Talk:Artificial intelligence

"Decomputing" listed at Redirects for discussion

The redirect Decomputing has been listed at redirects for discussion to determine whether its use and function meets the redirect guidelines. Readers of this page are welcome to comment on this redirect at Wikipedia:Redirects for discussion/Log/2026 February 24 § Decomputing until a consensus is reached. मल्ल (talk) 16:10, 24 February 2026 (UTC)

Outdated?

It is striking that the paragraph beginning "High-profile applications of AI include..." mentions several Google products, but doesn't mention ChatGPT. Does this article need an overhaul? Furius (talk) 18:43, 19 March 2026 (UTC)

It needs people to add sourced suitable content. Sincerely, North8000 (talk) 19:56, 19 March 2026 (UTC)

Yes, it seems outdated. We could potentially replace "virtual assistants (e.g., Google Assistant, Siri, and Alexa)" to instead mention modern chatbots or LLMs.

I believe this sentence is also outdated: "However, many AI applications are not perceived as such: "A lot of cutting-edge AI has filtered into general applications, often without being called AI because once something becomes useful enough and common enough it's not labeled AI anymore."" This was relevant in the 2000s, when people avoided the term "AI" as it still carried stigma from the overpromises and the AI winters (see the AI effect). But nowadays, the term AI is widely used and the trend may even have reversed. Alenoach (talk) 09:19, 20 March 2026 (UTC)

Removed subsection "Substitute for human interaction"

This added subsection appears to be AI-generated (flagged as 100% AI-generated by all three AI content detectors I tried, GPTZero, ZeroGPT and copyleaks.com). I don't think this is a false positive because the initial version of the article Illusion of understanding by the same contributor ElliotSvenssonPE was also AI-generated. I checked Salles et al. and it doesn't seem to support the specific claims being made about users disclosing personal information more readily to AI systems or preferring them in situations involving embarrassment or vulnerability. Some other sources seem too old for supporting the corresponding statements. I also think using the term "illusion of understanding" isn't neutral as it assumes that there is no real understanding, and the question of whether modern chatbots understand is debated and depends on how you define "understanding" (see Talk:Illusion_of_understanding#POV_and_scope_of_the_article). Alenoach (talk) 07:12, 27 March 2026 (UTC)

I readily admit the text was 100% AI-generated, but it's not what I wouldn't have written. Even so, sourcing issues are totally real: one of the original references that Copilot gave me was a complete fake (or else it was methodically scrubbed from the Internet, including the Wayback Machine). Therefore I consider this a lesson in lazy Wikipedia contributing and won't allow the same thing to happen again.

As for the issues of POV, I think the point is a true point and well-supported by the sources, and even if that is far from a consensus it doesn't seem at all clear to me that these sources and their ideas must therefore be censored. ElliotSvenssonPE (talk) 17:25, 27 March 2026 (UTC)

Outdated sentence?

Hi @ApLundell. I saw that you disagree on whether this sentence is outdated. My understanding is that in the 2000s, using the term "AI" was often avoided, as there was still some stigma from overpromises of the AI winter. Nowadays, the term AI is broadly used. Companies are rather incentivized to indicate that their product is AI-powered. The examples of applications that the source gives, "search engines, bank software for processing transactions and in medical diagnosis" are generally not something that companies would avoid branding as AI as long as there is some machine learning involved.

Regardless, I don't think the CNN source remains applicable: it's from 2006, and the situation two decades later is very different. So I think the quote should be dropped.

What do you think? Alenoach (talk) 01:43, 30 March 2026 (UTC)

I was hoping someone else would comment.

The crux of my argument that that, as described in this article, AI is an entire field of study, and over the past few years that field has grown, but the type of thing that's regularly described as "AI" has shrunk. The aspects of the field that were incorrectly not thought of as AI when that source was written are still not thought of as AI. If anything we've only added to the bucket of things that computer scientists consider to be with within the AI field, but are not normally described that way. ApLundell (talk) 02:32, 31 March 2026 (UTC)

As this is only given half a paragraph in the body of the article, placing it in the lead is premature. This should be more clearly explained with better sources and without editorializing and without unattributed quotes before being re-added to the lead. Being out-dated does matter, per WP:AGEMATTERS, but sourcing, editorializing/WP:OR, and weight issues also need to be cleared. Grayfell (talk) 09:26, 31 March 2026 (UTC)

The CNN source seems fine. It is still true that there are many previous, more commonplace AI systems that often don't get labeled as AI because people often think AI has to be newer, more cutting-edge. This is in part because a common thinking is what Alenoach describes, where people mistakenly think that something is "AI as long as there is some machine learning involved," when reality is that there are many types of AI beyond and before machine learning. So, yes, the CNN source as well as the idea attributed to it could be kept, could be supported by other more recent sources, and as Grayfall says could likely be expanded in the body of the article before considered for the lead. I honestly would have removed the list of products (Netflix, Alexa, Waymo, etc.) from the lead before I removed this description that there are common misperceptions of what is and isn't AI. Asparagusstar (talk) 22:32, 31 March 2026 (UTC)

There's too much WP:OR in the article, and in the wider topic on Wikipedia. If it's still true now, we need to start from reliable, modern sources. We cannot start from what we know to be "true" and work backwards. That's too sloppy.

The opinions of Nick Bostrom in 2006 would, at bare minimum, need to be contextualized as the opinions of Nick Bostrom in 2006, not presented without attribution as a simple statement of fact, and especially not in the lead. Grayfell (talk) 01:01, 1 April 2026 (UTC)

Absolutely agree. Should have been cut a decade ago. ---- CharlesTGillingham (talk) 09:35, 8 April 2026 (UTC)

Wiki Education assignment: Composition II LS

This article was the subject of a Wiki Education Foundation-supported course assignment, between 24 March 2026 and 15 May 2026. Further details are available on the course page. Student editor(s): Natusiamaczek (article contribs).

— Assignment last updated by Ostudent (talk) 22:18, 9 April 2026 (UTC)

Should this article have a more detailed overview of symbolic artificial intelligence?

@Alenoach: Because this article hardly mentions any algorithms used in automated reasoning, I added several excerpts about these algorithms (which have been removed).

Should this section describe these algorithms in more detail? It briefly mentions forward chaining and backward chaining, but fails to mention other important algorithms like conflict-driven clause learning and unification. Jarble (talk) 16:31, 5 May 2026 (UTC) Jarble (talk) 15:42, 5 May 2026 (UTC)

The "Techniques" section is a dense summary of the existing techniques. And the section's coverage of logic and state space search is already significant relative to the overall field, which has become dominated by machine learning. There is also already a sentence that roughly illustrates the concepts of forward and backward chaining and provides links: "In the case of Horn clauses, problem-solving search can be performed by reasoning forwards from the premises or backwards from the problem."

I'm concerned that concepts like unification and the DPLL algorithm are difficult to understand (especially for people who don't have a technical background) and not helpful for non-expert readers interested in things like how ChatGPT and other modern AI models work.

It seems better to add this kind of technical details to more specific articles, like maybe inference engine, symbolic artificial intelligence or expert system. Not only would this be more directly notable to the topic, the audience also likely has more interest in and ability to understand these topics. Alenoach (talk) 22:07, 5 May 2026 (UTC)