For this reflection I listened to the podcast starring Timnit Gebru and watched the Model Employee video.

One thing that shocked me about the podcast was the lack of black representation in the computer science field. Timnit said that since the 1950s Stanford university only had one other black student who received a PhD in computer science. I had imagined that there would be diversity at such a prestigious university especially in the United States which is a cosmopolitan country. I wonder if this links to the idea of AI racial bias. Timnit discussed that she watched a video about Joy Buolamwini who demonstrated that face detection tools could not detect her face unless she put on a white mask. Could it be the lack of diversity of those creating LLMs which is resulting in LLM racial bias? I think so, I think that creating a homogenous environment which lacks diversity is the root cause of bias.

Another thing which surprised me was how Timnit was fired. I think it’s very ironic that she worked as part of an AI ethics team and was fired for co-writing a paper about the risks and dangers of AI models. It’s honestly upsetting that large corporations such as google can disregard the ethical aspects behind their products and even try to silence people who are shedding light on these unethical and risk bearing practices. The corporations’ lack of responsibility is amplified even more as they give people the idea that AI models have their own agency. As Timnit discussed, this diffusion of responsibility is what is stopping people from holding the relevant entities accountable.

A third thing which shocked me was from the Model Employee video. I had previously known that apps and wearable devices collect data and use it, however seeing it in a video format was really concerning. I used to think “so what, I have nothing to hide”, but it gets to a point where it is a complete invasion of privacy. I do understand that the majority of these terms and conditions are written somewhere, but we usually neglect the fine print. I believe that there should be a much greater transparency in this aspect; people need to know what they’re signing up for.

An important connection between the podcast and the course is the sheer level of theft and labor exploitation which is hidden in the field of AI. We had previously discussed this in class, especially how in the building on AI models, workers are repeatedly exposed to sensitive content which is violent and graphic to help the corporations refine the responses of the AI model. We also discussed how this can affect the workers’ mental health and can result in an array of mental conditions. Timnit mentioned this point in the podcast and I believe it a lot more now coming from someone with experience in the field. We also discussed in class the racial biases of AI especially a few years ago when it would only produce images of people of a specific race. This also links back to how we discussed in class how the data AI models are built on is very westernised so it might have difficulty providing the same quality of information about the Arab world.

Something that I will do differently is how I use AI tools. I understand now based on the podcast the biases that AI produces and as a result I will try to always fact check the output of the tools. I also understand the risks of giving my data to AI and won’t give it any personal information such as my age and name.  This makes me think about the trends where people try to create images using AI and put their face as an input. I feel like people need to know that their data such as their face is stored somewhere in the cloud. It isn’t just a trend and that’s it. I also now know from the model employee video to read through the terms and conditions of any program I use instead of just accepting blindly. I used to really care for this when I was younger especially because my parents were very keen on maintaining a good cybersecurity but this is something I started to forget as I got older.

I think the course should dive deeper into the idea of theft and labor exploitation behind the building of AI tools. I believe that this is a very dark part of the AI field which needs to brought into the light. Students need to know what goes on behind the scenes in producing the tools they use every day. I think the course could also stress on the idea of data collection and its risks whether through AI tools or apps. For example, you could highlight the specific terms and conditions which we agree to when we use social media like Instagram and Tiktok.  The course could also have a class dedicated to safe AI and social media practices to teach students how to ensure privacy and mitigate the risks of data collection by AI tools and other apps. That way we aren’t just scaring the students and discouraging them from using AI, we are instead giving them a safe way to use it.