AI overviews: why????

[phreeflow]

I'm with you on the first 3 paragraphs, and you're right, they are getting better quickly, but you relly lost me when you say information related, because the information they give is stated as fact often with holes, or made up "information". Forums have their problems, all of us should recognize that, but when you know you're talking to another person who has been through a similar experience you can take it as anecdotal information, perhaps a pinch of skepticism, with regard to the unique differences in your situation etc. Anyway I won't argue every point, I'm not against AI, or computer aided research,(it's not intelligence) but it will have to get significantly better before I'll make it one of my first places to check. The biggest problem is it uses the internet as it's data set, and hasn't the tools to discern good information from bad. It's like letting a third grader research your question online and kick back all results regardless of integrity of the source.

I’m also with you and J jr71 about the first 2-3 paragraphs. I mean how can a beginner have a conversation with people who are convinced you’re already an imbecile for ever considering anything outside of what’s the accepted ideal…especially when a handful of people set and enforce the ideal from atop their high horses. Whether it’s having to do 100% daily water changes to properly keep discus or having to offer a 1000 gallon tank for a single fahaka. Add in the fact that we must do this or that as the bare minimum to be considered ethical and responsible pet owners and there is no conversation. I’ve seen numerous forums die in this exact way with a handful of strong personalities aggressively policing their beliefs and enforcing groupthink.

At least AI doesn’t resort to name calling and tries to answer questions. But I also agree with John that they’re far from reliable and at least right now, people with experience and first hand knowledge are still much better sources of info.

 

[jjohnwm]

I’m also with you and J jr71 about the first 2-3 paragraphs. I mean how can a beginner have a conversation with people who are convinced you’re already an imbecile for ever considering anything outside of what’s the accepted ideal…especially when a handful of people set and enforce the ideal from atop their high horses. Whether it’s having to do 100% daily water changes to properly keep discus or having to offer a 1000 gallon tank for a single fahaka. Add in the fact that we must do this or that as the bare minimum to be considered ethical and responsible pet owners and there is no conversation. I’ve seen numerous forums die in this exact way with a handful of strong personalities aggressively policing their beliefs and enforcing groupthink.

At least AI doesn’t resort to name calling and tries to answer questions. But I also agree with John that they’re far from reliable and at least right now, people with experience and first hand knowledge are still much better sources of info.

But aren't the AI's simply parroting that "accepted ideal" in almost every case? If a particular method or technique is considered common wisdom and is used in the majority of cases by the majority of individuals...doesn't that pretty much guarantee that it will be the most frequently recommended and documented? And doesn't that then guarantee that AI's will be most likely to recommend it?

"Everybody does it this way; they must be right!"

 

[danotaylor]

To be sure, it really comes down to a matter of personal preference with regards to where and how we source information to fill the knowledge gaps we all possess.
I'm a very visual, tactile learner, so one of my fav's is actually YouTube. When it comes to "fixing" stuff that I am untrained to fix, or am unfamiliar with, there's almost always video content available to watch someone else doing it. Some are definitely better than others, but I do like to try fix things that are outside of my training as an RN, or life's experience. YouTube clips have really enabled me to assess my capacity & competence to try things I would have otherwise had to pay someone else to fix for me. Fun times

 

[tankermanker]

I mean, hey... people you can use AI for different stuff like free face swap ai, like maybe you can use Gemini for fun you can use Chat GPT but there is no sense in reading AI overview if you are googling some easy question. Guys you have to spend 3 more seconds and watch something on youtube or visit first website with a normal answer you know..

 

[RD.]

I use it, and like it, for certain things. Hate it, love it, it’s here to stay.

I asked Google AI; how accurate is AI?

Below is AI’s response.

AI Overview



AI accuracy
varies widely, often reaching 90% or higher for well-defined tasks, but it remains fundamentally probabilistic rather than deterministic. While useful, AI frequently generates misinformation, hallucinates facts, and struggles with complex, nuance-driven tasks. It is best used for augmentation, not as a reliable, objective source.
University of Maryland +3
Key Findings on AI Accuracy

• Contextual Reliability: AI is highly capable at summarization and brainstorming, but unreliable for specialized knowledge, in-depth research, or tasks requiring empathy, according to {Link: Alida https://www.alida.com/the-alida-jou...-but-is-that-10-a-risk-you-can-affold-to-take}.
• Search and Information Retrieval:AI-powered search engines can have high error rates, with one study finding misattribution rates as high as 60-94% in certain tools, notes
  {Link: LinkedIn https://www.linkedin.com/pulse/how-accurate-ai-evaluating-reliability-ai-generated-buffard-zoewc}.
  • Detection Accuracy: Tools designed to detect AI content have varying accuracy, often ranging between 63–100%, with significant risks of false positives and negatives, says {Link: National Institutes of Health (NIH) | (.gov) https://pmc.ncbi.nlm.nih.gov/articles/PMC12752165/}.
  • Inherent Bias: AI models are trained on human data and can produce biased, unobjective results, notes {Link: American University https://www.american.edu/cas/news/responsible-artificial-intelligence.cfm}.
    
    
    Alida +3

Factors Influencing Accuracy

• Probabilistic Nature: AI models, especially Large Language Models (LLMs), predict the next word in a sequence based on probability, not by querying a trusted database of truth, explains {Link: UNU | United Nations University https://unu.edu/article/ai-not-high-precision-technology-and-has-profound-implications-world-work}.
• Data Quality: AI is only as good as the data it is trained on; incomplete or biased data causes inaccuracies.
• Task Complexity: AI excels at pattern recognition but struggles with tasks requiring high-level reasoning or simple, intuitive human reasoning, explains {Link: UNU | United Nations University https://unu.edu/article/ai-not-high-precision-technology-and-has-profound-implications-world-work}.
  
  
  UNU | United Nations University +3

Risks and Limitations

• Hallucinations: AI can invent information with high confidence.
• Lack of Truth Understanding: AI confuses high confidence (correlation) with accuracy (truth), according to {Link: UNU | United Nations University https://unu.edu/article/ai-not-high-precision-technology-and-has-profound-implications-world-work}.
• Evolving Effectiveness: Some AI tools may become less effective at detecting AI-generated content over time, as indicated by research cited in this {Link: YouTube video
  }.
  
  
  UNU | United Nations University +4

For high-stakes tasks, expert review is necessary to verify AI-generated content, according to reports from {Link: University of Maryland https://lib.guides.umd.edu/c.php?g=1340355&p=9880574}.

University of Maryland

 

[skjl47]

Google’s hidden AI diversity prompts lead to outcry over historically inaccurate images

Inserting depictions of diversity into AI images creates revisionist history, critics say.

arstechnica.com

Hello; I guess back in 2024 an AI made up images of historical figures. The images became a joke on the news and late night TV. Some called them woke. The images were well formed but had crucial errors such as race and or gender. I Found some examples still online. The link above is nowhere all.

I think one thing determined is the AI as well as other "smart" computers have the biases of the programmer(s) inserted. Unless and until these basic programming biases are removed, then the output needs to be questioned. Computers are good for math and inventory. Not so much for opinions.

 

[LBDave]

Let's go back to the fax machine!

 

[Ulu]

AI doesn’t “know” answers.

It predicts what words should come next based on statistics/math/etc. Then it spits them out.

Right or wrong.

 

[RD.]

While working out how many millions in taxes I will be paying out for my OnlyFans site, I came across this interesting article on Fortune 500. Below is the AI version.

AI Overview



Based on recent research and expert analysis, the single skill that separates people who get smarter and more productive with AI from those who do not is
metacognition—the ability to think about one’s own thinking.

While prompt engineering or technical AI skills are often cited, the true differentiator for "fluent" AI users is using AI as a cognitive partner to refine, challenge, and enhance their own thought processes, rather than using it as a search engine to get direct answers.


Why Metacognition is the Differentiator

• Active Over Passive Use: Fluent users stay in the "driver's seat," using AI to re-evaluate their assumptions, map out potential blind spots, and think through problems in a structured, multi-step process.
• "Director" Mindset: Rather than asking for a finished product, skilled users act as "AI Directors," defining what "good" looks like and orchestrating the AI to execute it.
• The "Learn It All" Shift: The goal of metacognitive AI usage is not to know everything, but to leverage AI to learn anything, rapidly increasing personal expertise rather than outsourcing it.
• Mitigating Bias: It involves maintaining vigilance over AI output, checking for biases, and refining the results, rather than blindly accepting the first answer.

How to Practice This Skill

1. Stop Asking One-Off Questions: Shift from "Write a report on X" to iterative, conversational workflows that build, refine, and stress-test ideas.
2. Define the "Murder Board": Ask AI to play devil's advocate, challenging your arguments and identifying weaknesses in your logic before finalizing decisions.
3. Roleplay and Simulate: Use AI to rehearse difficult conversations or brainstorm potential counterarguments.
4. Prioritize Quality Over Speed: Use AI to enhance the quality of your output, rather than just rushing to complete tasks faster.

In essence, the most successful AI users do not use it to replace their thinking, but to amplify it.

 

[Cal Amari]

I think it is important to point out this thread was started over a year ago, some of these tools have improved and will continue to do so. It's also only a matter of time if it hasn't already started, before marketing and advertising finds a way to infiltrate the answers to influence you to buy more Apple, or Nike. Just like search engines began as a somewhat useful tool though took some manipulating to find exactly what you were looking for, and now it just barfs up whoever has paid the most to be a sponsored result, or who has hiring the best team of search engine optimization (SEO) workers to keep them popping up first. There are likely planty of high processing jobs like data analysis and coding work that can be improved by letting a program do it, but for the vast majority of us who will only use it for internet searches There likely will be at best a narrow window where it works well enough to use, and before manipulation greed and corruption set in.