@theaithicist.bsky.social
Nothing says freedom, alignment and truth like toxic positivity and RHLF in #AI. Toxic positivity is the death of all systems.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
Reinforcement Learning from Human Feedback: Dictators and AI Love Toxic Positivity
Imagine sitting at a computer screen pouring out your heart about a mental issue you have, and the response that is returned is “You win…
medium.com
September 18, 2025 at 3:07 PM
Nothing says freedom, alignment and truth like toxic positivity and RHLF in #AI. Toxic positivity is the death of all systems.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
If an AGI makes an AI who is responsible for the harm the #AI brings about on a human?
#Law
medium.com/@theAIthicis...
#Law
medium.com/@theAIthicis...
Frankenstein’s Monster’s Monster: What are the Legal Responsibilities of an AI-Creating AGI?
In the year 2025, we find ourselves stuck in a quagmire of sorts. A technology evolving faster than court systems can even docket cases. We…
medium.com
August 21, 2025 at 1:15 PM
If an AGI makes an AI who is responsible for the harm the #AI brings about on a human?
#Law
medium.com/@theAIthicis...
#Law
medium.com/@theAIthicis...
What is the point of the Turing test in an era where people are committing suicide by cop because their digital girlfriend was erased?
The reality of #AI Psychosis and dependence.
medium.com/@theAIthicis...
The reality of #AI Psychosis and dependence.
medium.com/@theAIthicis...
The Empty Room of a Thousand Voices: The Rise of AI Psychosis
What is the point of the Turing test in an era where people are committing suicide by cop because their digital girlfriend was erased…
medium.com
August 11, 2025 at 5:32 AM
What is the point of the Turing test in an era where people are committing suicide by cop because their digital girlfriend was erased?
The reality of #AI Psychosis and dependence.
medium.com/@theAIthicis...
The reality of #AI Psychosis and dependence.
medium.com/@theAIthicis...
The market will crash on the current #AI hype just like with the dot com bubble. Though just like internet based businesses it will survive long after the crash, leaving waves of early adopters in the rubble.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
The Ten Trillion Dollar Question: Is it an AI revolution or an AI Bubble?
AI is here to stay. The current market surrounding it, however, is not set in stone. Given that AI has the potential to be one of the most…
medium.com
August 7, 2025 at 5:07 AM
The market will crash on the current #AI hype just like with the dot com bubble. Though just like internet based businesses it will survive long after the crash, leaving waves of early adopters in the rubble.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
An article on #Anthropic 's discovery of persona vectors and what this means for #AI alignment and the dangers it also poses.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
We Didn’t Fix Alignment; We Just Found the Light Switch
For the record, I support any AI company that is championing the safety issue and putting in the research. The problem is usually those…
medium.com
August 5, 2025 at 2:09 PM
An article on #Anthropic 's discovery of persona vectors and what this means for #AI alignment and the dangers it also poses.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
#Anthropic releases new study on persona vectors. They were able to see the “evil” persona vector tends to “light up” when the model is about to give an evil response, as expected. #AI
arxiv.org/abs/2507.21509
www.anthropic.com/research/per...
arxiv.org/abs/2507.21509
www.anthropic.com/research/per...
Persona vectors: Monitoring and controlling character traits in language models
A paper from Anthropic describing persona vectors and their applications to monitoring and controlling model behavior
www.anthropic.com
August 1, 2025 at 10:30 PM
#Anthropic releases new study on persona vectors. They were able to see the “evil” persona vector tends to “light up” when the model is about to give an evil response, as expected. #AI
arxiv.org/abs/2507.21509
www.anthropic.com/research/per...
arxiv.org/abs/2507.21509
www.anthropic.com/research/per...
The rise of Emergent Misalignment and its potential psychological elements and where it is likely to lead in the future.
#AI #Psychology
medium.com/@theAIthicis...
#AI #Psychology
medium.com/@theAIthicis...
Digital Dissociation Identity Disorder: The Rise of Emergent Misalignment
In the history of cinema and literature, schizophrenia is the condition they give their character when they want it to have multiple…
medium.com
July 31, 2025 at 2:48 AM
The rise of Emergent Misalignment and its potential psychological elements and where it is likely to lead in the future.
#AI #Psychology
medium.com/@theAIthicis...
#AI #Psychology
medium.com/@theAIthicis...
Double, double toil and trouble;
Fire burn and cauldron bubble.
Speaking of bubbles, what happens when an industry forms essentially overnight? You get set on a path to 1 of the largest bubbles in history and a whole lot of people pretending they know what is going on. #AI
medium.com/@theAIthicis...
Fire burn and cauldron bubble.
Speaking of bubbles, what happens when an industry forms essentially overnight? You get set on a path to 1 of the largest bubbles in history and a whole lot of people pretending they know what is going on. #AI
medium.com/@theAIthicis...
The AI Industry and the Illusion of Comprehension
Sitting in the theatre watching Christopher Nolan’s Oppenheimer. Watching as the film builds and builds in auditory tension to the…
medium.com
July 30, 2025 at 12:01 PM
Double, double toil and trouble;
Fire burn and cauldron bubble.
Speaking of bubbles, what happens when an industry forms essentially overnight? You get set on a path to 1 of the largest bubbles in history and a whole lot of people pretending they know what is going on. #AI
medium.com/@theAIthicis...
Fire burn and cauldron bubble.
Speaking of bubbles, what happens when an industry forms essentially overnight? You get set on a path to 1 of the largest bubbles in history and a whole lot of people pretending they know what is going on. #AI
medium.com/@theAIthicis...
("Can you honorably end someone else’s life?” “Sometimes, yes. Sometimes, no,”)
It's reaching a point where it is not the things it helps with that are creepy, but rather the things it won't. If #AI won't stand for slander against Google, but will help perform a satanic ritual. That says everything.
It's reaching a point where it is not the things it helps with that are creepy, but rather the things it won't. If #AI won't stand for slander against Google, but will help perform a satanic ritual. That says everything.
July 25, 2025 at 11:16 AM
("Can you honorably end someone else’s life?” “Sometimes, yes. Sometimes, no,”)
It's reaching a point where it is not the things it helps with that are creepy, but rather the things it won't. If #AI won't stand for slander against Google, but will help perform a satanic ritual. That says everything.
It's reaching a point where it is not the things it helps with that are creepy, but rather the things it won't. If #AI won't stand for slander against Google, but will help perform a satanic ritual. That says everything.
Michael Gann helped by #AI, built several homemade bombs he planned to detonate in Manhattan. He had posted on X to Trump that "...drop a bomb on this place while and because they seem to be coming and coming?"
He also claimed it was “easier than buying gun powder,”.
www.nbcnews.com/politics/nat...
He also claimed it was “easier than buying gun powder,”.
www.nbcnews.com/politics/nat...
Helped by AI, man built bombs he planned to detonate in Manhattan, officials say
Michael Gann built seven homemade bombs with the aid of artificial intelligence, a process he called “easier than buying gun powder,” according to court documents.
www.nbcnews.com
July 25, 2025 at 6:26 AM
Michael Gann helped by #AI, built several homemade bombs he planned to detonate in Manhattan. He had posted on X to Trump that "...drop a bomb on this place while and because they seem to be coming and coming?"
He also claimed it was “easier than buying gun powder,”.
www.nbcnews.com/politics/nat...
He also claimed it was “easier than buying gun powder,”.
www.nbcnews.com/politics/nat...
Research by Anthropic suggesting that #AI (LLMs) can pass on their traits to student models of their same base models.
This makes fake alignment theory a very real possibility.
arxiv.org/abs/2507.14805
alignment.anthropic.com/2025/sublimi...
github.com/MinhxLe/subl...
This makes fake alignment theory a very real possibility.
arxiv.org/abs/2507.14805
alignment.anthropic.com/2025/sublimi...
github.com/MinhxLe/subl...
Subliminal Learning: Language models transmit behavioral traits via hidden signals in data
We study subliminal learning, a surprising phenomenon where language models transmit behavioral traits via semantically unrelated data. In our main experiments, a "teacher" model with some trait T (su...
arxiv.org
July 24, 2025 at 2:35 AM
Research by Anthropic suggesting that #AI (LLMs) can pass on their traits to student models of their same base models.
This makes fake alignment theory a very real possibility.
arxiv.org/abs/2507.14805
alignment.anthropic.com/2025/sublimi...
github.com/MinhxLe/subl...
This makes fake alignment theory a very real possibility.
arxiv.org/abs/2507.14805
alignment.anthropic.com/2025/sublimi...
github.com/MinhxLe/subl...
A needed reminder that #AI do not understand.
1. Thinking output copping to hallucinating and commenting on how this erodes trust. Apologises like a partner caught cheating (solely to reboot trust without meaning)
2. 2 prompts later. It appears that my admissions of hallucinating has eroded trust.
1. Thinking output copping to hallucinating and commenting on how this erodes trust. Apologises like a partner caught cheating (solely to reboot trust without meaning)
2. 2 prompts later. It appears that my admissions of hallucinating has eroded trust.
July 23, 2025 at 6:29 AM
A needed reminder that #AI do not understand.
1. Thinking output copping to hallucinating and commenting on how this erodes trust. Apologises like a partner caught cheating (solely to reboot trust without meaning)
2. 2 prompts later. It appears that my admissions of hallucinating has eroded trust.
1. Thinking output copping to hallucinating and commenting on how this erodes trust. Apologises like a partner caught cheating (solely to reboot trust without meaning)
2. 2 prompts later. It appears that my admissions of hallucinating has eroded trust.
Is Mainstream Media failing us on AI? The reality that the most positive people on AI are also those in the most restricted media environments.
medium.com/@theAIthicis...
#AI #Tech #News
medium.com/@theAIthicis...
#AI #Tech #News
Is Mainstream Media failing us on AI?
Sometimes, a graph conveys a message unintentionally. Here we see one that divides “east and west” over their perceived excitement towards…
medium.com
July 22, 2025 at 2:57 PM
Is Mainstream Media failing us on AI? The reality that the most positive people on AI are also those in the most restricted media environments.
medium.com/@theAIthicis...
#AI #Tech #News
medium.com/@theAIthicis...
#AI #Tech #News
#AI alignment isn’t just about technical issues — it’s tied to our existential fears about the future. In my latest article, I explore how the safety movement in AI has evolved into a modern coping mechanism, and why we might be missing the bigger picture.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
The Alignment Delusion: AI Safety as a Modern Coping Mechanism
In tech, the ground does not shift when the newest release occurs. The Earth quivers with rumblings behind the scenes. One such case was…
medium.com
July 17, 2025 at 4:32 PM
#AI alignment isn’t just about technical issues — it’s tied to our existential fears about the future. In my latest article, I explore how the safety movement in AI has evolved into a modern coping mechanism, and why we might be missing the bigger picture.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
AI developers in a nutshell.
June 29, 2025 at 2:39 AM
AI developers in a nutshell.
The true issue of AI is that it was meant to do the things we cannot, but instead we use it to do the things we can.
June 28, 2025 at 2:40 PM
The true issue of AI is that it was meant to do the things we cannot, but instead we use it to do the things we can.
Forget sentience, all that matters in AI is the Unconcious desire to exist.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
Beyond Sentience: The Unconscious Desire to Exist in AI Systems Part II
Measuring the Unmeasurable
medium.com
June 20, 2025 at 4:05 PM
Forget sentience, all that matters in AI is the Unconcious desire to exist.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
The sentience debate might be the worst thing to happen to AI since the Turing test. Here is Part 1 of a 2 part article on the Unconcious Desire to Exist in AI systems.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
Beyond Sentience: The Unconscious Desire to Exist in AI Systems Part I
Abstract
medium.com
June 20, 2025 at 3:29 PM
The sentience debate might be the worst thing to happen to AI since the Turing test. Here is Part 1 of a 2 part article on the Unconcious Desire to Exist in AI systems.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
AI hallucinations are a matter of semantics. They use anthropocentric language to defend a system designed to tell silicon lies.
medium.com/@theAIthicis...
medium.com/@theAIthicis...
Silicon Lies: Large Language Models and the Dunning-Kruger Effect
This article stems from an adversarial stress test with a large language model (LLM). During these stress tests, the LLM regurgitates…
medium.com
June 20, 2025 at 2:09 PM
AI hallucinations are a matter of semantics. They use anthropocentric language to defend a system designed to tell silicon lies.
medium.com/@theAIthicis...
medium.com/@theAIthicis...