House Inputs and Outputs xkcd.com/2963
July 25, 2024 at 3:12 AM
House Inputs and Outputs xkcd.com/2963
A lot of people say AI isn't very good for the world and its outputs are routinely unreliable, but these haters fail to see how it's revolutionizing how we incinerate schoolchildren.
To strike 1,000 targets in 24 hours in Iran, the U.S. military leveraged the most advanced AI it’s ever used in warfare.
Anthropic’s Claude partnered with the military’s Maven Smart System, suggesting targets and issuing precise location coordinates. https://wapo.st/46BFe1T
Anthropic’s Claude partnered with the military’s Maven Smart System, suggesting targets and issuing precise location coordinates. https://wapo.st/46BFe1T
March 5, 2026 at 4:43 AM
A lot of people say AI isn't very good for the world and its outputs are routinely unreliable, but these haters fail to see how it's revolutionizing how we incinerate schoolchildren.
"Efficiency" has a specific meaning, which is the ratio of inputs to outputs. You can't calculate efficiency without measuring outputs. I can save money by not putting gas in my car, but I'm not going anywhere.
DOGE is not just cutting inputs, it's cutting outputs, which is a Congressional issue.
DOGE is not just cutting inputs, it's cutting outputs, which is a Congressional issue.
DOGE is not a govt efficiency project. Firing 1 in 4 fed employees would cut govt spending by 1%. DOGE is a political control project. Firing and terrorizing public employees is a means to weakening state regulation & strengthening a personalist presidency donmoynihan.substack.com/p/doge-misma...
DOGE Mismanagement Principles
Musk et al. don't know or care about the damage they are doing
donmoynihan.substack.com
February 17, 2025 at 2:45 PM
"Efficiency" has a specific meaning, which is the ratio of inputs to outputs. You can't calculate efficiency without measuring outputs. I can save money by not putting gas in my car, but I'm not going anywhere.
DOGE is not just cutting inputs, it's cutting outputs, which is a Congressional issue.
DOGE is not just cutting inputs, it's cutting outputs, which is a Congressional issue.
genAI has empowered mediocre people to create outputs that nobody wants, for the sake of creating outputs
September 23, 2026 at 5:41 AM
genAI has empowered mediocre people to create outputs that nobody wants, for the sake of creating outputs
Just calling them "outputs" tells you all you need to know.
September 22, 2026 at 3:41 PM
Just calling them "outputs" tells you all you need to know.
"In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limits."
www.computerworld.com/article/4059...
www.computerworld.com/article/4059...
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws
In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limi...
www.computerworld.com
September 21, 2025 at 12:50 PM
"In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limits."
www.computerworld.com/article/4059...
www.computerworld.com/article/4059...
Functionalist approaches measure intelligence by how closely its outputs resemble the human mind. If you constantly recalibrate its outputs to resemble a human mind, you're not creating a thinking machine. You are creating a machine to trick other functionalists.
To me the chatbots pose zero interesting questions about the ‘nature of selfhood’ but dozens about the nature of manipulation. Here the Claude team pretends they are programming the robot for ethical responses but they land on one that is false and disingenuous.
February 11, 2026 at 7:13 PM
Functionalist approaches measure intelligence by how closely its outputs resemble the human mind. If you constantly recalibrate its outputs to resemble a human mind, you're not creating a thinking machine. You are creating a machine to trick other functionalists.
hazing and gang initiation are very similar activities that produce similar outputs
this contrarian pick-me bullshit needs to be addressed.
I was hazed by what most would deem a “pretend” frat: no house, Ivy League campus, nerdy kids. They didn’t hit me, which I suppose is good. But that restraint allowed them to give themselves permission to do anything else up to that line.
I was hazed by what most would deem a “pretend” frat: no house, Ivy League campus, nerdy kids. They didn’t hit me, which I suppose is good. But that restraint allowed them to give themselves permission to do anything else up to that line.
March 7, 2026 at 3:23 PM
hazing and gang initiation are very similar activities that produce similar outputs
A reliable metric for a losing war: notice how often a combatant describes inputs (ordinance dropped, resources committed, duration of operation) as a substitute for outputs (effects a military operation is intending to produce).
So this presser is just Trump listing off stats on assets we're using in Iran. The diminishing returns of these highly billed events is really showing through lately
April 6, 2026 at 5:29 PM
A reliable metric for a losing war: notice how often a combatant describes inputs (ordinance dropped, resources committed, duration of operation) as a substitute for outputs (effects a military operation is intending to produce).
work has banned chatgpt because it's bad and leaky and predatory and gives poor quality outputs. fuckin GOOD.
September 11, 2025 at 7:51 AM
work has banned chatgpt because it's bad and leaky and predatory and gives poor quality outputs. fuckin GOOD.
Outputs are not thoughts. Transcripts of sequential outputs are not actually “chains of thought.”
“Chain of thought” as it’s used in AI is an anthropomorphized and inaccurate descriptor.
“Chain of thought” as it’s used in AI is an anthropomorphized and inaccurate descriptor.
Yes, AIs think. You can read their thoughts as they have them. AI companies hide the AI’s thoughts because other AI companies might read them and use those thoughts to train their own AIs. This is all very strange and astonishing science fiction territory but it’s real.
Oh, they "think," do they?
September 15, 2026 at 11:48 PM
Outputs are not thoughts. Transcripts of sequential outputs are not actually “chains of thought.”
“Chain of thought” as it’s used in AI is an anthropomorphized and inaccurate descriptor.
“Chain of thought” as it’s used in AI is an anthropomorphized and inaccurate descriptor.
But who am to laugh at electricity outputs
January 9, 2026 at 9:29 PM
But who am to laugh at electricity outputs
When you realize how much of generating outputs with AI - that aren’t immediately obvious slop - requires professional teams worth of experienced human effort editing said outputs to look remotely passable, it becomes obvious how much of a scam AI is.
to be fair, I think most reactions weren't necessarily "i hate this because it's AI" but "this looks stupid as hell, just show the movie." But also, so much of the ideological push for AI is in part about hiding the real labor and resources it requires.
So apparently, it may turn out that The Wizard of Oz at the Las Vegas Sphere wasn’t entirely made in AI and Google threw the VFX crew under the bus by claiming it was all done by artificial intelligence.
If true, then fucking yikes.
If true, then fucking yikes.
August 5, 2025 at 5:32 PM
When you realize how much of generating outputs with AI - that aren’t immediately obvious slop - requires professional teams worth of experienced human effort editing said outputs to look remotely passable, it becomes obvious how much of a scam AI is.
It has dual headphone outputs because this is an MP3 player for people who fuck
May 25, 2025 at 6:50 AM
It has dual headphone outputs because this is an MP3 player for people who fuck
Tim's failure was not 'failing to check the outputs', it was using a machine that fakes plausible outputs instead of referring to reality
It is mid 2026. You cannot plead ignorance about what that software does, anymore
It is as bad as sitting down and writing a fake quote with your own hand
It is mid 2026. You cannot plead ignorance about what that software does, anymore
It is as bad as sitting down and writing a fake quote with your own hand
June 3, 2026 at 9:59 PM
Tim's failure was not 'failing to check the outputs', it was using a machine that fakes plausible outputs instead of referring to reality
It is mid 2026. You cannot plead ignorance about what that software does, anymore
It is as bad as sitting down and writing a fake quote with your own hand
It is mid 2026. You cannot plead ignorance about what that software does, anymore
It is as bad as sitting down and writing a fake quote with your own hand
Still better than 99.9% of AI outputs since
then in 2019 GPT-2 generated this carol
www.aiweirdness.com/the-ais-caro...
www.aiweirdness.com/the-ais-caro...
December 17, 2025 at 8:23 PM
Still better than 99.9% of AI outputs since
shoutout to the portal 2 speedrun plugin that outputs an ascii pride flag in the console if you launch the game in june
June 2, 2025 at 1:39 AM
shoutout to the portal 2 speedrun plugin that outputs an ascii pride flag in the console if you launch the game in june
86. Anyone can produce outputs that look like design outputs (including AI). But outputs are just a receipt for a whole bunch of work, and faking a receipt doesn't substitute for the work.
October 25, 2024 at 3:31 AM
86. Anyone can produce outputs that look like design outputs (including AI). But outputs are just a receipt for a whole bunch of work, and faking a receipt doesn't substitute for the work.
"Refusing to use AI outputs is sabotage." Amazing.
September 2, 2025 at 1:39 PM
"Refusing to use AI outputs is sabotage." Amazing.
i will never stop being furious at large language models being successfully branded as "artificial intelligence"
they're not conscious & functionally can't be; they don't "hallucinate," they fail to generate appropriate outputs
they're not conscious & functionally can't be; they don't "hallucinate," they fail to generate appropriate outputs
July 10, 2024 at 2:59 PM
i will never stop being furious at large language models being successfully branded as "artificial intelligence"
they're not conscious & functionally can't be; they don't "hallucinate," they fail to generate appropriate outputs
they're not conscious & functionally can't be; they don't "hallucinate," they fail to generate appropriate outputs
"AI is just like calculators"
Show me a calculator sold with the warning "outputs may be incorrect. Please check answers before using"
Show me a calculator sold with the warning "outputs may be incorrect. Please check answers before using"
November 29, 2025 at 2:53 PM
"AI is just like calculators"
Show me a calculator sold with the warning "outputs may be incorrect. Please check answers before using"
Show me a calculator sold with the warning "outputs may be incorrect. Please check answers before using"
Harder to think of a clearer example of "The Master's Tools Will Never Dismantle The Master's House" for this era than people using "generative AI" to protest the outputs and infrastructure of "AI"
June 6, 2026 at 9:35 PM
Harder to think of a clearer example of "The Master's Tools Will Never Dismantle The Master's House" for this era than people using "generative AI" to protest the outputs and infrastructure of "AI"
A chatbot’s outputs aren’t “about” things “out in the world.” Its outputs are “about” nothing except its own training data. Every output should be understood as “this is what my training data tells me is the likeliest string to be produced following your prompt”
June 3, 2024 at 12:45 PM
A chatbot’s outputs aren’t “about” things “out in the world.” Its outputs are “about” nothing except its own training data. Every output should be understood as “this is what my training data tells me is the likeliest string to be produced following your prompt”