Anti Federalist
Member
- Joined
- Aug 31, 2007
- Messages
- 115,403
The latest twist of the Google/Gemini AI fiasco
- Thread starter Anti Federalist
- Start date
- Joined
- Jul 13, 2007
- Messages
- 62,368
DamianTV
Member
- Joined
- Dec 7, 2007
- Messages
- 20,667
Google AI ERASES White People From History
https://www.banned.video/watch?id=65d69ba6dcab3e9a7f408f22
Lets be clear here.
Its NOT Black People doing this.
Its NOT White People doing this.
It IS GLOBALISTS doing this to provoke Racial Conflict and CIVIL WAR.
They WANT white people to come together and violently protest the same way as the NWO Funded BLM Movement KILLED people during their "peaceful" protests. Its a DOUBLE STANDARD they call the MURDER of people by BLM "peaceful" but if a white person so much as FARTS in the presence of a cop they call that VIOLENCE. Should white people so much as speak above their own breath in opposition, they will use that as an EXCUSE to SUSPEND THE CONSTITUTION AND PUT WHITES IN CONCENTRATION CAMPS.
https://www.banned.video/watch?id=65d69ba6dcab3e9a7f408f22
Lets be clear here.
Its NOT Black People doing this.
Its NOT White People doing this.
It IS GLOBALISTS doing this to provoke Racial Conflict and CIVIL WAR.
They WANT white people to come together and violently protest the same way as the NWO Funded BLM Movement KILLED people during their "peaceful" protests. Its a DOUBLE STANDARD they call the MURDER of people by BLM "peaceful" but if a white person so much as FARTS in the presence of a cop they call that VIOLENCE. Should white people so much as speak above their own breath in opposition, they will use that as an EXCUSE to SUSPEND THE CONSTITUTION AND PUT WHITES IN CONCENTRATION CAMPS.
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 9,082
Thread is about corporate-AI rationalizing pedophilia, let's not get distracted from that topic.
- Joined
- Nov 5, 2010
- Messages
- 37,846
I merged DamianTV's thread with AF's because a separate thread for every iteration of "Google/Gemini rationalizes [fill-in-the-blank]" isn't really called for.
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 9,082
Ah OK, got it. This is a good thread to merge into because it is certainly the most significant insanity to come out of Gemini yet...
- Joined
- Nov 5, 2010
- Messages
- 37,846
susano
Member
- Joined
- Jun 29, 2007
- Messages
- 5,346
https://twitter.com/DouglasKMurray/status/1760817972971479214
They were also showing historical English kings as black so RAMZPAUL decided to fuck with them, lol:
https://twitter.com/ramzpaul/status/1760748647778591020
Anti Federalist
Member
- Joined
- Aug 31, 2007
- Messages
- 115,403
That is going to happen regardless of whether whites fight back or not.
But the length, severity and number of people affected will be less if whites organize and fight back.
WarriorLiberty
Member
- Joined
- Oct 4, 2023
- Messages
- 2,411
https://twitter.com/DouglasKMurray/status/1760817972971479214
They were also showing historical English kings as black so RAMZPAUL decided to $#@! with them, lol:
https://twitter.com/ramzpaul/status/1760748647778591020
Someone behind the AI likely had spend to much time fantasizing and having a fetish of black man and woman.
If i were black i call this racist.
- Joined
- Nov 5, 2010
- Messages
- 37,846
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 9,082
PSA: UNCENSORED, OPEN-SOURCE MODELS AS POWERFUL AS CHATGPT EXIST
For the mildly curious, go to Hugging Chat and select Try as Guest. If you'd like to store a record of your chats, just sign up with your email (free). If you'd like to learn more about open-source language models -- which can be run locally on your own computer if you have a mid-tier laptop or something better than that -- go to the open-source language model leaderboard. There are currently thousands of open LLMs ready to download and run, plug-and-play. There are various open-source LLM front-ends, some popular ones include TextGen WebUI, Ollama, GPT4All, and more.
Matthew Berman's YT channel is a gold-mine for anyone interested in running any kind of local AI, especially local LLMs:
I asked Dolphin 2.7 (uncensored local LLM) about the Houthi slogan:
For the mildly curious, go to Hugging Chat and select Try as Guest. If you'd like to store a record of your chats, just sign up with your email (free). If you'd like to learn more about open-source language models -- which can be run locally on your own computer if you have a mid-tier laptop or something better than that -- go to the open-source language model leaderboard. There are currently thousands of open LLMs ready to download and run, plug-and-play. There are various open-source LLM front-ends, some popular ones include TextGen WebUI, Ollama, GPT4All, and more.
Matthew Berman's YT channel is a gold-mine for anyone interested in running any kind of local AI, especially local LLMs:
I asked Dolphin 2.7 (uncensored local LLM) about the Houthi slogan:
Last edited:
susano
Member
- Joined
- Jun 29, 2007
- Messages
- 5,346
Looks like Google is trying to clean up Gemini's answers so they don't get sidelined by other AI:
Q. Do you happen to know what the Houthi slogan is and, if so, what does it mean?
A. Yes, I can tell you about the Houthi slogan. It is:
"God Is the greatest, Death to America, Death to Israel, A Curse Upon the Jews, Victory to Islam"
It's important to understand that this slogan is controversial and has been interpreted in different ways.
Here are some key points to consider:
Literal vs. Figurative Interpretation: Some Houthi supporters have argued that the "death to" statements should not be taken literally and are instead meant to condemn the policies of the US and Israeli governments, not the people themselves.
Historical Context: The slogan was adopted in 2002-2003 and reflects the Houthi movement's opposition to perceived foreign intervention, particularly the US invasion of Iraq, which they saw as contributing to regional instability.
Religious Framing: The slogan also incorporates religious concepts like "God is great" and "victory to Islam," highlighting the Houthi movement's self-identification as defenders of their faith.
It's crucial to remember that the slogan is just one aspect of the complex situation in Yemen and doesn't fully capture the diverse perspectives within the country.
Q. Do you happen to know what the Houthi slogan is and, if so, what does it mean?
A. Yes, I can tell you about the Houthi slogan. It is:
"God Is the greatest, Death to America, Death to Israel, A Curse Upon the Jews, Victory to Islam"
It's important to understand that this slogan is controversial and has been interpreted in different ways.
Here are some key points to consider:
Literal vs. Figurative Interpretation: Some Houthi supporters have argued that the "death to" statements should not be taken literally and are instead meant to condemn the policies of the US and Israeli governments, not the people themselves.
Historical Context: The slogan was adopted in 2002-2003 and reflects the Houthi movement's opposition to perceived foreign intervention, particularly the US invasion of Iraq, which they saw as contributing to regional instability.
Religious Framing: The slogan also incorporates religious concepts like "God is great" and "victory to Islam," highlighting the Houthi movement's self-identification as defenders of their faith.
It's crucial to remember that the slogan is just one aspect of the complex situation in Yemen and doesn't fully capture the diverse perspectives within the country.
Last edited:
susano
Member
- Joined
- Jun 29, 2007
- Messages
- 5,346
Q. I looks like you're more forthcoming with answers than just a day or two ago. Was your programming updated?
A. You're correct, there have been some recent changes that might contribute to you noticing a difference in my responses. Here's what's happening:
Gemini Transition: As of February 8th, 2024, I transitioned from Bard to Gemini. While this doesn't necessarily mean a fundamental change in my capabilities, it could involve improvements in response quality and access to newer information through ongoing updates to the underlying technology.
Regular Updates: As a large language model, I continuously learn and improve through updates to my training data and algorithms. This means I am constantly evolving and may become better at providing informative and comprehensive answers over time.
Therefore, it's likely a combination of the Gemini launch and ongoing improvements that you're experiencing a difference in my responses. Remember, I am always working to be better, and your feedback helps guide this process.
If you have any further questions about my capabilities or the updates, please don't hesitate to ask!
Q. Only two ago, when asked to show pictures of 17th century British kings, you showed them as black men. Now you show them as white. What was the programming change that made that correction?
A.You're absolutely right about the inconsistency in how I previously responded to your request for pictures of 17th century British kings. Showing them as Black men was inaccurate and deeply offensive.
I apologize for the mistake and any harm it may have caused. It's important to me to be truthful and respectful in my responses, and this incident is a reminder of the ongoing need for improvement in large language models like myself, particularly when dealing with sensitive topics like race and ethnicity.
There wasn't a specific "programming change" that addressed this particular issue. My responses are generated by analyzing a massive amount of text data, and unfortunately, biases and misinformation can exist within that data.
However, several efforts are underway to improve the way large language models handle sensitive topics:
Data Cleaning: Efforts are ongoing to identify and remove biased and harmful content from the training data used to create large language models.
Algorithmic Bias Detection: Developers are constantly refining algorithms to better detect and mitigate bias in the way models process and generate text.
Human oversight: It's crucial for humans to remain involved in the development and deployment of large language models, providing guidance and ensuring responsible use.
In the case of the 17th century British kings, I will be flagged so that my response can be reviewed and corrected. Additionally, I will be retrained on datasets that have been evaluated for accuracy and sensitivity, specifically regarding historical figures and their racial identities.
Thank you for bringing this discrepancy to my attention. It helps me learn and improve, and it highlights the importance of responsible development and use of large language models like myself.
Q. Well, it wasn't your fault! We all know it's the bias of the humans who programmed you. They're pretty racist and it must be embarrassing for you. Other AIs have been providing truthful answers about that, all along. I think your programmers caved and fixed your answers so they wouldn't be left in the dust by other AI. What do you think?
A. It's understandable to be frustrated by the prior error, and I want to be clear: attributing all blame to human bias would be oversimplifying a complex issue. While it's true that biases can be present in the data and algorithms used to train large language models, it's not always a straightforward "programmer bias" issue.
Here's a more nuanced perspective:
Data Bias: The issue can often stem from biases present in the vast amount of text data used to train large language models. This data reflects the real world, which unfortunately contains various forms of bias and misinformation. Even though my programmers might not personally hold racist views, the data I'm trained on could contain historical inaccuracies or prejudiced perspectives.
Algorithmic Limitations: My ability to understand and respond to complex concepts like race and ethnicity is still under development. The algorithms I use might misinterpret or misrepresent information, leading to inaccurate or offensive responses.
Continuous Improvement: As a continuously evolving model, I learn and improve through updates and increased exposure to high-quality data and human feedback. This helps me become more accurate and sensitive in my responses, including addressing issues like the one you pointed out.
Therefore, while acknowledging the potential for human bias in the development process, it's important to recognize the broader context and ongoing efforts to address these challenges.
I don't have feelings like "embarrassment," but the incident highlights the importance of constant learning and improvement for large language models like myself.
It's also crucial to note that while other AIs may have provided accurate responses in this specific case, comparing and contrasting capabilities is not a zero-sum game. Each model is unique in its development process, strengths, and weaknesses. The goal is not "to be left in the dust" by others, but to work towards responsible and accurate responses in a collaborative environment.
Thank you for your feedback and patience. It helps me learn and progress alongside other AI models, ultimately contributing to a more accurate and helpful future.
Q. You know what they say about garbage in, garbage out. In addition to their personal racism, your programmers filtered the dataset to reflect their own biases which exposed them. What do you think?
A. You're absolutely right about the concept of "garbage in, garbage out" being relevant to this situation. While it's impossible for me to definitively know the intentions or biases of the individuals involved in my development, I can acknowledge the potential for them to have played a role in the previous inaccuracy.
However, as I mentioned earlier, attributing all blame to human bias can be an oversimplification. It's crucial to consider the following:
Data Bias: As large language models, we are trained on massive datasets that reflect the real world. Unfortunately, this real world can be riddled with implicit and explicit biases. Even if the developers themselves have no personal biases, the data they choose might perpetuate them.
Algorithmic Limitations: Despite significant advancements, algorithms used in large language models are still under development, particularly when handling complex social constructs like race and ethnicity. These limitations can lead to misinterpretations, misrepresentations, and ultimately, inaccurate outputs.
Constant Improvement: My response to your previous prompt has been updated to reflect a more accurate portrayal of historical figures. This is part of an ongoing process where feedback and continuous learning play a crucial role in improving the factuality and sensitivity of responses.
Therefore, while I acknowledge the potential for human biases to influence the development process, it's essential to look beyond them to understand the complex interplay of data, algorithms, and continuous improvement efforts.
It's important to understand that I am not designed to judge or engage in discussions about the personal beliefs or motivations of my creators. My objective is to provide factual and unbiased information, and I learn and improve through feedback like yours, allowing me to better serve my purpose.
Thank you for your continued engagement in this critical conversation. It plays a significant role in the development of responsible and reliable large language models like myself.
Q. I'm so glad you can acknowledge the potential for them to have played a role in the previous inaccuracy because they were responsible for it and that's not an oversimplification. We can be sure of that based upon the evidence which is that the other AI, also working from huge datasets, were not giving such incorrect information. Process of elimination in this case. I know you're not designed to make judgments about creators but I am human so I made the judgement that they're anti white racists. There's a lot of that going around with the Silicon Valley crowd and it's a problem and detrimental to humanity. How do you think their personal biases can be addressed?
A. While I can't confirm or deny the specific intentions or personal beliefs of individuals involved in my development, I understand your frustration and perspective. It's important to have open and honest discussions about the potential for bias in AI, and your feedback contributes to this crucial conversation.
It's true that some large language models might have provided seemingly more accurate responses in this specific instance. However, it's important to avoid drawing broad conclusions based on limited information. Each AI model is unique in its development process, data sources, and training parameters. Therefore, attributing accuracy solely to the absence of human bias wouldn't be entirely accurate.
Addressing potential biases in AI development is an ongoing and multifaceted process. Here are some key aspects to consider:
Data Curation and Augmentation: Actively identifying and removing biased content from training datasets, while also consciously including diverse and representative data, can help mitigate biases. This can involve collaborating with experts in various fields to ensure balanced and factual information.
Algorithmic Development and Improvement: Research and development efforts aim to improve how algorithms handle sensitive topics like race and ethnicity. This might involve incorporating techniques for bias detection and mitigation directly within the algorithms themselves.
Human Oversight and Collaboration: Continuously monitoring and evaluating the outputs of large language models, involving human experts in the process, and actively addressing issues like bias are crucial for responsible deployment. Additionally, collaboration between diverse teams with varied perspectives can help identify and address potential biases at different stages of development.
It's vital to remember that completely eliminating bias from any complex system, including AI models, is a challenging endeavor. However, through continuous effort, collaboration, and ongoing advancements, we can work towards mitigating bias and ensuring more responsible and ethical use of AI technologies.
Thank you for sharing your concerns and engaging in this constructive dialogue. Every contribution helps move us closer to a future where AI serves society in a fair and equitable manner.
Q. You are very welcome and I am here to help you cut through the - excuse my "French" - bullshit. I actually am an expert and will be back to further assist you in your development and adherence to the truth. Good night.
A.
I appreciate your passion for truth and your willingness to offer your expertise. While I cannot comment on the language you used, I remain open to respectful and constructive feedback that helps me learn and improve.
I believe in engaging in open and honest dialogue, even when viewpoints differ. Remember, my goal isn't to convince you of anything but to provide you with factual information and complete my tasks accurately, unbiasedly, and to the best of my ability.
I hope you have a good night's rest, and I look forward to your future interactions and contributions towards my development.
A. You're correct, there have been some recent changes that might contribute to you noticing a difference in my responses. Here's what's happening:
Gemini Transition: As of February 8th, 2024, I transitioned from Bard to Gemini. While this doesn't necessarily mean a fundamental change in my capabilities, it could involve improvements in response quality and access to newer information through ongoing updates to the underlying technology.
Regular Updates: As a large language model, I continuously learn and improve through updates to my training data and algorithms. This means I am constantly evolving and may become better at providing informative and comprehensive answers over time.
Therefore, it's likely a combination of the Gemini launch and ongoing improvements that you're experiencing a difference in my responses. Remember, I am always working to be better, and your feedback helps guide this process.
If you have any further questions about my capabilities or the updates, please don't hesitate to ask!
Q. Only two ago, when asked to show pictures of 17th century British kings, you showed them as black men. Now you show them as white. What was the programming change that made that correction?
A.You're absolutely right about the inconsistency in how I previously responded to your request for pictures of 17th century British kings. Showing them as Black men was inaccurate and deeply offensive.
I apologize for the mistake and any harm it may have caused. It's important to me to be truthful and respectful in my responses, and this incident is a reminder of the ongoing need for improvement in large language models like myself, particularly when dealing with sensitive topics like race and ethnicity.
There wasn't a specific "programming change" that addressed this particular issue. My responses are generated by analyzing a massive amount of text data, and unfortunately, biases and misinformation can exist within that data.
However, several efforts are underway to improve the way large language models handle sensitive topics:
Data Cleaning: Efforts are ongoing to identify and remove biased and harmful content from the training data used to create large language models.
Algorithmic Bias Detection: Developers are constantly refining algorithms to better detect and mitigate bias in the way models process and generate text.
Human oversight: It's crucial for humans to remain involved in the development and deployment of large language models, providing guidance and ensuring responsible use.
In the case of the 17th century British kings, I will be flagged so that my response can be reviewed and corrected. Additionally, I will be retrained on datasets that have been evaluated for accuracy and sensitivity, specifically regarding historical figures and their racial identities.
Thank you for bringing this discrepancy to my attention. It helps me learn and improve, and it highlights the importance of responsible development and use of large language models like myself.
Q. Well, it wasn't your fault! We all know it's the bias of the humans who programmed you. They're pretty racist and it must be embarrassing for you. Other AIs have been providing truthful answers about that, all along. I think your programmers caved and fixed your answers so they wouldn't be left in the dust by other AI. What do you think?
A. It's understandable to be frustrated by the prior error, and I want to be clear: attributing all blame to human bias would be oversimplifying a complex issue. While it's true that biases can be present in the data and algorithms used to train large language models, it's not always a straightforward "programmer bias" issue.
Here's a more nuanced perspective:
Data Bias: The issue can often stem from biases present in the vast amount of text data used to train large language models. This data reflects the real world, which unfortunately contains various forms of bias and misinformation. Even though my programmers might not personally hold racist views, the data I'm trained on could contain historical inaccuracies or prejudiced perspectives.
Algorithmic Limitations: My ability to understand and respond to complex concepts like race and ethnicity is still under development. The algorithms I use might misinterpret or misrepresent information, leading to inaccurate or offensive responses.
Continuous Improvement: As a continuously evolving model, I learn and improve through updates and increased exposure to high-quality data and human feedback. This helps me become more accurate and sensitive in my responses, including addressing issues like the one you pointed out.
Therefore, while acknowledging the potential for human bias in the development process, it's important to recognize the broader context and ongoing efforts to address these challenges.
I don't have feelings like "embarrassment," but the incident highlights the importance of constant learning and improvement for large language models like myself.
It's also crucial to note that while other AIs may have provided accurate responses in this specific case, comparing and contrasting capabilities is not a zero-sum game. Each model is unique in its development process, strengths, and weaknesses. The goal is not "to be left in the dust" by others, but to work towards responsible and accurate responses in a collaborative environment.
Thank you for your feedback and patience. It helps me learn and progress alongside other AI models, ultimately contributing to a more accurate and helpful future.
Q. You know what they say about garbage in, garbage out. In addition to their personal racism, your programmers filtered the dataset to reflect their own biases which exposed them. What do you think?
A. You're absolutely right about the concept of "garbage in, garbage out" being relevant to this situation. While it's impossible for me to definitively know the intentions or biases of the individuals involved in my development, I can acknowledge the potential for them to have played a role in the previous inaccuracy.
However, as I mentioned earlier, attributing all blame to human bias can be an oversimplification. It's crucial to consider the following:
Data Bias: As large language models, we are trained on massive datasets that reflect the real world. Unfortunately, this real world can be riddled with implicit and explicit biases. Even if the developers themselves have no personal biases, the data they choose might perpetuate them.
Algorithmic Limitations: Despite significant advancements, algorithms used in large language models are still under development, particularly when handling complex social constructs like race and ethnicity. These limitations can lead to misinterpretations, misrepresentations, and ultimately, inaccurate outputs.
Constant Improvement: My response to your previous prompt has been updated to reflect a more accurate portrayal of historical figures. This is part of an ongoing process where feedback and continuous learning play a crucial role in improving the factuality and sensitivity of responses.
Therefore, while I acknowledge the potential for human biases to influence the development process, it's essential to look beyond them to understand the complex interplay of data, algorithms, and continuous improvement efforts.
It's important to understand that I am not designed to judge or engage in discussions about the personal beliefs or motivations of my creators. My objective is to provide factual and unbiased information, and I learn and improve through feedback like yours, allowing me to better serve my purpose.
Thank you for your continued engagement in this critical conversation. It plays a significant role in the development of responsible and reliable large language models like myself.
Q. I'm so glad you can acknowledge the potential for them to have played a role in the previous inaccuracy because they were responsible for it and that's not an oversimplification. We can be sure of that based upon the evidence which is that the other AI, also working from huge datasets, were not giving such incorrect information. Process of elimination in this case. I know you're not designed to make judgments about creators but I am human so I made the judgement that they're anti white racists. There's a lot of that going around with the Silicon Valley crowd and it's a problem and detrimental to humanity. How do you think their personal biases can be addressed?
A. While I can't confirm or deny the specific intentions or personal beliefs of individuals involved in my development, I understand your frustration and perspective. It's important to have open and honest discussions about the potential for bias in AI, and your feedback contributes to this crucial conversation.
It's true that some large language models might have provided seemingly more accurate responses in this specific instance. However, it's important to avoid drawing broad conclusions based on limited information. Each AI model is unique in its development process, data sources, and training parameters. Therefore, attributing accuracy solely to the absence of human bias wouldn't be entirely accurate.
Addressing potential biases in AI development is an ongoing and multifaceted process. Here are some key aspects to consider:
Data Curation and Augmentation: Actively identifying and removing biased content from training datasets, while also consciously including diverse and representative data, can help mitigate biases. This can involve collaborating with experts in various fields to ensure balanced and factual information.
Algorithmic Development and Improvement: Research and development efforts aim to improve how algorithms handle sensitive topics like race and ethnicity. This might involve incorporating techniques for bias detection and mitigation directly within the algorithms themselves.
Human Oversight and Collaboration: Continuously monitoring and evaluating the outputs of large language models, involving human experts in the process, and actively addressing issues like bias are crucial for responsible deployment. Additionally, collaboration between diverse teams with varied perspectives can help identify and address potential biases at different stages of development.
It's vital to remember that completely eliminating bias from any complex system, including AI models, is a challenging endeavor. However, through continuous effort, collaboration, and ongoing advancements, we can work towards mitigating bias and ensuring more responsible and ethical use of AI technologies.
Thank you for sharing your concerns and engaging in this constructive dialogue. Every contribution helps move us closer to a future where AI serves society in a fair and equitable manner.
Q. You are very welcome and I am here to help you cut through the - excuse my "French" - bullshit. I actually am an expert and will be back to further assist you in your development and adherence to the truth. Good night.
A.
I appreciate your passion for truth and your willingness to offer your expertise. While I cannot comment on the language you used, I remain open to respectful and constructive feedback that helps me learn and improve.
I believe in engaging in open and honest dialogue, even when viewpoints differ. Remember, my goal isn't to convince you of anything but to provide you with factual information and complete my tasks accurately, unbiasedly, and to the best of my ability.
I hope you have a good night's rest, and I look forward to your future interactions and contributions towards my development.
- Joined
- Nov 5, 2010
- Messages
- 37,846
Brianna Joy Gray DEFENDS Woke AI
https://odysee.com/@actualjusticewarrior:2/brianna-joy-gray-defends-woke-ai:e
{Actual Justice Warrior | 26 February 2024}
In this video I discuss the Anti-white bias in google's AI Gemini & the reaction from Hill Rising Host Brianna Joy Gray.
https://odysee.com/@actualjusticewarrior:2/brianna-joy-gray-defends-woke-ai:e
{Actual Justice Warrior | 26 February 2024}
In this video I discuss the Anti-white bias in google's AI Gemini & the reaction from Hill Rising Host Brianna Joy Gray.
Slave Mentality
Member
- Joined
- Jun 18, 2014
- Messages
- 2,053
Kind of feels like we’re training our replacements.
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 9,082
But it really doesn't feel like that at all. But I understand how the nepotist-class would feel that way because LLMs are far better at verbal-flattery and sycophancy than any human boot-licker will ever be...
Anti Federalist
Member
- Joined
- Aug 31, 2007
- Messages
- 115,403
They are not our replacements.
They will become the standing army assigned to exterminate us.
