@zagan1.bsky.social
"At the heart of science is an essential balance between two seemingly contradictory attitudes--an openness to new ideas, no matter how bizarre or counterintuitive they may be, and the most ruthless skeptical scrutiny of all ideas, old and new." Carl Sagan
Ja, deutsche "Patrioten", die Europa am liebsten zugleich an die Moskauer und Pekinger Machthaber verscherbeln würden, sitzen definitiv im Glashaus, wenn sie Korruption beklagen.
February 4, 2026 at 1:12 AM
Ja, deutsche "Patrioten", die Europa am liebsten zugleich an die Moskauer und Pekinger Machthaber verscherbeln würden, sitzen definitiv im Glashaus, wenn sie Korruption beklagen.
Fronturlaubsjob für ukrainische Bodendrohnen?
defenceleaders.com/news/ukraini...
defenceleaders.com/news/ukraini...
Ukrainian Combat Robot Holds Frontline Position for Six Weeks in Sign of Growing UGV Maturity
Uncover the significance of the UGV holds position for 6 weeks and its impact on frontline military strategies.
defenceleaders.com
January 22, 2026 at 3:33 AM
Fronturlaubsjob für ukrainische Bodendrohnen?
defenceleaders.com/news/ukraini...
defenceleaders.com/news/ukraini...
Tell me more about how golf courses have achieved any of these 15 scientific feats lately:
youtu.be/aYppKInSS4I
youtu.be/aYppKInSS4I
Top 15 New Discoveries MADE By AI
YouTube video by AI Uncovered
youtu.be
January 22, 2026 at 3:23 AM
Tell me more about how golf courses have achieved any of these 15 scientific feats lately:
youtu.be/aYppKInSS4I
youtu.be/aYppKInSS4I
You repeat all these lame anti-AI talking points without critical thought and then accuse me of letting AI reply for me, and the irony is somehow lost on you. Hilarious.
January 22, 2026 at 3:21 AM
You repeat all these lame anti-AI talking points without critical thought and then accuse me of letting AI reply for me, and the irony is somehow lost on you. Hilarious.
Don't get me wrong, current AI is jagged and uneven with deficits in physical and spatial reasoning in particular, and its ability to generalize from training data is still limited, but it's a lot better than nothing.
January 22, 2026 at 3:19 AM
Don't get me wrong, current AI is jagged and uneven with deficits in physical and spatial reasoning in particular, and its ability to generalize from training data is still limited, but it's a lot better than nothing.
New drug candidates for cancer research are being discovered in record time:
www.thelancet.com/journals/lan...
www.thelancet.com/journals/lan...
Cancer drug discovery at warp speed: can AI deliver?
In April, 2025, Demis Hassabis, the Nobel Prize-winning CEO of Google DeepMind, appeared
on the US television show 60 Minutes and suggested that all diseases could be curable
with the help of artifici...
www.thelancet.com
January 22, 2026 at 3:19 AM
New drug candidates for cancer research are being discovered in record time:
www.thelancet.com/journals/lan...
www.thelancet.com/journals/lan...
They accelerate research on mathematical proofs, identify useful new materials, improve computing infrastructure, chip designs and mathematical algorithms beyond prior limits.
www.arxiv.org/abs/2506.13131
www.arxiv.org/abs/2506.13131
AlphaEvolve: A coding agent for scientific and algorithmic discovery
In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific...
www.arxiv.org
January 22, 2026 at 3:18 AM
They accelerate research on mathematical proofs, identify useful new materials, improve computing infrastructure, chip designs and mathematical algorithms beyond prior limits.
www.arxiv.org/abs/2506.13131
www.arxiv.org/abs/2506.13131
The next level of machine intelligence is to produce new knowledge not just in tests but in real-world cutting-edge science, and again: multimodal LLMs generate hypotheses, design experiments, and iterate based on results.
Plan->test->evaluate->revise->repeat.
Plan->test->evaluate->revise->repeat.
January 22, 2026 at 3:17 AM
The next level of machine intelligence is to produce new knowledge not just in tests but in real-world cutting-edge science, and again: multimodal LLMs generate hypotheses, design experiments, and iterate based on results.
Plan->test->evaluate->revise->repeat.
Plan->test->evaluate->revise->repeat.
..success depends on logical rigor and builder creativity impossible to fake by a parrot. These competitions pose original problems whose solutions cannot be lifted from AI training data. Yet the AIs did well, which demonstrates intelligence: the ability to figure out new and unanticipated problems.
January 22, 2026 at 3:16 AM
..success depends on logical rigor and builder creativity impossible to fake by a parrot. These competitions pose original problems whose solutions cannot be lifted from AI training data. Yet the AIs did well, which demonstrates intelligence: the ability to figure out new and unanticipated problems.
Well, technically yes, but so is the human brain. Your implicit assertion here is that AI has no intelligence whatsoever and only spits out rote-learned text blocks, which is demonstrably wrong.
General-purpose generative AIs recently won gold medals at international math and coding olympiads where
General-purpose generative AIs recently won gold medals at international math and coding olympiads where
January 22, 2026 at 3:14 AM
Well, technically yes, but so is the human brain. Your implicit assertion here is that AI has no intelligence whatsoever and only spits out rote-learned text blocks, which is demonstrably wrong.
General-purpose generative AIs recently won gold medals at international math and coding olympiads where
General-purpose generative AIs recently won gold medals at international math and coding olympiads where
... (successfully completed tasks like AI generating a correct 3D protein structure or mathematical proof) every 9 months. It's not nothing.
Then you claim that AI systems are "just compression systems built of the most likely answer". This half-truth can be more deceptive than a plain lie.
Then you claim that AI systems are "just compression systems built of the most likely answer". This half-truth can be more deceptive than a plain lie.
January 22, 2026 at 3:13 AM
... (successfully completed tasks like AI generating a correct 3D protein structure or mathematical proof) every 9 months. It's not nothing.
Then you claim that AI systems are "just compression systems built of the most likely answer". This half-truth can be more deceptive than a plain lie.
Then you claim that AI systems are "just compression systems built of the most likely answer". This half-truth can be more deceptive than a plain lie.
Power efficiency of Nvidia's new Vera Rubin platform surged five-fold over Blackwell in just 22 months.
Next, multiply that with the rate of improving algorithmic efficiency as AI models get pruned and distilled down. The same compute budget yields about twice the performance...
Next, multiply that with the rate of improving algorithmic efficiency as AI models get pruned and distilled down. The same compute budget yields about twice the performance...
January 22, 2026 at 3:10 AM
Power efficiency of Nvidia's new Vera Rubin platform surged five-fold over Blackwell in just 22 months.
Next, multiply that with the rate of improving algorithmic efficiency as AI models get pruned and distilled down. The same compute budget yields about twice the performance...
Next, multiply that with the rate of improving algorithmic efficiency as AI models get pruned and distilled down. The same compute budget yields about twice the performance...
"Resource" in my previous post is not a basic resource like copper and sugar but computing power and its ability to solve problems in STEM fields, which does become exponentially more affordable. Hardware efficiency has been doubling every 2 years, an estimate that looked conservative recently.
January 22, 2026 at 3:06 AM
"Resource" in my previous post is not a basic resource like copper and sugar but computing power and its ability to solve problems in STEM fields, which does become exponentially more affordable. Hardware efficiency has been doubling every 2 years, an estimate that looked conservative recently.
Es ist wie bei Putin: je mehr Europa zurückweicht, desto härter schlägt er zu und desto ungeheuerlicher werden die Methoden und Risiken, die er auf seinem imperialen Kreuzzug eingeht. Bleibt man aber unbeugsam, wird er vorsichtig und sucht sich ein leichteres Opfer.
January 21, 2026 at 2:10 AM
Es ist wie bei Putin: je mehr Europa zurückweicht, desto härter schlägt er zu und desto ungeheuerlicher werden die Methoden und Risiken, die er auf seinem imperialen Kreuzzug eingeht. Bleibt man aber unbeugsam, wird er vorsichtig und sucht sich ein leichteres Opfer.
when resources get cheap and plentiful, more can be forked out to basic research that doesn't generate cash flow immediately. The AI revolution is driving an efficiency revolution in computation and the energy grid that facilitates access for everyone including research labs.
January 21, 2026 at 1:34 AM
when resources get cheap and plentiful, more can be forked out to basic research that doesn't generate cash flow immediately. The AI revolution is driving an efficiency revolution in computation and the energy grid that facilitates access for everyone including research labs.
... they were computationally prohibitive until AI-accelerated modelling came of age using intelligent approximations.
Science is rarely the primary goal of economic activity. Rather, it's enabled by the economies of scale (cost advantages gained from scaled-up production):
Science is rarely the primary goal of economic activity. Rather, it's enabled by the economies of scale (cost advantages gained from scaled-up production):
January 21, 2026 at 1:29 AM
... they were computationally prohibitive until AI-accelerated modelling came of age using intelligent approximations.
Science is rarely the primary goal of economic activity. Rather, it's enabled by the economies of scale (cost advantages gained from scaled-up production):
Science is rarely the primary goal of economic activity. Rather, it's enabled by the economies of scale (cost advantages gained from scaled-up production):
And lots of folks use the internet to look at p0rn and cat pictures, yet Alphafold and AlphaEvolve were developed at a web search company.
Molecular biology simulations aren't small but mind-blowingly intricate. They took so long to take off after the human genome project precisely because...
Molecular biology simulations aren't small but mind-blowingly intricate. They took so long to take off after the human genome project precisely because...
January 21, 2026 at 1:26 AM
And lots of folks use the internet to look at p0rn and cat pictures, yet Alphafold and AlphaEvolve were developed at a web search company.
Molecular biology simulations aren't small but mind-blowingly intricate. They took so long to take off after the human genome project precisely because...
Molecular biology simulations aren't small but mind-blowingly intricate. They took so long to take off after the human genome project precisely because...
... and should not be gated like physical objects.
Copyright claims encompassing neural patterns representing general styles of art are far more extensive than even the demands of copyright maximalists back in the day.
Copyright claims encompassing neural patterns representing general styles of art are far more extensive than even the demands of copyright maximalists back in the day.
January 21, 2026 at 1:25 AM
... and should not be gated like physical objects.
Copyright claims encompassing neural patterns representing general styles of art are far more extensive than even the demands of copyright maximalists back in the day.
Copyright claims encompassing neural patterns representing general styles of art are far more extensive than even the demands of copyright maximalists back in the day.
(literally, as in Very Concerned Citizens playing vinyl records backward to listen for demonic self-harm messages) and mp3 filesharing on FTP servers ("Thieves! Pirates! Criminals!") with similar arguments. We should not blame terrible life choices on tools and media, and information cannot be owned
January 21, 2026 at 1:24 AM
(literally, as in Very Concerned Citizens playing vinyl records backward to listen for demonic self-harm messages) and mp3 filesharing on FTP servers ("Thieves! Pirates! Criminals!") with similar arguments. We should not blame terrible life choices on tools and media, and information cannot be owned
No invention, from the simplest to the most sophisticated, has ever been risk-free. More lives are saved each day by AI alerting users to a medical problem than lost by someone committing suicide after a chat session.
I'm old enough to remember when people were demonising Heavy Metal music...
I'm old enough to remember when people were demonising Heavy Metal music...
January 21, 2026 at 1:22 AM
No invention, from the simplest to the most sophisticated, has ever been risk-free. More lives are saved each day by AI alerting users to a medical problem than lost by someone committing suicide after a chat session.
I'm old enough to remember when people were demonising Heavy Metal music...
I'm old enough to remember when people were demonising Heavy Metal music...
Knowing of closed-loop water-cooled PCs, I could guess before factchecking that the water issue must be vastly exaggerated, and lo and behold: even golf courses eclipse datacenters when it comes to water use. Andy Masley has the lowdown:
open.substack.com/pub/andymasl...
open.substack.com/pub/andymasl...
The AI water issue is fake
On the national, local, and personal level
open.substack.com
January 21, 2026 at 1:18 AM
Knowing of closed-loop water-cooled PCs, I could guess before factchecking that the water issue must be vastly exaggerated, and lo and behold: even golf courses eclipse datacenters when it comes to water use. Andy Masley has the lowdown:
open.substack.com/pub/andymasl...
open.substack.com/pub/andymasl...
Your reaction reminds me there are Luddite atheists as willfully ignorant as creationists. I do not need a chatbot to point out developments that have been in my news feed since the noughties, and neither does anyone else with an ounce of curiosity about science and the future of humanity.
January 21, 2026 at 1:16 AM
Your reaction reminds me there are Luddite atheists as willfully ignorant as creationists. I do not need a chatbot to point out developments that have been in my news feed since the noughties, and neither does anyone else with an ounce of curiosity about science and the future of humanity.
"parasitic" datacenters only bring tutoring to underserved communities, major coding powers to minor indie developers, better materials to electronics & machinery, efficiency to residential heating and the healthcare system, reliable biomolecular simulations to drug discovery and cancer research,...
January 20, 2026 at 4:12 AM
"parasitic" datacenters only bring tutoring to underserved communities, major coding powers to minor indie developers, better materials to electronics & machinery, efficiency to residential heating and the healthcare system, reliable biomolecular simulations to drug discovery and cancer research,...