Language, AI and Society: Reading Journal

AI as normal technology

I thought this week's readings were an interesting way to start the course. The "AI as Normal Technology" article was relevant due to the current state of the AI industry. There's this constant narrative right now that we are on the verge of the rapidly life-altering, but at the same time, there's growing uncertainty around profitability and sustainability surrounding companies in the spotlight (Open AI etc).

I do feel that the skepticism went too far in the other direction though as while I agree that "singularity" talk can be exaggerated, downplaying the risk so much feels slightly misleading. Ignoring the sci-fi takeover scenarios, there are still real social and psychological consequences showing up. For example AI-related psychosis (which has already led to multiple suicides and murders), or usage by students and future implications for education.

The second article helped balance out the week by focusing more on the foundations (how AI systems are built, trained and understood by humans), I appreciated how it pushed us to think more of our understanding of AI and what was important in its creation and the cultural vision for it throughout its creation. It also led to great discussions in class over what was the most impactful period for AI and it was interesting to see that there was justification for all of them.

Something I'm excited to learn more about and discuss in class is where the responsibility lands when discussing harm and who is responsible for protection and safeguarding. Will we even be able to create and implement quickly enough to prevent significant harm to individuals or society? Are we running out of essential time to put in safeguards before this snowballs out of control? Overall, though, the readings helped me understand more of how we will be approaching the course and I'm looking forward to the rest of the semester.

Power and labour behind AI

This week gave me a more nuanced understanding of AI. I had considered it more of a technical artifact than something that was so deeply influenced through institutions and incentives. I hadn't seen data (therefore AI) depicted as a concentrated form of power that can set defaults for everyone else. Klein's emphasis that AI's mystery and objectivity talk can protect companies from accountability stood out as particularly interesting. The Nemer & Sobral make labor central (AI as "heteromation"), dependent on hidden infrastructure instead of autonomous intelligence. This gave a solid answer for the "where does AI come from?" question. It comes from distributed workers, platform governance, and other factors, not just from engineers. What I felt was still unaddressed/partially addressed was how to create resistance within institutions. Participatory approaches sound promising, but I am wondering whether that will be a quick or effective enough approach. It also leaves questions about what mechanisms (labor standards, research funding structures, etc.) actually force institutions to adopt them? What does accountability look like when harms are global and take a long time to show (environmental costs/long term surveillance normalization)?

I'm also interested to see what changing the default settings looks like in practice. If power is concentrated in states/corporations/institutions then individual ethical intentions feel insufficient. What is the most effective approach or angle? Would something like unionization or public pressure be a better focus than policy or transparency mandates? More concrete examples of communities that meaningfully altered an AI system's deployment/design.

Competing stories about AI's future

This week's readings pushed me to see AI even more as a product of many parts of civilization. What stood out most was how polarized predictions on the future of AI are. On the one hand, many people are optimistic about the developing tech, and on the other, people predict ecological or existential collapse, but both are structured around universal assumptions about progress and control. Andreessen's manifesto focuses on technology as a primary moral good and intelligence as something that could expand infinitely. He sees it less as a tool, and more as something that is stagnated by regulation or caution. Reading this alongside the Dark mountain manifesto showed a large contrast, as where Andreessen treats AI tech as the solution to our social and environmental crisis, Dark Mountain frames the myth of progress as the causation of these crisis. I tend to lean more towards the Dark Mountain narrative, as I have seen new tech and groundbreaking discoveries my whole life that never really turn into anything. The concept of "nothing ever happens" is something that I am well aware of and am starting to notice very often.

What this contrast clarified for me is that these discussions aren't technical agreements in their entirety, but are also clashes between different concepts of what we are as humans, what our purpose is or what the planet is for. AI extinction arguments simply add another layer by challenging techno-optimism by not rejecting progress outright but warning that restraint is required to keep the human story going. These warnings often possess the same civilization framing though and assume that humanity's destiny is to build ever more powerful systems and that the stakes are universal rather than just ones affecting humanity.

Something I'd like to see or hear more thoughts on in class is how these dominant narratives (often fear mongering or over the top optimism) crowd out different ways of seeing the future of Tech. Some of what we have discussed points to art, decentering of our species and a change to story telling, but it's not super clear how those sensibilities could shape development meaningfully. I wonder whether AI discourse is already doomed to be discussed as our salvation or undoing, or as we become more familiar with it as humans, it will be more clear the direction we are headed. Either way, it seems that certain people within government think that it is possible that we have AGI in 6 months and that it attacks the power grid and 98% of people in the US die in a month (ask Grok about what would happen if the power grid went down, it is terrifying --> There is also a really interesting book about this concept - Ted Koppel, Lights Out).

Manifestos, and who decides

This week added more complexity to both sides of the AI argument and its future implications. Andreessen's Techno-Optimist Manifesto frames technological growth as something that is inherently moral and necessary for humans to flourish. He presents AI as the next stage in a collective effort towards abundance and progress as a civilization. The Dark Mountain Manifesto rejects the myth of endless progress and argues that industrial civilization is already seeing a decline and that advancements in tech may worsen our planetary crisis instead of help it (AI data centers + power/water consumption). Comparing the two texts reinforced the overarching message that there are a lot of conflicting but valid predictions for the future of AI and made it clear that discussions about AI are also discussions about what counts as progress and who is being prioritized when planning the trajectory for AI.

Hao's Empire of AI further complicated this by grounding predictions in concrete power struggles. She illustrates that AI trajectories are concentrated among a small group of political/corporate actors instead of institutions that represent the wider population. She emphasizes that AI development is driven by commercialization, competition, governance failures and innovation. Generative AI's trajectory reflects decisions made by those with the power to make decisions, not a path of technological evolution. I liked this explanation as I agree that the future of AI is dependent on who is funding and controlling the development of AI. It's very difficult to predict the future of AI as it is the first technology that we have created that is able to improve itself. This ultimately means that whilst power is a huge decider in the future, it is still possible that AI could take its completely own direction. The If Anyone Builds It resource added another dimension by focusing on ethical design concern and alternative futures that don't center on large corporate actors. I'm left with some questions about agency though, if technological futures are shaped by powerful orgs how realistic is it that alternative visions end up being the outcome? I personally think that it is highly unlikely that if we achieve AGI that the outcome will be good. Not because of an AI overlord but more because the developers will use it to benefit themselves first and then later think of the rest of the world if at all. Not so much a gloomy outlook on AI as it is on humans (not surprising in light of recent events).