@urbanowicz.bsky.social
By the way, this problem didn't exist in CRTs, because they only flash a line for a moment and then the line goes black, so there's no real image to smear.
November 25, 2025 at 8:50 AM
By the way, this problem didn't exist in CRTs, because they only flash a line for a moment and then the line goes black, so there's no real image to smear.
The biggest offender is mismatch between smooth eye movement and the image which moves in jumps. Real moving object appears sharp when you track it, but a 60 Hz animated moving target could be seen without smear only if you somehow perform 60 saccades per second.
November 25, 2025 at 8:46 AM
The biggest offender is mismatch between smooth eye movement and the image which moves in jumps. Real moving object appears sharp when you track it, but a 60 Hz animated moving target could be seen without smear only if you somehow perform 60 saccades per second.
The world if we knew how to take a pixel and make more of that exact pixel:
November 24, 2025 at 9:02 AM
The world if we knew how to take a pixel and make more of that exact pixel:
thrilled, next-gen
November 18, 2025 at 11:16 AM
thrilled, next-gen
"Voluntarily submitted the form at this date" would be an additional datapoint, which wasn't there.
October 23, 2025 at 7:35 PM
"Voluntarily submitted the form at this date" would be an additional datapoint, which wasn't there.
Is it cheaper than electric lights?
October 22, 2025 at 3:51 PM
Is it cheaper than electric lights?
Fifteen? Merlin's beard, Tom! Isn't it bad enough to consider a second take? To rip the soul into fifteen pieces...
October 11, 2025 at 5:41 AM
Fifteen? Merlin's beard, Tom! Isn't it bad enough to consider a second take? To rip the soul into fifteen pieces...
Nor do we ask taxi drivers to pass an exam before taking a ride, even though we don't see their "previous work".
October 11, 2025 at 5:22 AM
Nor do we ask taxi drivers to pass an exam before taking a ride, even though we don't see their "previous work".
It was a joke. Those people really believe in "just a tool". They even literally use this knife example, as if there's no difference between steel weapons and cutlery.
October 6, 2025 at 8:30 PM
It was a joke. Those people really believe in "just a tool". They even literally use this knife example, as if there's no difference between steel weapons and cutlery.
... you see, this is just a tool. It is neutral by itself, like knives. A knife can be used for bad (like murder) or for good (like cooking). That’s why every kitchen must keep at least one sword.
October 6, 2025 at 7:56 PM
... you see, this is just a tool. It is neutral by itself, like knives. A knife can be used for bad (like murder) or for good (like cooking). That’s why every kitchen must keep at least one sword.
We know how to create (craft, if you will) perfect copies of ML models. We don't know how to copy persons. We aren't even sure it is possible without destroying the original person.
October 4, 2025 at 7:18 AM
We know how to create (craft, if you will) perfect copies of ML models. We don't know how to copy persons. We aren't even sure it is possible without destroying the original person.
There’s probably newly created department of Vertical Videos within youtube, and that team is afraid of being laid off.
September 30, 2025 at 10:20 AM
There’s probably newly created department of Vertical Videos within youtube, and that team is afraid of being laid off.
584942417 years ought to be enough for anyone.
September 11, 2025 at 5:10 PM
584942417 years ought to be enough for anyone.
And in PAL they made a speedup to 25 fps. (Nobody noticed too.)
September 4, 2025 at 4:12 AM
And in PAL they made a speedup to 25 fps. (Nobody noticed too.)
No, read the explanation carefully.
September 3, 2025 at 7:11 AM
No, read the explanation carefully.
While more accurate, even more accurate would have been "example-driven parameters optimisation". ML is also a marketing move, which tries to hide that someone comes up with and fixes the model, someone gathers the examples. When they say ML, it's as if machine is growing and learning all by itself.
August 28, 2025 at 6:49 AM
While more accurate, even more accurate would have been "example-driven parameters optimisation". ML is also a marketing move, which tries to hide that someone comes up with and fixes the model, someone gathers the examples. When they say ML, it's as if machine is growing and learning all by itself.
I would let that region of spacetime containing him drift away. Why would I want continuity?
August 24, 2025 at 7:59 AM
I would let that region of spacetime containing him drift away. Why would I want continuity?
As in good cop/bad cop?
August 20, 2025 at 7:39 AM
As in good cop/bad cop?
We finally understood what a Jacobian is thanks to you.
August 18, 2025 at 7:48 PM
We finally understood what a Jacobian is thanks to you.
What misinformation? A simple question has been asked, total bullshit followed. That you can torture it into giving right answer is irrelevant. Imagine you bought a bottle of water to drink, which upon tasting turned out to be urine. ("You need to distillate this type of water to taste right!")
August 9, 2025 at 7:52 AM
What misinformation? A simple question has been asked, total bullshit followed. That you can torture it into giving right answer is irrelevant. Imagine you bought a bottle of water to drink, which upon tasting turned out to be urine. ("You need to distillate this type of water to taste right!")
Will they bother us with "I asked AI and it says you are wrong"?
August 8, 2025 at 4:44 AM
Will they bother us with "I asked AI and it says you are wrong"?
This isn't about technicalities of LLMs. It's about OpenAI acting like frauds. Do you think that the target users will use "PhD level AI" for simple questions, with already known answers? Do you think they will know what to do to make it answer correctly? Will they even notice that the answer is BS?
August 8, 2025 at 4:43 AM
This isn't about technicalities of LLMs. It's about OpenAI acting like frauds. Do you think that the target users will use "PhD level AI" for simple questions, with already known answers? Do you think they will know what to do to make it answer correctly? Will they even notice that the answer is BS?
Isn't this view dependent on unproven impossibility of efficient classical simulations of QM?
July 30, 2025 at 8:32 PM
Isn't this view dependent on unproven impossibility of efficient classical simulations of QM?
They can do some boilerplate code just fine... instead of people stopping to think why there's so much boilerplate and whether anything can be done to reduce it.
July 30, 2025 at 7:13 AM
They can do some boilerplate code just fine... instead of people stopping to think why there's so much boilerplate and whether anything can be done to reduce it.
The problem is that photos aren't uniform light spots and eyes aren't cameras. For example, looking at colorimetrically accurate photo of an LED-screen in a well-lit room almost always looks wrong with respect to what you'd see in person.
July 21, 2025 at 6:54 PM
The problem is that photos aren't uniform light spots and eyes aren't cameras. For example, looking at colorimetrically accurate photo of an LED-screen in a well-lit room almost always looks wrong with respect to what you'd see in person.