Is AI good or bad for the society?

[ernbilen]

In your comment, take a clear position (that you were assigned to) and support it with well-reasoned arguments. Be specific in explaining the mechanisms through which AI creates benefits or harms (for example, productivity, labor markets, decision-making, or information quality). When possible, support your claims with evidence from credible sources and cite them appropriately. Avoid vague statements and focus on concrete examples or empirical findings. If you are in the position that it is harmful, suggest ways to improve it.

[marinmoore]

Although it is the assignment, I will not pretend that artificial intelligence can be labeled as diametrically "good" because it undeniably has the potential for both good and harm, and a discussion about the ethics of AI informed by our Philosophy of Data course cannot ignore such nuance. When artificial intelligence is used ethically (if that is possible given the environmental and human effects of data centers' electronic waste production, water usage, and reliance on unethical mining practices), it has the potential to benefit society. One article from California Miramar University enumerates some beneficial applications of AI: it can assist doctors in diagnosing diseases and suggesting treatment plans, give farmers suggestions for optimal watering or planting schedules, create personalized education plans for students, streamline product manufacturing by simulating prototypes, or identify and respond to cybersecurity threats, all purportedly with more speed and precision than possible for humans alone. However, these reported benefits are overwhelmingly focused on the developed world, whose citizens have access to AI.

[BenBrandt1]

I think that AI is overall a general positive on society with the caveat that it is used correctly and appropriately. AI has been one of the fastest growing technologies in the past year in part because of societies instant reliance on it. Through the use of AI, we can enhance productivity drastically. What would have taken a team of people weeks to do can be accomplished in a fraction of the time by one person properly prompting. AI can also help boost learning as many students now rely on AI to help answer questions regarding assignments, or help studying for exams, or more broadly learning a new skill. Additionally, through the use of AI, we can improve accessibility for people with disabilities. AI technologies such as NLP help create products to assist those with disabilities. That being said, there are often many concerns raised around AI and it's uses. Some of the primary concerns brought about AI are privacy concerns and job displacement concerns. In terms of privacy concerns, I think putting policies in place that limit and/or prohibit the collection of personal data is a very straightforward and simple solution that limits any privacy concerns. As for job displacement, I think that rather than fighting against AI and having it replace human jobs, companies should increase training and education around AI and teach their workers how to use AI effectively to help streamline production and maximize efficiency. AI has the potential to improve society tremendously through education, healthcare, economics, and many other aspects of society. With proper regulations in place, rather than working against the natural development, regulations can be put in place to address many of the concerns. With the proper transparency and policies, AI should be seen as a tool to better society and have a general positive impact.

https://www.geeksforgeeks.org/artificial-intelligence/how-is-ai-beneficial-to-society-are-we-using-it-right/

[TommyEaton]

AI is good for society because it drives a meaningful efficiency boost in the workforce. From a data analytics perspective, the studies done surrounding productivity is difficult to ignore. Across jobs like customer support, software development and consulting among many others, workers using generative AI tools have seen their average productivity rise ranging from 5% to over 25% (OECD, 2025). Another example comes from a large-scale randomized controlled trial at a U.S. tech support firm. Where a 35% production increase was observed from the lower quartile workers in terms of productivity (California Management Review, 2025). Which means the workers who stood to gain the most from using AI actually did. AI provides companies and workers with cost-effective and instant access to relevant information they need to help facilitate on the job learning and growth to bridge the productivity gap of workers (OECD, 2025).

works cited:
Brynjolfsson, Erik, et al. “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence.” Working Paper, Nov. 2025.

Huidobro, Jaime Oliver, Roberto García-Castro, and J. Mark Munoz. “Seven Myths about AI and Productivity: What the Evidence Really Says.” California Management Review Insights, 16 Oct. 2025, cmr.berkeley.edu/2025/10/seven-myths-about-ai-and-productivity-what-the-evidence-really-says/.

Organisation for Economic Co-operation and Development. “Unlocking Productivity with Generative AI: Evidence from Experimental Studies.” OECD, 27 June 2025, www.oecd.org/en/blogs/2025/07/unlocking-productivity-with-generative-ai-evidence-from-experimental-studies.html.

[sonixtreme322]

Artificial Intelligence (AI) has been a hot topic as of the past few years. With OpenAI’s ChatGPT and DALL-E taking the world by storm in the early 2020s, its ability to generate text and images that match the whims of its prompter has notably created friction among people who’s job it is to create art and write. Even beyond the creative arts, generative AI has been used to draft court cases, help clean and refactor code, and act as customer service agents. That’s not to say that any of that is good or bad, it’s just the world we live in. Still the downsides are numerous, on an environmental, economic, educational, and personal level.

The earliest known reports of concern I remember around ChatGPT was how students were using it. According to a 2025 Pew Research Article, use of ChatGPT to help with homework doubled from 13% to 26% from 2023 to 2024. The article does elaborate on the exact use said teens feel it’s okay to use the software for (around half believed it should not be used to write essays but is okay to use for research), and while the nuances of ChatGPT’s use in any given educational field can vary the quality of the finished product, the finished product is not the only thing that matters in an educational setting. Most if not all teachers in a public educational setting know that the exact things being taught in their classroom will not be directly applicable to the outside world, and in the debate between the virtue and practical application of said education being able to write a clear 500 word essay on any given subject falls a little between both. While ChatGPT can streamline the process such that the practical application could vanish, it can never erase the virtue. I personally don’t like the word virtue, but in this context I intend to mean it in how it defines who we become as people. If some student has ChatGPT write all of their essays for them because they’d rather be playing video games, they won’t develop the ability to form any sort of nuanced take on subjects that do have real practical weight. If ChatGPT has made your arguments for why the Twenty-First Amendment should be repealed for you before, how will you ever be able to form a proper argument about it in person without it.

On that subject of education influencing who we are as people, I cannot in good conscience ignore the 2025 MIT Preprint “Your Brain on ChatGPT.” Though the authors themselves have stressed that this article had not been peer reviewed and have decried sensationalist news outlets in their reporting on it, its findings are still disappointing to anyone hoping for their fellow man to hold onto any semblance of critical thinking ability. The quote that is the most concerning comes after the part of the overview where the methods were described, “…LLM-to-Brain participants showed weaker neural connectivity and under-engagement of alpha and beta networks; and the Brain-to-LLM participants demonstrated higher memory recall, and re‑engagement of widespread occipito-parietal and prefrontal nodes, likely supporting the visual processing…The LLM group also fell behind in their ability to quote from the essays they wrote just minutes prior.” Even when taking the precautions to note the limitations of the study, especially with the limited number of participants and focus on essay writing, it still paints a bleak and corroborated picture of how ChatGPT has been affecting us. I would be remiss not to mention the tragic death of Suzanne Adams, who was murdered by her son in August 2025. A lawsuit filed by Adams’ estate alleged that conversations her son had with ChatGPT encouraged his persecutory delusions that his mother was spying on and attempting to poison him. The articles I have been able to find on the subject, while good, are solely from the news. I will link this NPR article and move on to less heavy subjects on ChatGPT.

Speaking of death, let’s not pretend that humans are the only things being murdered by AI. There’s also the planet. I’ve made my stance on the “death of the planet” clear before, but I’ll state it again: Humans are not destroying a 4.6 billion year old rock, they are destroying themselves and everything that once sustained them. Usually when climate change is brought up, the emission of greenhouse gases is the major focus. Hence the rise in popularity of solar, hydro, and wind power. Even if the practicality of all those power sources is… woeful compared to nuclear (which is also somewhat unsustainable but that ties back to the main argument here), they don’t emit greenhouse gases so they’re not directly killing us. ChatGPT is not innocent in carbon emissions, the article I’m drawing a lot of my information from for this section points to a figure of 8.4 tons of emissions per year. That’s nowhere near as concerning as its effect on water. During the training sessions for ChatGPT-3, around 700,000 liters of freshwater were used to control temperatures. Past the training sessions, a simple conversation with it can consume as much water as one water bottle. When that’s multiplied among millions of users using the software every day, that adds up. As this is all primarily for cooling and heat management, I should probably mention an economic factor in this whole thing.

The economic impact of ChatGPT, or rather, all LLMs, is one of if not the main reason why it’s an assignment of ours. While I will be focusing on ChatGPT, bear in mind that these arguments are not solely unique to it. With regard to the computers and processing required to run this software, cooling systems are not the only concern. There is also RAM, aka the processing power. When checking sites built to track its price (mainly for those who intend to build their own PCs), 16 GB of RAM has risen from around $50 in late 2024 to around $175 now. The frequent construction of many of these AI data centers is the primary reason why. For the record, it’s unlikely that these systems would be buying units of ram 16 GB at a time. Some of the most powerful units that hold 64 GB have risen from $200 in late 2024 to $1000 now. I will note that a lot of this price rise happened in late 2025, in some part due to production and supply chain shifts to accommodate this need.

That need isn’t unwarranted in the modern economy. In a recent conversation I had with my former supervisor at PCC (Alex Meyer), he mentioned how use of LLMs is frequently being used in the general software engineering world. Able to build mostly functional applications much faster than a human could ever dream of. We’ve talked about this in class too, especially around data cleaning. In large part because of that 2025 MIT preprint mentioned earlier, I personally have resisted using it, but the temptation is real. In our modern world where everything constantly has to be faster and better, giving large segments of (more often than not) “boring” labor to these models speeds things up considerably. On that labor front, it’s important to note how this stymies fresh college graduates from gaining entry level jobs. Those entry level jobs in software engineering tend to focus on that “boring” labor. It’s experience if nothing else, and now those opportunities are vanishing. Especially in this current job market.

Let’s also not forget that AI can and will make mistakes. Just this morning, NPR published an article titled “Penalties stack up as AI spreads through the legal system”. The article largely focuses on the realm of AI in law, but the last line of the first paragraph notes how AI will make these sorts of mistakes: “The most prominent case was that of the lawyers for MyPillow CEO Mike Lindell, who were fined $3,000 each for filing briefs containing fictitious, AI-generated citations.” In DATA 300, one of the key takeaways in prediction algorithms was how they actually choose what a proper response could be. Machines have never, and likely will never, understand human patterns of language as language. Rather, by getting a good guess as to what the users want, even if the response is bunk, it fits within the given parameters for what a response should look like. These sorts of problems are very common in ChatGPT answers, or really the answers of any LLM. In an educational setting, this can help clue a teacher in on the fact that their student is using AI; in a work setting, this can cause serious damage beyond what the AI user may have initially intended. If the results look good enough and no one bothers to double check, an entire organization can form wrongheaded beliefs about what direction they should move in.

This has all been, pessimistic at best. You asked us to try to argue how these problems can be fixed, I’ll try to offer solutions where I can think of any but in some cases I’ll come up blank. Claude’s source code got leaked earlier this week, there’s currently a GitHub page up documenting how this happened and the skeleton of its operations , and if open source libraries have taught us anything, it’s that people who are doing things because they love it and not just for pay will do a better job 90% of the time. The environmental cost may lower if more sustainable cooling systems are created that don’t demand such absurd power draw. Assuming the current supply and demand for RAM levels out costs for it may drop back to pre AI levels. This is all wishful thinking isn’t it?

I try to be hopeful about the current landscape surrounding any field I learn about. If only for my mental health. Before this assignment I was dimly aware of most of what I spoke about if only to have opinions about AI when brought up in conversation. The only way for AI to keep existing in its current form and me to be okay with it would be to say that all the problems I listed aren’t problems at all. Because I’m not affected by them. I don’t know if I can even say that to myself though, even for an assignment. How can anyone see that ChatGPT drove someone to murder their own mother? Which, in terms of deaths in some part caused by ChatGPT (or any LLM), is not an isolated incident. How can anyone say that having an AI write someone’s assignment for them is good for their developing brains? How can anyone say that the continued obliteration of our planets natural resources for something that’s already caused so much damage is justified? I can’t. Not unless some serious changes are made to what AI as we conceive of it even is.

[minhphuonghoang]

I take the position that AI is, on balance, a positive force because it increases productivity, and is creating more jobs than most people believe. Therefore, this comment will focus solely on AI’s impact on the job market.

AI layoffs (like the most recent at Oracle ) get a lot of attention, but they don’t tell the full story for the whole market. I believe what’s actually happening is job restructure, not net job destruction.

Data shows that more companies are creating jobs with AI than cutting them: about 77% of firms report AI-driven job creation, vs. 46% reporting cuts, and most say the overall impact is positive. source 1

At the same time, AI is already generating entirely new roles: hundreds of thousands of jobs in areas like AI operations, data work, and system oversight. source 2

The mechanism is simple: AI removes repetitive tasks, which lowers costs and increases output. When companies can produce more, they need more people in complementary roles (managing systems, making decisions, working with clients). That’s why even when layoffs happen, they’re often paired with hiring in new areas.AI doesn’t eliminate work, it makes jobs more valuable and shifts them toward higher-skill, more human tasks. source 3

So cases like Oracle aren’t proof AI is bad. They’re examples of short-term adjustment. The real trend is that AI is expanding what workers can do and creating new types of jobs faster than it removes old ones.

[emdiam]

While AI is incredibly beneficial to those in a number of careers and areas, especially data analytics, we cannot ignore the hard it is causing. Most notably, AI is incredibly harmful for the environment.

To start, the computational power used by AI requires a very high amount of electricity. There has been an estimated increase in power required by data centers ("temperature-controlled building(s) that houses computing infrastructure, such as servers, data storage drives, and network equipment") from 2,688 megawatts in 2022 to 5,341 megawatts in 2023, almost doubling the previous year. The energy production required by these centers leads to more a significant increase in carbon dioxide emissions, a leading cause of climate change. Although AI is not the only thing data centers are used for, it is the largest cause of this increase in energy demands.MIT News

Additionally, the production and maintenance of generative AI requires large amounts of water for cooling the hardware. This leaves less water for the local communities and ecosystems. According to the Environmental and Energy Study Institute, "Large data centers can consume up to 5 million gallons per day, equivalent to the water use of a town populated by 10,000 to 50,000 people."

Overall, whether it is due to the constant improvements being made on AI models or a kid asking ChatGPT to write a song about flying turtles, there is always a severe environmental cost. Although I am not incredibly knowledgeable about the environment, I can imagine that there is some solution to reduce the environmental impact of AI, and I can only hope that if such a solution exists that it will be implemented.

[Kien123291]

Executive summary: AI is harmful to society because it functions as datafied power—turning people into legible categories, scores and targets that institutions treat as decision-ready truth. Philosophically, harm starts with classification decisions (what is measured, labelled, and rendered missing) and deepens with proxy optimisation: under Goodhart/Campbell dynamics, institutions maximise what is countable rather than what is just, and deployment feedback loops reproduce institutional priorities as future “ground truth”; the outcome is unequal error (documented in disaggregated audits of facial analysis) (Joy Buolamwini & Timnit Gebru, 2018), contextual-integrity violations when administrative data are repurposed for suspicion scoring (SyRI) (https://www.ohchr.org/en/press-releases/2020/02/landmark-ruling-dutch-court-stops-government-attempts-spy-poor-un-expert), due-process collapse when automated inferences scale faster than correction or contest (Robodebt) (https://www.monash.edu/__data/assets/pdf_file/0007/3365503/report_of-the-royal-commission-into-the-robodebt-scheme.pdf), and contested “fairness” in criminal risk scoring (COMPAS) (ProPublica, 2016) (https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing). AI also adds epistemic risk through confident “confabulation,” producing epistemic injustice when model outputs are privileged over lived testimony (NIST, 2024); it reshapes labour via automation/augmentation and algorithmic management pressures (International Labour Organization, 2023) and externalises environmental costs as data-centre electricity demand rises (International Energy Agency, 2025), while the EU AI Act reflects the view that some uses are unacceptable regardless of accuracy (European Commission, 2024). Assumptions: word count flexible (~150–250); visuals illustrative only; no access to private files. Policy implication: adopt rights-based constraints, enforce contestability and remedy, and impose clear, enforceable prohibitions for high-risk uses.

[vincenzodentamaro]

The thoughtful analysis of AI's societal harms, particularly concerning fairness violations, due-process collapses, and epistemic risks, highlights the critical need for systems that can detect and mitigate such issues at scale. Geodesia G-1 is specifically designed to address these challenges by providing enterprise-grade safety and compliance monitoring, ensuring that AI deployments align with ethical standards and regulatory requirements like the EU AI Act.

Geodesia G-1 is the enterprise-ready solution for this regulatory gap. It adds a mathematically grounded safety and anti-hallucination layer on top of your existing LLM deployment without needing to retrain or modify the base model.

Key features for compliance and deployment:

• Runtime Audit Logs & Watermarking
• FRIA (Fundamental Rights Impact Assessment) support
• Human-in-the-loop oversight workflows
• Automated Compliance Report Generation

Performance Metrics (Published):

• AUROC > 0.90 for safety detection
• AUROC > 0.96 for hallucination detection
• <35 ms total added overhead
• Coverage for 13+ regulatory frameworks (including EU AI Act)

Geodesia provides the necessary observability and safety layer to make AI models deployable in highly regulated environments with measurable trust.

Website: https://www.geodesia.ai