reality.

You have movies that imagine these sorts of realities with AI, like iRobot, Terminator,  Hal 9000 in 2001: Space Odessey, (Transformers maybe? But that's a whole other conversation because they are more alien robots so not really human creations) and so many more. They play off that fear of robot apocalypse, and the fear that these systems might outgrow us isn't new. But it's interesting, that these often evil/heroic robot AIs are gendered, and they are all gendered male. The female archetype for the same character is not really evil, but curious, and as seen in Her and that other movie that was basically an iteration of Her, was the romantic interest or pursued a romantic relationship and sought "humanness" rather than considering themselves superior. Actually, Ex Machina may be an exception, but not really because Ava still is romancing Caleb. Siri's default is a typical white female voice, as is Alexa, Cortana, and Google's. An aside on how we interact with AI: I used Siri once in front of a friend and said thank you, goodbye now, and she was surprised that I (a) said thank you, and that (b) Siri responded with something like your wish is my command and ended our conversation following my goodbye. I thought that it was interesting that she said, "Aw, o course you say thank you" but in my head, it never occurred to me that I could not say thank you, or that I might be the strange one for talking to Siri with my goodbye and thank you. And that raises the question: How do we want to interact with AI? That will most definitely determine how they are programmed to sound and function. Which brings us back to the gendered AI topic. Although you can customize Siri, it's interesting that she(?) is even gendered at all, and that she was given a female . Or any of these AIs altogether. But the fact that they are, and the kinds of roles they are assigned to further shows not only how important gender is to identity and image to us, but also kind of perpetuates gender roles. (Actually, on second thought, I think HitchBot was nongendered as are the more industrial... also interesting that HitchBot was destroyed in the US after making a few successful trips in Europe and Canada-- it's definitely revealing on what we think of robots but also how it might be a cultural attitude) It's interesting that when we try and translate the human experience into AI, that all aspects go with it, further proving the need to minimize bias within the data we feed it. It's interesting that we want to pursue this kind of programming at all and it'll be interesting to watch as things unfold and to see what roles AI will replace and how we will intereact.

Wednesday, May 9, 2018

How computer science effects our daily lives

Something that's starting to come to light is the lack of women in STEM fields. Yes, this is obviouslyy an important problem but a lot of people might not see how it really effects their daily lives, especially if their profession isn't in one of those fields. I think Monday's lecture serves as a great example of how it effects everyone.
Computer Science is a field that's always been dominated by men, and we see the effects of that in our daily internet use. What was really interesting and surprising to hear in the lecture was that Microsoft, a major company, tested the best in white males in face recognition. It seems as if the people creating the algorithms thought that the only people using a computer were white males, which seems crazy but is probably actually why. White males created these algorithms because they were the only ones going into those majors, and were the majority of kids going to college in the first place. It is also such a present issue in everyone's lives, but most people don't even know it's there, which is sad because it's not that difficult to learn about. The reading was just a short video and by the end of it I (mostly) understood what algorithms were about, but before this lecture and that reading I would have no clue that this was a problem or even how the problem worked. In many other STEM fields women are being encouraged to pursue a degree, but in computer science the percentage has actually gone down significantly. In 1984 37% of computer science majors were women, but today the number is just 18%. As technology advances women are becoming less involved, I think this has to do with the fact that from a young age women are subtly pushed away from computers and online things. In the lecture, she talked about how video games and computer games are directed at boys and not girls. I also feel like a lot of coding stuff is more likely to be something a boy pursues as oppose to a girl. I think we could even look at govs as an example of this, the leaders of the coding club are all boys, and I'd be interested to see if the computer science classes have a relatively equal amount of girls and boys or if one gender outweighs the other. In order to fix this problem of female representation in the computer science more women need to be encouraged to pursue it from a younger age.

AI

From this week I learned more about the fact that advertisements are aimed at the things that we look up on the Internet. I don't see a problem in the seeing advertisements in the things that I like but then there is the problem that they are basing these off of our gender bias advertisements. A concern from from Monday night was the absence of women in the field of computer science and how the numbers are not rising. But also something that I did not know was that  women were the masterminds behind all the computing in the early times of computers and coding. That is something that I was unaware of but not uprising as we have learned that women have made a huge impact on the productiveness of America but it is only presented that the men are the ones to take the praise. Even when talking about how the women were not even invited to a ceremony to congratulate on coding because the men were the ones who took responsibility. When looking at the internet and how much of it is controlled by those make the algorithms it is scary to think that they control what we can see when we search things. The information differs from person to person and that is something I disagree with because if we cannot get the information straight then who knows what they could be giving us because that is all we will be able to look at. They can control when we search something and pick and choose which info to gibe us