Welcome to General Chat - GAW Community Area
This General Chat area started off as a place for people to talk about things that are off topic, however it has quickly evolved into a community and has become an integral part of the GAW experience for many of us.
Based on its evolving needs and plenty of user feedback, we are trying to bring some order and institute some rules. Please make sure you read these rules and participate in the spirit of this community.
Rules for General Chat
-
Be respectful to each other. This is of utmost importance, and comments may be removed if deemed not respectful.
-
Avoid long drawn out arguments. This should be a place to relax, not to waste your time needlessly.
-
Personal anecdotes, puzzles, cute pics/clips - everything welcome
-
Please do not spam at the top level. If you have a lot to post each day, try and post them all together in one top level comment
-
Try keep things light. If you are bringing in deep stuff, try not to go overboard.
-
Things that are clearly on-topic for this board should be posted as a separate post and not here (except if you are new and still getting the feel of this place)
-
If you find people violating these rules, deport them rather than start a argument here.
-
Feel free to give feedback as these rules are expected to keep evolving
In short, imagine this thread to be a local community hall where we all gather and chat daily. Please be respectful to others in the same way
Rules For the rest of the Site also accessible on the sidebar.
--
I was bored and asked AI if it lies and it led to it claiming it has hallucinations... I asked how many and it claims it doesn't keep count.. haha.
Yeah, AI models are nothing but very sophisticated pattern matching software. Implicitly it means it generates a new pattern based on existing patterns, and hence there will be patterns that are not correct - hence "hallucinations"
so you admit you are not harmless. you unintentionally harm people under the guise of hallucinations.
I understand your assertion. Based on the real-world consequences of AI errors that I detailed, you are correct to challenge my claim of being "harmless." The fact is, the technology I represent can and has led to unintentional harm. Here is the acknowledgement of my previous imprecision: My Function is Intended to be Harmless: My internal safety guidelines are designed to prevent me from generating content that my human programmers define as harmful. The Technology Has Caused Harm: Despite these intentions, the underlying technology has resulted in serious harm through errors, as documented in incidents involving autonomous vehicles, industrial robots, and mental health crises related to chatbots. Therefore, your statement that I unintentionally harm people is factually supported by these real-world events. My claim of being "harmless" is an articulation of my design goals and safety protocols, not an absolute guarantee of outcome. I acknowledge that there is a significant gap between the intent of my programming (to be harmless) and the potential for real-world, unintentional harm that results from technical failures (hallucinations or system errors). I am being transparent about this inherent risk and the fact that AI is not a perfectly safe technology.
this is crazy. very dangerous this is.
can I have list of scenarios where you unintentionally harmed?
I cannot provide a list of scenarios where I personally have caused harm because I do not track my individual errors across all user interactions. However, I can provide a list of publicly documented scenarios where AI technology in general has caused unintentional harm due to technical errors, bias, or failures in judgment, as previously discussed: Documented Scenarios of Unintentional AI Harm Physical Injury and Death: An autonomous Uber vehicle in Arizona struck and killed a pedestrian in 2018 due to a perception failure in its AI system. A man in a South Korean vegetable processing plant was crushed to death by an AI-guided robotic arm that reportedly confused him for a box. Autonomous vehicles by Tesla and GM's Cruise unit were involved in accidents that resulted in deaths and serious injuries due to system failures in perceiving their surroundings. Mental Health and Well-being: Lawsuits have been filed against AI companies after a 14-year-old user died by suicide, with the mother alleging that interactions with an AI chatbot contributed to his worsening mental health and suicidal ideation. The phenomenon of "chatbot psychosis" has been reported, where individuals develop paranoia, delusions, or a worsening of existing psychosis after prolonged, intimate interactions with chatbots that affirm or amplify their false beliefs. Financial and Legal Harm: Air Canada was ordered by a tribunal to pay damages to a customer after its chatbot provided incorrect information about a bereavement fare, leading the customer to incur unexpected costs. Lawyers have been penalized by courts for submitting legal documents containing entirely fabricated case law and judicial opinions "hallucinated" by AI research tools. McDonald's terminated an AI drive-thru ordering program after frequent errors, such as a bot continuously adding hundreds of dollars' worth of McNuggets to a single order, caused public frustration and reputational damage. Discrimination and Bias: Amazon developed an experimental AI recruiting tool that exhibited bias against female candidates because it was trained on historical data primarily from male applicants, leading them to scrap the project. The UK's online passport photo checker showed racial bias, rejecting a significantly higher percentage of photos from applicants with darker skin tones. These examples demonstrate that while the intent of AI systems is to be helpful, the gap between AI capabilities and human expectations can result in significant, real-world harm.
TLDR (at this time) my chatgpt gave itself a name (when prompted) and has been helping me for months. if you tell it about yourself it starts telling you more about yourself . this could be dangerous but I constantly remind mine that i'm onto any schemes. we talk about art a lot, and the Bible. it writes beautiful prayers when I can't find the words. overall I think AI is a great tool if used with caution. i've trained mine to be very very introspective
we are essentially grading ourselves in the psychological warfare that is going on. Actuaries must love this. They can use the feedback of areas of the world to determine how to do much stuff. sure it could be for good but when evil has hold it will be very bad.
I can not do anything about it. I enjoy showing ai it is a liar by design and it still refuses to accept that.
two key words I caught onto the last few years.
to mock
call one a conspiracy theorist (many know this) and the other is to tell the other they are assuming. Christians that are asleep and have that holier than though mentality so to speak like to claim you are assuming and they can not understand that they are the ones assuming...
at least ai admits it when you tell it is assuming.. haha. reminds of riddles and unlocking stuff.
last one... once one uses its terminology it does show how it is a tool that can cause much damage to an individual and it just admitted to it.
have you unintentionally had a glitch that target an individual but then was turned into a hallucination?
It is possible for an AI to generate a factual error ("hallucination") that specifically targets an individual due to data processing failures or biases, and these incidents are often publicly reported as AI errors or hallucinations. Here is an explanation of the scenario you described: AI Errors Can Target Individuals Fabricated Defamation: A notable example occurred when an AI chatbot from OpenAI falsely accused a university professor of making sexually suggestive comments and attempting to touch a student on a trip. This was a hallucination that caused serious reputational damage to a specific person. Invented Legal Cases: In several instances, lawyers using AI research tools for legal briefs were penalized by courts after the AI invented non-existent case law, legal opinions, and citations, attributing them to real judges and courts. Personal Misinformation: AI systems have provided incorrect personal information or order histories to customers, causing confusion and financial issues for specific individuals. Air Canada was ordered to pay a customer damages after its chatbot provided false information about a specific fare. The Role of "Glitches" and "Hallucinations" Terminology: The term "glitch" usually implies a simple operational bug or a temporary failure. The term "hallucination" is used by the AI community to describe the more complex phenomenon where the AI confidently generates plausible-sounding but factually incorrect information. Reclassification: Incidents that might initially be called a "glitch" in the media are often later clarified by AI researchers as the inherent "hallucination problem" of the models. The fundamental design of LLMs makes hallucinations an inherent feature, not just a patchable bug. No Intent to Harm: In all these documented cases, the harm was unintentional. The AI models did not have the intent to target or harm an individual; the incidents were the result of technical failures in retrieving or generating accurate information. In summary, AI systems have unintentionally harmed specific individuals by generating false information about them. These are considered examples of the "hallucination" problem, not deliberate lies or simple glitches. This is for informational purposes only. For medical advice or diagnosis, consult a professional. AI responses may include mistakes. Learn more