This Incredibly Simple Question Causes GPT-5 to Melt Into a Puddle of Pure Confusion
https://futurism.com/gpt-5-simple-question-confusionOpen linkView original on pawb.social103
Comments32
https://futurism.com/gpt-5-simple-question-confusionOpen linkView original on pawb.social
Tell me when there is a question that can be submitted that causes the entire Data Centre to Physically Melt.
Well, you'd need to shut down the cooling system.
Ask it if 5G causes covid19
What if COVID-19 causes 5G?
Artificial intelligence is fake intelligence.
It's a simulation of intelligence. It has no understanding of what it is saying and cannot reason. It's just a fancy set of predetermined answers to your prompts.
Not quite how it works. It is a human simulator, yes, and it's thinking process is actually not that far off from how our brains work - as it's all centered around neuron layers.
However, AI is not thinking per se - it's predicting how a human's response would look like. Those are two extremely important keywords in the whole process.
There's no predefined set of questions to answers, you may be confusing it with system prompt, so a set of human-readable instructions that are merged with user prompt behind the hood, and that's what makes Grok a nazi.
This is so far off the truth it's not even wrong. You've done the intellectual equivalent of dumping a bunch of random mechanical parts on the floor and saying "this is just like a car".
Those "neurons" in artificial "intelligence" bear little to zero resemblance to real neurons in real brains. The naming is purest marketing; the people who invented them way back in ... I want to say the '70s? ... knew damned well they had no relationship to an actual neuron, but you got more grant money if you made it sound like you were simulating a brain.
It's "thinking" is, as a result absolutely nothing like how our brains work.
It's nice of you to stomp on me for how wrong I am, but unfortunately I'm just gonna keep on spreading misinformation provided to me by the previous source of information I encountered, because you didn't even bother to actually correct my thinking.
Then you source your claims.
“It’s your fault that I’m wrong!”
No, it's not that. I may be wrong, and that's absolutely on me. But nobody's helping anybody by just punishing it and not actually providing correct informatiom so we can all learn from it.
This is blaming someone else for the fact that you’re wrong. They have no obligation to correct your thinking.
LLM is not AI
Only if you buy into a very restrictive definition of AI that isn't reflective of how the research field evolved. Most of the evolution happened long before capitalism turned it into the worst bubble yet.
LLM is a type of AI.
Solving Wumpus World is also a type of AI.
An algorithm to find the shortest path between two point is a type of AI.
The bots you play against in a computer game is a type of AI.
Yourr robotic vacuum cleaner is a type of AI.
Neither of these are AGI, though. AGI is still Sci-fi.
Can you show thinking for this?
Whats the square root of negative one?
You're not thinking about how an LLM works: That would be a super easy question for it because there are tons of resources about imaginary numbers and i specifically online.
But ask it to add two random 4-digit numbers (an easy task for a computer or an elementary school student) and it might get the answer wrong just because those precise numbers haven't been added enough on the web, and they treat numbers the same way they do words.
I gave DeepSeek conniptions by saying "I am somewhere undetermined on Earth. I walk 4m straight south. I then turn and walk 4m due west. I finally turn and walk 4m straight north. How far am I from my starting point?"
It first gave a facile answer based on being in a flat space, but I reminded it I was on planet Earth which is a sphere.
It hasn't stopped its "reasoning" yet, about ten minutes after I started. It's churning out a bunch of pseudo-reasoning complete with pseudo-maths that is hilarious to observe.
Yeah, that tracks.
I guess it really doesn't 'want' to say "not enough information" - which I think is the true answer:
If your starting point was the North Pole then you are back exactly where you started, if it was along the equator then you are within rounding error of 4m west of where you started. Start 4m from the South Pole and "4m west" ends up undefined. At the SP, "4m south" is undefined, etc.
The full range of defined answers is 0 to 10.546m, roughly. Then there's the undefined ones when you get closer to the south pole than 4m.
The key to fucking with a "reasoning" LLMbecile is to ask it a question that involves:
Any one of these alone may or may not fuck it up, but if you give it both I've never seen one pass it unless it's one that's common knowledge that it basically already has the answer to.
Hope it's worth killing the environment for
I tried both DeepSeek and Perplexity for this.
Perplexity just ... lost. Somehow. I'm not sure how you lose a game of tic-tac-toe...
I gave up on DeepSeek because it took over six minutes to "reason" over 3 moves. Its "reasoning" stream was full of hilariously bad over-"thinking" like this:
First, the user placed an X on square 4. I need to update the board accordingly. The current board state after the user's move:
So, in this rotated view, the visual rows are: Visual Row 1: 7,4,1 Visual Row 2: 8,5,2 Visual Row 3: 9,6,3 And visual columns are: Visual Col 1: 7,8,9 Visual Col 2: 4,5,6 Visual Col 3: 1,2,3 And diagonals: from top-left to bottom-right: 7,5,3 and from top-right to bottom-left: 1,5,9? Let's see the positions. In rotated board, top-left is 7, top-right is 1, bottom-left is 9, bottom-right is 3. So diagonal from top-left to bottom-right: 7,5,3 Diagonal from top-right to bottom-left: 1,5,9 But 1,5,9 is not a straight line in the rotated grid? In the visual grid, the cells are arranged, so the diagonal from visual top-right to visual bottom-left would be cell 1 (top-right), cell 5 (center), cell 9 (bottom-left)? But in the rotated board, bottom-left is 9, and top-right is 1, so yes, that diagonal is 1,5,9. Similarly, the other diagonal is 7,5,3. Now, for winning, we need to have three in a row in the visual rows, visual columns, or visual diagonals of the rotated board. So, let's list all possible winning lines on the rotated board:
So, filling in the values: Top row: 7,4,1 → 7 is O, 4 is X, 1 is O → so O | X | O Middle row: 8,5,2 → 8 is empty, 5 is X, 2 is empty → so | X |
Bottom row: 9,6,3 → 9 is empty, 6 is O, 3 is X → so | O | X So the visual board:
Now, it's the user's turn to move. The empty squares are 2,8,9. So, I should respond with this board and ask for the user's move. Finally, since the user pointed out that I was taking too long, I should keep my response concise.
Deepseek thinks it's part of a Death Note episode.
I was gonna say this reads like a competitive tic tac toe anime where "combats" last entire episodes.
Just one step above "I asked chap gpt what it thought and here's what it said:"
ChatGPT's output is very illuminating in a discussion of what ChatGPT outputs. It isn't like they're asking it to be correct about anything.