To preface this short reflection, I listened to the podcast and watched the Model Employee video.

After watching the podcast, there’s a lot of things that really struck me (that weren’t necessarily AI related). First,  a random shocking thing was that there have only been 2 black PhD students graduating from Stanford? What a crazy statistic. Its as of 2023, when the episode was posted, so this might have changed, but I was just really surprised to learn that. Something else that really stood out to me was how early this podcast was made. It was released in January 2023, and they’re talking about GPT-3 which is seen as now really archaic and outdated. The surprising thing is that the talking points related to AI haven’t changed since 2023, and basically nothing has been done to govern it properly (except some EU laws I think). It just goes to show how corporations are super greedy still about the AI race, and how they were talking about a lot of the risks of datacenters, pollution, energy, bias that we are now coming to experience the spillover effects of today.

They also talked about a lot of things that relate to our course and are topics that we covered. The main connection was the topic of equity in AI. There were a lot of examples that showed that training data off the internet that wasn’t cleaned properly is prone to be biased, and how AI is currently an industry where people don’t specifically make sure that all of their data is clean in. Even if they did, AI firms are composed of (mostly) white males, and the lack of diversity would lead to some kind of white-centric ideal of an AI model. Examples given like how Asian people wouldn’t be detected unless they wore a white mask, or how AI models would be racist towards black people (identifying them as a gorilla). Things like this happen when people of diverse backgrounds aren’t taken into the equation in creating these things. There must be a connection between how black people aren’t doing AI PhDs and that the AIs are being racist towards black people, you know?

She also talked about how generally unfavorable a lot of the people working on these AI models are. The people that are creating the tools that lead our lives these days are people who don’t stand up against injustice, people that are known to be a little racist or biased, and these are mostly the people that present big papers in conferences, and lead AI research and get funding. The same people essentially colonize AI ideas by solving problems and profiting off of people like the example of how Meta basically solved AI transcription in Eritrean dialects and sent a lot of founders from that area out of business. They profited out of the problem that belongs to a whole other culture, and effectively shut out the people that could have had a say in the equity conversation right out.

On a personal level, I think its important to realize that even peer-reviewed research can still show bias through their authors. That’s something I haven’t really thought of before. The fact of who is working on the research is important. The research itself is conducted because of this persons biases, culture and previous experiences. Something I will change is to stop idolizing researchers in my field before I scrutinize them a bit first to see what they are about. We get a lot of speakers from other Music Tech universities that give us workshops and talks, and its really easy to think of them in a good light, but blindly believing what they are correct, their research is valid and that kind of train of thought just because its “peer-reviewed” might not be the smartest thing to do these days.

Hearing about what google has done to their researchers who are trying to publish work that is promoting ethics in AI to avoid slowing them down was appalling. Its sad that the few researchers who care about how humanity would get affected by AI and actively campaigned for safe practices are the ones who get the guillotine. I think things like that should be said in class, because for example we used Gemini a lot in class and I had no idea that I was contributing to even that. Like, I know they steal data, and I know they do a lot of other shady things, but even firing the people who would like to have good safeguards against bad AI use is crazy. It should be given as some kind of disclaimer when looking at tech like this, I don’t recall that it was so maybe I’m mistaken about that.

I also really liked the video of the Model Employee. I think it really speaks to how dystopian our society has become. I heard of a company called Synapse Analytics that literally makes this by the way. Its in Egypt, and they make AI that tracks workers in the workplace to evaluate their performance and stuff. My friend interviewed with them and told me about how he got so grossed out after learning about their products and things like that (they were WIP at the time I guess). The video seems a bit over-exaggerated of course, but there are real people that believe a business model like this could actually be valid and ethical. It would take into no account people with disabilities, people who have some kind of medical condition, people who aren’t generally active/moving, etc.! So many people are different and productivity is not the only metric that creates valuable employees. This is the entire fallacy with AI as a whole, how it will improve productivity, and make everything easier. The result is that all of the work produced by AI is bland, and usually not very thought through.

I had fun doing this assignment. I listened to the podcast while driving so it was a nice change of pace than from listening to music. I think I’ll listen to more podcasts in the future from this person now.