Alexander Kruel
@dakruel.bsky.social
“We are more often frightened than hurt; and we suffer more from imagination than from reality” ― Seneca
More: https://axisofordinary.substack.com/
More: https://axisofordinary.substack.com/
Watch Google's Waymo self-driving car avoid calamity with a scooter rider
December 17, 2024 at 4:01 PM
Watch Google's Waymo self-driving car avoid calamity with a scooter rider
Google's new state-of-the-art AI video model is absolutely brilliant: blog.google/technology/g...
Prompt for the video: A pirate canon shooting meatballs into a spicy hotpot
Prompt for the video: A pirate canon shooting meatballs into a spicy hotpot
December 17, 2024 at 8:44 AM
Google's new state-of-the-art AI video model is absolutely brilliant: blog.google/technology/g...
Prompt for the video: A pirate canon shooting meatballs into a spicy hotpot
Prompt for the video: A pirate canon shooting meatballs into a spicy hotpot
🌘 Earthrise was stunning as Alex’s ship neared the moon base. Fully automated docking gave him time to enjoy the view—yet, unease lingered.
December 15, 2024 at 12:50 PM
🌘 Earthrise was stunning as Alex’s ship neared the moon base. Fully automated docking gave him time to enjoy the view—yet, unease lingered.
OpenAI announces a new version of its video generation model Sora.
THE VIDEO BELOW IS 100% AI GENERATED.
You can watch the live announcement here: www.youtube.com/watch?v=2jKV...
You will be able to access the model at sora.com later today.
THE VIDEO BELOW IS 100% AI GENERATED.
You can watch the live announcement here: www.youtube.com/watch?v=2jKV...
You will be able to access the model at sora.com later today.
December 9, 2024 at 6:15 PM
OpenAI announces a new version of its video generation model Sora.
THE VIDEO BELOW IS 100% AI GENERATED.
You can watch the live announcement here: www.youtube.com/watch?v=2jKV...
You will be able to access the model at sora.com later today.
THE VIDEO BELOW IS 100% AI GENERATED.
You can watch the live announcement here: www.youtube.com/watch?v=2jKV...
You will be able to access the model at sora.com later today.
Meet Willow: Google’s state-of-the-art quantum chip. It's the first quantum chip to show exponential error reduction as qubits scale, paving the way for large-scale, fault-tolerant quantum computers. blog.google/technology/r...
December 9, 2024 at 5:42 PM
Meet Willow: Google’s state-of-the-art quantum chip. It's the first quantum chip to show exponential error reduction as qubits scale, paving the way for large-scale, fault-tolerant quantum computers. blog.google/technology/r...
via via Riley Goodside
December 7, 2024 at 10:58 PM
via via Riley Goodside
Google just introduced 🧞Genie 2 🧞: A large-scale foundation world model creating a path to unlimited environments for training and evaluating embodied agents.
Read more: deepmind.google/discover/blo...
Read more: deepmind.google/discover/blo...
December 4, 2024 at 5:45 PM
Google just introduced 🧞Genie 2 🧞: A large-scale foundation world model creating a path to unlimited environments for training and evaluating embodied agents.
Read more: deepmind.google/discover/blo...
Read more: deepmind.google/discover/blo...
The reason is that LLMs "see" tokens instead of letters. Of course, they can get it right if they "concentrate". But it's much more difficult for them than for humans.
December 2, 2024 at 1:43 AM
The reason is that LLMs "see" tokens instead of letters. Of course, they can get it right if they "concentrate". But it's much more difficult for them than for humans.
Counting the number of black dots in this image is the human equivalent of asking an LLM how many 'r's there are in the word strawberry.
December 1, 2024 at 11:29 PM
Counting the number of black dots in this image is the human equivalent of asking an LLM how many 'r's there are in the word strawberry.
This phase of our prehistory is preserved in instincts and in cultural memory in the form of several ancient myths.
The uncanny valley is a remnant of this. Evolution instilled a fear of the not-quite-human in us due to millions of years of inter-hominid-species conflicts.
The uncanny valley is a remnant of this. Evolution instilled a fear of the not-quite-human in us due to millions of years of inter-hominid-species conflicts.
November 29, 2024 at 2:20 PM
This phase of our prehistory is preserved in instincts and in cultural memory in the form of several ancient myths.
The uncanny valley is a remnant of this. Evolution instilled a fear of the not-quite-human in us due to millions of years of inter-hominid-species conflicts.
The uncanny valley is a remnant of this. Evolution instilled a fear of the not-quite-human in us due to millions of years of inter-hominid-species conflicts.
Newly discovered footprints show that at least two hominid species were walking through the muddy submerged edge of a lake in Kenya’s Turkana Basin at the same time, about 1.5 million years ago.
www.science.org/doi/10.1126/...
www.science.org/doi/10.1126/...
November 29, 2024 at 2:20 PM
Newly discovered footprints show that at least two hominid species were walking through the muddy submerged edge of a lake in Kenya’s Turkana Basin at the same time, about 1.5 million years ago.
www.science.org/doi/10.1126/...
www.science.org/doi/10.1126/...
November 29, 2024 at 1:02 PM
QwQ: Reflect Deeply on the Boundaries of the Unknown (32B beats O1 mini and competes with O1 preview) qwenlm.github.io/blog/qwq-32b...
“What does it mean to think, to question, to understand? These are the deep waters that QwQ (Qwen with Questions) wades into.”
“What does it mean to think, to question, to understand? These are the deep waters that QwQ (Qwen with Questions) wades into.”
November 27, 2024 at 10:03 PM
QwQ: Reflect Deeply on the Boundaries of the Unknown (32B beats O1 mini and competes with O1 preview) qwenlm.github.io/blog/qwq-32b...
“What does it mean to think, to question, to understand? These are the deep waters that QwQ (Qwen with Questions) wades into.”
“What does it mean to think, to question, to understand? These are the deep waters that QwQ (Qwen with Questions) wades into.”
November 20, 2024 at 9:36 PM
OpenAI has released a new version of ChatGPT-4o that excels at creative writing.
Here you can see the creativity difference in the MinecraftBench (how good are models at building in minecraft).
Left: old model
Right: new model
Prompt: "create a massive church of yourself"
Here you can see the creativity difference in the MinecraftBench (how good are models at building in minecraft).
Left: old model
Right: new model
Prompt: "create a massive church of yourself"
November 20, 2024 at 9:32 PM
OpenAI has released a new version of ChatGPT-4o that excels at creative writing.
Here you can see the creativity difference in the MinecraftBench (how good are models at building in minecraft).
Left: old model
Right: new model
Prompt: "create a massive church of yourself"
Here you can see the creativity difference in the MinecraftBench (how good are models at building in minecraft).
Left: old model
Right: new model
Prompt: "create a massive church of yourself"
In Estonia, a gravel road called the "Saatse Boot" crosses into Russian territory for about 1 km. Drivers can pass without border checks, but stopping or walking there is prohibited. Breaking the rules may lead to arrest by Russian border guards. en.wikipedia.org/wiki/Saatse_...
November 20, 2024 at 5:59 PM
In Estonia, a gravel road called the "Saatse Boot" crosses into Russian territory for about 1 km. Drivers can pass without border checks, but stopping or walking there is prohibited. Breaking the rules may lead to arrest by Russian border guards. en.wikipedia.org/wiki/Saatse_...
Chinese replication of the OpenAI o1 paradigm. They also replicated the o1 scaling laws (longer reasoning leads to better performance).
Note: It appears to be less efficient than the original OpenAI implementation and may not be as intelligent as OpenAI o1-preview.
Try it now: chat.deepseek.com
Note: It appears to be less efficient than the original OpenAI implementation and may not be as intelligent as OpenAI o1-preview.
Try it now: chat.deepseek.com
November 20, 2024 at 4:17 PM
Chinese replication of the OpenAI o1 paradigm. They also replicated the o1 scaling laws (longer reasoning leads to better performance).
Note: It appears to be less efficient than the original OpenAI implementation and may not be as intelligent as OpenAI o1-preview.
Try it now: chat.deepseek.com
Note: It appears to be less efficient than the original OpenAI implementation and may not be as intelligent as OpenAI o1-preview.
Try it now: chat.deepseek.com
“One way to test the robustness of the policy is to mess with the robot. Some of the π₀ corrections are pretty impressive.”
More: www.physicalintelligence.company
More: www.physicalintelligence.company
November 19, 2024 at 4:07 PM
“One way to test the robustness of the policy is to mess with the robot. Some of the π₀ corrections are pretty impressive.”
More: www.physicalintelligence.company
More: www.physicalintelligence.company
Robot watches how-to videos and becomes an expert surgeon: hub.jhu.edu/2024/11/11/s...
The team from Johns Hopkins and Stanford Universities harnessed imitation learning. It allowed the robot to learn from a vast archive of surgical videos, eliminating the need for programming each move.
The team from Johns Hopkins and Stanford Universities harnessed imitation learning. It allowed the robot to learn from a vast archive of surgical videos, eliminating the need for programming each move.
November 17, 2024 at 1:41 PM
Robot watches how-to videos and becomes an expert surgeon: hub.jhu.edu/2024/11/11/s...
The team from Johns Hopkins and Stanford Universities harnessed imitation learning. It allowed the robot to learn from a vast archive of surgical videos, eliminating the need for programming each move.
The team from Johns Hopkins and Stanford Universities harnessed imitation learning. It allowed the robot to learn from a vast archive of surgical videos, eliminating the need for programming each move.