Talk:Generative AI

Revision in the hardware/software and related sections

Myself, @Wu001922, and @Chet19 were looking at the "Software and hardware" section, and it looks like there might be some room for additions with more recent work on video generation systems, particularly approaches that combine diffusion models with transformer-based architectures. This seems like a less discussed piece of the article, that could be added to as further advancements keep being introduced. We also noticed that the "Concerns" section feels a bit fragmented and could be clearer being grouped into just a few broader categories (e.g., technical limitations, ethical risks, environmental impact), which would better match the way other articles seem to discuss it. Additionally, with very recent developments in the law section a potential addition in recent legislation (e.g., 2026 updates related to generative media and copyright.)

Just wanted to get a sense of whether this direction sounds reasonable before making larger edits.

~~~~ Long0727 (talk) 02:52, 19 March 2026 (UTC)

For the video-generation and legislation content, you are free to propose bold edits, but beware that it should be notable (important-enough to be mentioned), well-sourced and understandable to a non-specialist audience.

I agree that the "Concerns" section feels a bit fragmented. "Ethical risks" seems quite synonymous with "Concerns", which is already the section title; and it's unclear to me which of the subsections would fit in "technical limitations" besides the "Content quality" subsection. But maybe we could have a "Misuse" subsection that includes things like "Cybercrime", "Illegal imagery", "Deepfakes", "Information laundering", "RAG poisoning", and "Academic honesty". Perhaps "Use in journalism" could be a subsection of "Content quality". Let me know if you have another idea or opinion. Alenoach (talk) 04:11, 22 March 2026 (UTC)

In addition to what Alenoach said, I want to emphasize the importance of reliable sources, specifically independent sources. I recommend reading the Wikipedia essay WP:BACKWARDS. To briefly summarize, resist the temptation to start by adding information you personally know. Instead use what you know to find and summarize additioanl sources. If reliable sources (no pre-prints, please) explain this specific aspect of video generation, think about how you can summarize those sources. If sources don't emphasize a specific detail that you think is important, consider that you may be wrong about how important that detail actually is... or look for more sources, but make sure not to get lost in the weeds.

It's also worth mentioning WP:LLM. Wikipedia is intended to be a project for humans, and we are generally not interested in chatbot generated content. Grayfell (talk) 04:54, 22 March 2026 (UTC)

CUT Artificial Intelligence Labs design Report

CUT Artificial Intelligence Labs design Ivuma (talk) 20:24, 29 March 2026 (UTC)

CSCI 5125 Group Project: 2026 Technical Updates

Hello! I am Justin (User:Wu001922), working with Zachary (User:Long0727) and Cheston (User:chet19) for our CSCI 5125 project. We plan to update this article today to include recent 2026 developments in video generation and restructure the 'Concerns' section for better narrative flow. We will be making these edits incrementally. Feedback from the community is welcome! Wu001922 (talk) 23:45, 29 March 2026 (UTC)

AI and Bias

Hello! My name is Avery (ashelham), and I wanted to check and make sure it was okay to add some more information on AI's model being focused on the population's average, which is already discussed to impact minorities, but doesn't touch on disabled people. I'm also going to add some positive impacts AI has had to be unbiased towards it. I'll be adding information gradually and I would love to have feedback if any of you have it. Thank you! Ashelham (talk) 01:00, 1 May 2026 (UTC)

There could be a sentence on it, but if you want to add more detailed content, the article on generative AI may be too broad, it's probably better to add it to an article like Algorithmic bias or Fairness (machine learning). Alenoach (talk) 17:36, 1 May 2026 (UTC)

Thank you! Ashelham (talk) 17:58, 1 May 2026 (UTC)

Alternatively, Ethics_of_artificial_intelligence#Algorithmic_biases could be a good place to cover it, maybe the added content should be moved to a new subsection there. Alenoach (talk) 19:19, 1 May 2026 (UTC)

Ashelham, I agree with you overall. Larry068 (talk) 00:25, 3 May 2026 (UTC)

Wiki Education assignment: Technology and Culture

This article was the subject of a Wiki Education Foundation-supported course assignment, between 15 January 2026 and 1 May 2026. Further details are available on the course page. Student editor(s): Ashelham (article contribs). Peer reviewers: Bhanday, Daisy Kwon.

— Assignment last updated by Ashelham (talk) 18:23, 1 May 2026 (UTC)

COI edit request: § Bias generalises to image generation systems from one news report about one product

I have a conflict of interest here and am requesting rather than making this edit.

The issue: In § Bias the article states "Similarly, image generation systems prompted with terms such as "a photo of a CEO" have been observed to disproportionately generate images of white male individuals when trained on biased datasets." The only source attached is a July 2022 NBC News report by Jake Traylor about one product, OpenAI's DALL·E 2. The sentence generalizes to image generation systems as a class, and a single news report about a single system does not carry that. The section is a short list of individual examples with no peer-reviewed survey cited anywhere in it. There is no maintenance template in this article, so this is a sourcing defect in the text itself and nothing more.

Proposed change: Keep the existing NBC News reference and add a second one to the same sentence, then add one sentence after it:

Similarly, image generation systems prompted with terms such as "a photo of a CEO" have been observed to disproportionately generate images of white male individuals when trained on biased datasets.<ref>{{cite web |url=https://www.nbcnews.com/tech/tech-news/no-quick-fix-openais-dalle-2-illustrated-challenges-bias-ai-rcna39918 |title=No quick fix: How OpenAI's DALL·E 2 illustrated the challenges of bias in AI |author=Jake Traylor |date=July 27, 2022 |website=NBC News |access-date=January 26, 2024}}</ref><ref name="Ferrara2023">{{cite journal |last1=Ferrara |first1=Emilio |title=Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies |journal=Sci |year=2023 |volume=6 |issue=1 |page=3 |doi=10.3390/sci6010003}}</ref> A 2023 survey of fairness and bias in artificial intelligence treats bias in generative systems as a category separate from bias in classification and prediction, and reviews mitigation strategies for it.<ref name="Ferrara2023" />

Source: Ferrara, Emilio (2023). "Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies". Sci 6(1): 3. doi:10.3390/sci6010003 {{doi}}: unflagged free DOI (link). It is a survey rather than a measurement study, so it supports the class-level framing and the existence of the failure mode across text-to-image systems; the specific "a photo of a CEO" prompt and the DALL·E 2 result stay with the NBC News reference, which is why I am proposing the survey alongside it rather than in place of it.

Disclosure: This citation is to my own work (Ferrara, sole author). See User:Emilio Ferrara.

Happy for this to be declined or reworded; I will not make the edit myself. Emilio Ferrara (talk) 04:29, 27 July 2026 (UTC)