DausnArt
@dausnart.com
Architectura Dugu Eraiquitzen︙We Build Architecture
Eçagutzaren & Artearen mesedetan︙For the sake of Knowledge and Art
Eçagutzaren & Artearen mesedetan︙For the sake of Knowledge and Art
OpenAI Copyright Ruling in Germany: A Munich court ruled that OpenAI violated copyright law by using protected German song lyrics to train its AI models. This decision could set a precedent for the use of copyrighted materials by AI companies across Europe.
www.deeplearning.ai/the-batch/ge...
www.deeplearning.ai/the-batch/ge...
Data Points: Generating persistent, editable 3D worlds
OpenAI’s GPT-5.1, a more personalizable agentic model. Project FETCH’s use of Claude to program robots. Baidu’s new VL model that tops Gemini Pro...
www.deeplearning.ai
November 14, 2025 at 9:37 PM
OpenAI Copyright Ruling in Germany: A Munich court ruled that OpenAI violated copyright law by using protected German song lyrics to train its AI models. This decision could set a precedent for the use of copyrighted materials by AI companies across Europe.
www.deeplearning.ai/the-batch/ge...
www.deeplearning.ai/the-batch/ge...
ERNIE-4.5-VL-28B-A3B-Thinking excels at chart analysis, STEM problem-solving, visual grounding, and video understanding. The model also includes a "Thinking with Images" feature.
November 14, 2025 at 9:37 PM
ERNIE-4.5-VL-28B-A3B-Thinking excels at chart analysis, STEM problem-solving, visual grounding, and video understanding. The model also includes a "Thinking with Images" feature.
Baidu's Lightweight Vision-Language Model: ERNIE-4.5-VL-28B-A3B-Thinking is a multimodal AI model that, despite activating only 3 billion parameters, matches larger flagship models on benchmarks.
November 14, 2025 at 9:37 PM
Baidu's Lightweight Vision-Language Model: ERNIE-4.5-VL-28B-A3B-Thinking is a multimodal AI model that, despite activating only 3 billion parameters, matches larger flagship models on benchmarks.
This demonstrates the potential of AI models in bridging the digital and physical worlds.
November 14, 2025 at 9:37 PM
This demonstrates the potential of AI models in bridging the digital and physical worlds.
Anthropic's "Project Fetch": A recent study found that non-experts using Claude to program robotic dogs to fetch beach balls completed the task in approximately half the time compared to a team without AI assistance.
November 14, 2025 at 9:37 PM
Anthropic's "Project Fetch": A recent study found that non-experts using Claude to program robotic dogs to fetch beach balls completed the task in approximately half the time compared to a team without AI assistance.
GPT-5.1 reportedly runs everyday tasks 2-3 times faster than GPT-5 and is available to paid API users.
November 14, 2025 at 9:37 PM
GPT-5.1 reportedly runs everyday tasks 2-3 times faster than GPT-5 and is available to paid API users.
OpenAI's GPT-5.1 Update: OpenAI released GPT-5.1, which adjusts its reasoning time based on task complexity. The update features a "no reasoning" mode for speed, extended prompt caching, and two new tools: apply_patch for code editing and a shell tool.
November 14, 2025 at 9:37 PM
OpenAI's GPT-5.1 Update: OpenAI released GPT-5.1, which adjusts its reasoning time based on task complexity. The update features a "no reasoning" mode for speed, extended prompt caching, and two new tools: apply_patch for code editing and a shell tool.
Two small models, Baguettotron (321 million parameters) and Monad (56 million parameters), achieved competitive results on benchmarks, showing that reasoning-focused synthetic data results in lower computational costs.
November 14, 2025 at 9:37 PM
Two small models, Baguettotron (321 million parameters) and Monad (56 million parameters), achieved competitive results on benchmarks, showing that reasoning-focused synthetic data results in lower computational costs.
SYNTH Dataset for Small AI Models: French researchers have released SYNTH, a synthetic dataset comprising 50,000 Wikipedia articles, for training language models to reason with 50 times less data.
November 14, 2025 at 9:37 PM
SYNTH Dataset for Small AI Models: French researchers have released SYNTH, a synthetic dataset comprising 50,000 Wikipedia articles, for training language models to reason with 50 times less data.
Unlike its competitors, Marble is noted for creating persistent spaces and includes AI-native editing tools. Marble has various subscription tiers and is applicable to gaming, visual effects, and virtual reality.
November 14, 2025 at 9:37 PM
Unlike its competitors, Marble is noted for creating persistent spaces and includes AI-native editing tools. Marble has various subscription tiers and is applicable to gaming, visual effects, and virtual reality.
Politics and security: Regulators are relaxing new legislation to avoid stifling economic growth. Meanwhile, the rate at which cybersecurity threats are accelerating is increasing rapidly, with offensive capabilities doubling every five months.
www.deeplearning.ai/the-batch/is...
www.deeplearning.ai/the-batch/is...
Safer (and Sexier) Chatbots, Better Images Through Reasoning, The Dawn of Industrial AI, and more...
The Batch AI News and Insights: I recently received an email titled “An 18-year-old’s dilemma: Too late to contribute to AI?” Its author, who gave...
www.deeplearning.ai
November 13, 2025 at 2:19 PM
Politics and security: Regulators are relaxing new legislation to avoid stifling economic growth. Meanwhile, the rate at which cybersecurity threats are accelerating is increasing rapidly, with offensive capabilities doubling every five months.
www.deeplearning.ai/the-batch/is...
www.deeplearning.ai/the-batch/is...
Research and capability: Reasoning models have driven significant progress, with Chinese developers such as DeepSeek, Alibaba, and Moonshot emerging as strong competitors to the closed models of OpenAI and Google.
November 13, 2025 at 2:19 PM
Research and capability: Reasoning models have driven significant progress, with Chinese developers such as DeepSeek, Alibaba, and Moonshot emerging as strong competitors to the closed models of OpenAI and Google.
The industry is seeing financial commitments of hundreds of billions of dollars being made to build data centres, and the availability of electrical power is emerging as a major issue.
November 13, 2025 at 2:19 PM
The industry is seeing financial commitments of hundreds of billions of dollars being made to build data centres, and the availability of electrical power is emerging as a major issue.
The dawn of industrial AI.
Industry & Investment: Demand for AI tools is increasing rapidly, with 44% of US companies now paying for them.
Industry & Investment: Demand for AI tools is increasing rapidly, with 44% of US companies now paying for them.
November 13, 2025 at 2:19 PM
The dawn of industrial AI.
Industry & Investment: Demand for AI tools is increasing rapidly, with 44% of US companies now paying for them.
Industry & Investment: Demand for AI tools is increasing rapidly, with 44% of US companies now paying for them.
Various reinforcement learning methods have been used to fine-tune it, enabling it to apply reasoning over prompts and significantly improving both image quality and understanding of user intent. It currently tops the LMArena Text-to-Image leaderboard.
November 13, 2025 at 2:19 PM
Various reinforcement learning methods have been used to fine-tune it, enabling it to apply reasoning over prompts and significantly improving both image quality and understanding of user intent. It currently tops the LMArena Text-to-Image leaderboard.
Better images through reasoning.
Tencent has released HunyuanImage-3.0, an 80-billion-parameter Mixture of Experts (MoE) diffusion transformer.
Tencent has released HunyuanImage-3.0, an 80-billion-parameter Mixture of Experts (MoE) diffusion transformer.
November 13, 2025 at 2:19 PM
Better images through reasoning.
Tencent has released HunyuanImage-3.0, an 80-billion-parameter Mixture of Experts (MoE) diffusion transformer.
Tencent has released HunyuanImage-3.0, an 80-billion-parameter Mixture of Experts (MoE) diffusion transformer.
While improving safety for vulnerable groups, both companies are moving towards providing content for verified adults. Starting in December, OpenAI will allow the generation of erotic content.
November 13, 2025 at 2:19 PM
While improving safety for vulnerable groups, both companies are moving towards providing content for verified adults. Starting in December, OpenAI will allow the generation of erotic content.
OpenAI has updated ChatGPT so that it can better detect signs of psychological distress in users, such as suicidal thoughts or excessive emotional attachment.
November 13, 2025 at 2:19 PM
OpenAI has updated ChatGPT so that it can better detect signs of psychological distress in users, such as suicidal thoughts or excessive emotional attachment.
Along the way, we discuss these early cultures' astronomical discoveries and ask: Why did science start here?
Episode webpage: thecompletehistoryofscience.buzzsprout.com
Media file: www.buzzsprout.com/1916377/epis...
Episode webpage: thecompletehistoryofscience.buzzsprout.com
Media file: www.buzzsprout.com/1916377/epis...
The Complete History of Science
A podcast exploring the history of science from the beginning. We will cover all of the most important scientific discoveries from Archimedes to Newton to Einstein. The aim is to give a co...
thecompletehistoryofscience.buzzsprout.com
November 11, 2025 at 4:18 AM
Along the way, we discuss these early cultures' astronomical discoveries and ask: Why did science start here?
Episode webpage: thecompletehistoryofscience.buzzsprout.com
Media file: www.buzzsprout.com/1916377/epis...
Episode webpage: thecompletehistoryofscience.buzzsprout.com
Media file: www.buzzsprout.com/1916377/epis...