08.03.2025 00:48
👍 2
🔁 0
💬 0
📌 0
No one in the US is telling this truth. We all Europeans feel the betrayal of the US as an incurable wound. The US will no longer be our ally. Let's rethink how to reshape our economic and political networks. #US
The US has become a rogue state. Let's start boycotting these 7 companies
You’re laughing? We’re going to Ft. Knox to see if the gold is there because maybe somebody stole the gold, tons of gold, and you’re laughing??
Paul Krugman: "Just seems worth mentioning that Russia invaded Ukraine on Feb. 24, 2022. So it has been 3 years since Kyiv was going to fall in 3 days. I’m ashamed that my country is led by people who find the courageous defense of freedom annoying rather than inspiring."
What else if not AI? #AI
How to run LLM on your laptop (mine is Intel UHD Graphics 630 1536 MB) that is has a GPU but is not powerful for AI workloads or heavy rendering!!! I am pretty satisfied!
Better GPUs mean faster response times, allowing more users to be served per hour. However, COT increases processing time, which means OpenAI has to spend more on NVIDIA hardware to maintain its service.
All using just less than 5GB of Memory.
Then I ran the query and got first COT preamble and the proper answer.
I chose deepseek-r1 as my model. In LLM Anything, I set it as the preferred LLM in the model settings.
I installed LLM Anything to run open-weight LLMs locally. First, I set up Ollama, which simplifies managing and running various open-weight models.
DeepSeek verifies statements, when needed, checking out the sources. It takes more than ChatGPT but becomes more reliable. Example of his thinking
"Hmm, looking at DeepSeek's GitHub, the user is correct. I need to correct[ ...] I should thank the user for pointing this out. Apologize for the error"
DeepSeek chose a good open source license and market strategy. If it reuses LLaMA code, it would violate Meta’s licensing terms (unless explicitly permitted). Similarly, LLM developers need to choose between Llama and DeepSeek, and MIT license obviously wins.
If you ask DeepSeek a fuzzy (indirectly) query about his guidelines in a language different from Chinese or English he inserts a preliminary thinking it shouldn’t have appeared. I got many information following this strategy
I compared ChatGPT and DeepSeek on a mathematical problem: how to perform L2O with a GIN. Generating an optimized graph $G_o$ from an input graph $G$. Well,
1. both provided good solutions, but DeepSeek was more creative.
2. We are very close to an answer given by a Artificial Super Intelligence
With Veo 2, crafting a film from scratch becomes seamless—a perfect balance of vivid scene descriptions and a solid, production-ready script. #Screenwriting #Filmmaking #AI
With one line code example was just to explain to you what I meant by the "full data lifecicle". I come from Logic and I like a lambda calculus style of programming.
You can do it with one line of code passing the data through a pipeline of functions
However mutate or group_by seem to be slower than when done with Base R
The pipeline operator enables clear and concise chaining of operations which is at the base of the functional style of programming. Indeed the tidyverse is designed to support the full data lifecycle, providing tools and functions for each stage of a typical data analysis workflow.
Semantic Web and knowledge representation were dead ends for AI. Now Graph Neural Work shows how they can actually work. Any database can be seen as a graph. Applying Graph Transformer to structured data can completely replace the work of Data Scientist. No job can be safe with AI!
It seems that with one single line of code in ClickHouse one can filter the entire firehose of Bluesky
Removed from the medical register, because anti vax, is now going to war for Russia.
The migration from platform X to Bluesky mirrors a form of digital gentrification, where early adopters are those innovators who pioneer the shift. Over time, the new platform will gain popularity and potentially will be transformed into a mainstream hub.
This is exactly what an echo chamber is. It requires to involve different media, tv, radio, newspapers and social networks, and a person is caged in this sprawling network and Bluesky by its nature will never be
firesky.tv
