extremely important point that it is also a very necessary "critical thinking, reading comprehension, and logical reasoning" test
which, again, they are failing
which, again, they are failing
All this and the some.
Just to add, that as well as an ethics test, it’s also also a basic political competency test. Are you prepared, if someone fires an utterly predictable question with a gotcha framing at you, do you have the chops not to be trapped by the framing but to reframe it yourself
Just to add, that as well as an ethics test, it’s also also a basic political competency test. Are you prepared, if someone fires an utterly predictable question with a gotcha framing at you, do you have the chops not to be trapped by the framing but to reframe it yourself
September 11, 2026 at 6:12 AM
extremely important point that it is also a very necessary "critical thinking, reading comprehension, and logical reasoning" test
which, again, they are failing
which, again, they are failing
Lmao turns out that when you build a system in which instructions, inputs, outputs, and "reasoning" are indistinguishable it's piss easy to trick it into thinking your input is instructions and/or reasoning
Incredible scenes from Meta. Apparently you can prompt inject their Muse agent into giving you a root shell on the box it's running on by roleplaying that you are actually the Muse agent itself.
jonny (nonvenomous) (@jonny@neuromatch.social)
Attached: 1 image
Not only confirming this, its extremely easy. Almost no prompt injection resistance lmao
neuromatch.social
September 25, 2026 at 8:23 PM
Lmao turns out that when you build a system in which instructions, inputs, outputs, and "reasoning" are indistinguishable it's piss easy to trick it into thinking your input is instructions and/or reasoning
Hi, just because a software tells you it’s “reasoning” or “thinking” doesn’t mean it actually is.
You won’t believe this but programmers can make software lie about what it does.
Maybe programmers need more guardrails.
You won’t believe this but programmers can make software lie about what it does.
Maybe programmers need more guardrails.
June 6, 2026 at 2:11 AM
Hi, just because a software tells you it’s “reasoning” or “thinking” doesn’t mean it actually is.
You won’t believe this but programmers can make software lie about what it does.
Maybe programmers need more guardrails.
You won’t believe this but programmers can make software lie about what it does.
Maybe programmers need more guardrails.
what the FUCK am i reading
September 25, 2026 at 7:35 AM
what the FUCK am i reading
... or if not consciousness, then some thinking, reasoning, self-improving entity.
>>
>>
September 11, 2026 at 3:06 AM
... or if not consciousness, then some thinking, reasoning, self-improving entity.
>>
>>
This is the triumph of “tremendous writing“ over good reasoning, sensible thinking, or common sense. The death of the universe by hot takes
October 12, 2025 at 10:16 PM
This is the triumph of “tremendous writing“ over good reasoning, sensible thinking, or common sense. The death of the universe by hot takes
None of these models possess any thinking or reasoning capacity
August 8, 2025 at 6:29 AM
None of these models possess any thinking or reasoning capacity
So here's the thing I keep coming back to.
There are plenty of other things that have reasoning & language.
Hell, just in one letter of the alphabet - Corvids, Canids, Cetaceans, Cephalopods. We're surrounded by thinking, reasoning, feeling beings.
Why would that alter my understanding of self?
There are plenty of other things that have reasoning & language.
Hell, just in one letter of the alphabet - Corvids, Canids, Cetaceans, Cephalopods. We're surrounded by thinking, reasoning, feeling beings.
Why would that alter my understanding of self?
September 8, 2026 at 6:12 PM
So here's the thing I keep coming back to.
There are plenty of other things that have reasoning & language.
Hell, just in one letter of the alphabet - Corvids, Canids, Cetaceans, Cephalopods. We're surrounded by thinking, reasoning, feeling beings.
Why would that alter my understanding of self?
There are plenty of other things that have reasoning & language.
Hell, just in one letter of the alphabet - Corvids, Canids, Cetaceans, Cephalopods. We're surrounded by thinking, reasoning, feeling beings.
Why would that alter my understanding of self?
I bristle at the use of “thinking” applied to AI models.
Thinking is more than computation or even the reasoning inherent to problem-solving.
Thinking involves interaction between the interiority of the mind and the sensorial realities of the external world.
It involves curiosity and wonder.
Thinking is more than computation or even the reasoning inherent to problem-solving.
Thinking involves interaction between the interiority of the mind and the sensorial realities of the external world.
It involves curiosity and wonder.
September 20, 2026 at 5:34 PM
I bristle at the use of “thinking” applied to AI models.
Thinking is more than computation or even the reasoning inherent to problem-solving.
Thinking involves interaction between the interiority of the mind and the sensorial realities of the external world.
It involves curiosity and wonder.
Thinking is more than computation or even the reasoning inherent to problem-solving.
Thinking involves interaction between the interiority of the mind and the sensorial realities of the external world.
It involves curiosity and wonder.
her big lesson: being impressed she was manipulated by her own reasoning versus jfk what was I thinking
August 25, 2026 at 2:10 AM
her big lesson: being impressed she was manipulated by her own reasoning versus jfk what was I thinking
Globally, we are getting dumber, and it's by design. Capitalists want easy to control drones who lack critical thinking skills.
A study reveals that humanity is losing focus and logical thinking, with a decline in concentration and reasoning since 2010.
A study reveals that humanity is losing focus and logical thinking, with a decline in concentration and reasoning since 2010.
March 19, 2025 at 9:05 PM
Globally, we are getting dumber, and it's by design. Capitalists want easy to control drones who lack critical thinking skills.
A study reveals that humanity is losing focus and logical thinking, with a decline in concentration and reasoning since 2010.
A study reveals that humanity is losing focus and logical thinking, with a decline in concentration and reasoning since 2010.
I beg people to engage in more both/and thinking. So much conflictual discourse is driven by faulty either/or reasoning, and it’s tiresome.
April 21, 2025 at 1:42 AM
I beg people to engage in more both/and thinking. So much conflictual discourse is driven by faulty either/or reasoning, and it’s tiresome.
Responding to thoughtful comments from others, and thinking about it more, here is why I think Chutkan's reasoning doesn't hold up:
It can't be that when the Trump admin admits they did something wrong (or made a mistake), a court has jurisdiction, but when they don't admit it, then a court can't issue the same sort of order if the court finds that they did something wrong.
September 16, 2025 at 5:15 AM
Responding to thoughtful comments from others, and thinking about it more, here is why I think Chutkan's reasoning doesn't hold up:
me: i am a rational person and i evaluate the world around me with clear eyed skepticism of superstition, magical thinking, and faith based reasoning
also me: there are actual demons that live in SharePoint. the demons in SharePoint delight in our suffering
also me: there are actual demons that live in SharePoint. the demons in SharePoint delight in our suffering
May 6, 2025 at 8:20 PM
me: i am a rational person and i evaluate the world around me with clear eyed skepticism of superstition, magical thinking, and faith based reasoning
also me: there are actual demons that live in SharePoint. the demons in SharePoint delight in our suffering
also me: there are actual demons that live in SharePoint. the demons in SharePoint delight in our suffering
How can people continue to fundamentally NOT UNDERSTAND that genAI is not thinking or reasoning?
June 23, 2026 at 9:09 PM
How can people continue to fundamentally NOT UNDERSTAND that genAI is not thinking or reasoning?
I'm going to go to the outrageous length of assuming that if someone feels or acts one way or another, they might have done some reasoning, thinking, or living to get there
July 3, 2026 at 5:08 PM
I'm going to go to the outrageous length of assuming that if someone feels or acts one way or another, they might have done some reasoning, thinking, or living to get there
Defending MAHA and its pseudoscience is an exercise in motivated reasoning and mental gymnastics, not scientific thinking.
February 2, 2026 at 8:35 PM
Defending MAHA and its pseudoscience is an exercise in motivated reasoning and mental gymnastics, not scientific thinking.
Not to prop up doomer ignorant mansplainer Bill, but I used to joke with my former manager Samy Bengio on whether he doesn't get tired of showing that "reasoning claims are bullshit" (my words anyway).
machinelearning.apple.com/research/ill...
machinelearning.apple.com/research/ill...
September 7, 2026 at 3:31 AM
Not to prop up doomer ignorant mansplainer Bill, but I used to joke with my former manager Samy Bengio on whether he doesn't get tired of showing that "reasoning claims are bullshit" (my words anyway).
machinelearning.apple.com/research/ill...
machinelearning.apple.com/research/ill...
let models reason illegibly!! this rules! you may not like it but this is what peak reasoning looks like
June 9, 2026 at 7:08 PM
let models reason illegibly!! this rules! you may not like it but this is what peak reasoning looks like
Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought
arxiv.org/abs/2604.22709
arxiv.org/abs/2604.22709
Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought
While long, explicit chains-of-thought (CoT) have proven effective on complex reasoning tasks, they are costly to generate during inference. Non-verbal reasoning methods have emerged with shorter gene...
arxiv.org
April 28, 2026 at 3:08 PM
Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought
arxiv.org/abs/2604.22709
arxiv.org/abs/2604.22709
Religion harms healthy reasoning!
Religion demands:
🔸 Faith over evidence
🔸 Loyalty over doubt
🔸 Reverence over critique
🔸 Obedience over autonomy
Healthy reasoning thrives on
🔸 Evidence
🔸 Critical thinking
🔸 Willingness to question assumptions
🔸 Openness to being wrong
Seek TRUTH!
Religion demands:
🔸 Faith over evidence
🔸 Loyalty over doubt
🔸 Reverence over critique
🔸 Obedience over autonomy
Healthy reasoning thrives on
🔸 Evidence
🔸 Critical thinking
🔸 Willingness to question assumptions
🔸 Openness to being wrong
Seek TRUTH!
July 11, 2025 at 4:16 PM
Religion harms healthy reasoning!
Religion demands:
🔸 Faith over evidence
🔸 Loyalty over doubt
🔸 Reverence over critique
🔸 Obedience over autonomy
Healthy reasoning thrives on
🔸 Evidence
🔸 Critical thinking
🔸 Willingness to question assumptions
🔸 Openness to being wrong
Seek TRUTH!
Religion demands:
🔸 Faith over evidence
🔸 Loyalty over doubt
🔸 Reverence over critique
🔸 Obedience over autonomy
Healthy reasoning thrives on
🔸 Evidence
🔸 Critical thinking
🔸 Willingness to question assumptions
🔸 Openness to being wrong
Seek TRUTH!
There's a lot of big news today/this week, but Apple just dropped a nuke of a paper about LLMs & LRMs, specially around "high-complexity tasks where both models experience complete collapse." This is the biggest sign yet that if AI ever lives up the hype, it won't be via those approaches:
The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
Recent generations of frontier language models have introduced Large Reasoning Models
(LRMs) that generate detailed thinking processes…
machinelearning.apple.com
June 7, 2025 at 10:53 PM
There's a lot of big news today/this week, but Apple just dropped a nuke of a paper about LLMs & LRMs, specially around "high-complexity tasks where both models experience complete collapse." This is the biggest sign yet that if AI ever lives up the hype, it won't be via those approaches:
self-aware
emergent
complexity
consciousness
intelligence
thinking
meaning
understanding
qualia
sentience
sapience
ontology
mind
reasoning
creativity
emergent
complexity
consciousness
intelligence
thinking
meaning
understanding
qualia
sentience
sapience
ontology
mind
reasoning
creativity
August 29, 2025 at 8:20 PM
self-aware
emergent
complexity
consciousness
intelligence
thinking
meaning
understanding
qualia
sentience
sapience
ontology
mind
reasoning
creativity
emergent
complexity
consciousness
intelligence
thinking
meaning
understanding
qualia
sentience
sapience
ontology
mind
reasoning
creativity
“…the researchers argue that AI systems have given rise to a categorically different form of “cognitive surrender” in which users provide “minimal internal engagement” and accept an AI’s reasoning wholesale without oversight or verification.”
"Cognitive surrender" leads AI users to abandon logical thinking, research finds
Experiments show large majorities uncritically accepting "faulty" AI answers.
arstechnica.com
April 3, 2026 at 11:28 PM
“…the researchers argue that AI systems have given rise to a categorically different form of “cognitive surrender” in which users provide “minimal internal engagement” and accept an AI’s reasoning wholesale without oversight or verification.”