Close

Explore AI

Explore GPTs

Explore AI News

Explore AI Videos

Explore AI for Jobs

Latest Scikit-optimize AI Videos

Find the latest for Scikit-optimize company videos

Follow
... 10:33
SmolDocling OCR: The Best Open Source AI Model for OCR 🚀

AI Anytime

1+ months ago 2K views

Small Doling is an ultra-compact open-source vision language model designed for document understanding and OCR tasks. It boasts 256 million parameters and outperforms existing models like M in optical character recognition. The model provides efficient processing capabilities for various document types and can handle tables, equations, and other structures. Demonstrated through an app, users can convert images to text, markdown, and other formats quickly. Future improvements include enhanced chart recognition and multi-page inference. The accessible codebase supports broader experimentation and implementation in document processing tasks.
03:16 - 03:43

Overview of Small Doling OCR app functionalities for document processing.

Image Recognition Education
... 7:44
Best AI Investing Finance Tracker: Finance Professor Confirms

Investing Simplified - Professor G

1+ months ago 13K views

Quadratic simplifies finance management through AI-driven spreadsheets that allow users to easily track investments and consolidate data from multiple accounts. The tool is designed for both technical and non-technical users, enabling seamless incorporation of coding languages like Python, JavaScript, and SQL for complex analysis. Users can generate visual aids like pie and bar charts with simple text commands, making data visualization straightforward. Quadratic promotes collaboration across teams by guiding users in data management without requiring deep coding knowledge, enhancing productivity and supporting informed decision-making in finance and business operations.
04:52 - 05:18

Built-in AI empowers users across roles to handle data without coding.

Quadratic Predictive Analytics Finance
... 23:22
How to Choose the Best LLM (Grok 3) for Your AI Agent

Discover AI

1+ months ago 2K views

Optimization in multi-agent systems enhances performance by selecting the most suitable Large Language Model (LLM) for specific subtasks, even if individual models provide incorrect answers. Coordinating multiple LLMs can yield better outcomes by combining their strengths in solving tasks like arithmetic. A model selector framework has been introduced, which can significantly improve task performance, leveraging different strengths and weaknesses of LLMs based on task requirements and effectiveness, demonstrating that effective model selection leads to notable efficiency gains and more accurate results.
04:26 - 04:31

GRO-3 struggles with simple arithmetic, illustrating limitations in certain tasks.

Berkeley Natural Language Processing (NLP) Education
... 10:53
S* for AI CODE Generation: Plus 100%

Discover AI

1+ months ago 3K views

SAR methodology aims to enhance local AI coding performance by doubling normal output without cloud dependency. By transitioning from a 3 billion to a 14 billion parameter model, substantial improvements are seen in local code efficiency. The method incorporates test time scaling and iterative debugging, enabling parallel sampling and refining generated code effectively. Utilizing modern architecture models, the approach shows promising results, outperforming larger proprietary models while promoting independence from cloud resources. This advancement aims to improve accessibility and performance for individual developers and smaller teams, fostering a competitive environment in AI code generation.
06:27 - 07:35

The success of SAR in enhancing performance metrics across various models discussed.

Machine Learning Research & Innovations
... 11:6
New AI Just Beat DeepSeek With Almost No Effort! (This Shouldn't Be Possible!)

AI Revolution

1+ months ago 61K views

Open Thinker 32B is an open-source AI model that outperforms several proprietary models while using only 14% of the data from its competitors. Its training utilized a meticulously curated dataset of 114,000 samples, enhancing its performance in mathematical and logical problem-solving tasks. The model achieves impressive benchmark scores, showing potential for further optimization and development. In conjunction, Hugen 3.5D employs a unique approach of latent reasoning, using a recurrent architecture to enable iterative passes for problem-solving, making it efficient for complex queries while maintaining memory efficiency.
04:20 - 04:34

Open Thinker 32B outperforms Deep Seek R1 in primary benchmarks.

Machine Learning Research & Innovations
... 5:18
Make Your AI Agents Smarter, Faster and Cheaper

Ken Kai does AI

1+ months ago 2K views

The video presents techniques to optimize AI tool usage by reducing prompt sizes and enhancing query routing. It discusses the inefficiencies of using excessive tokens and irrelevant tool descriptions in AI prompts. By refining tool names and descriptions, the overall reliability and speed of the AI response improve significantly while also lowering costs. The speaker demonstrates practical examples by querying various tools and showcases the streamlined process, emphasizing the importance of concise and precise AI communication for better results.
02:34 - 04:25

Optimizing prompts increases response speed and AI reliability while cutting expenses.

Machine Learning Education
... 9:26
Enhancing AI Agents Through Fine Tuning & Model Customization

IBM Technology

1+ months ago 4K views

Model fine-tuning plays a critical role in enhancing the performance and reliability of agentic AI systems, which are designed to tackle complex, multi-step problems with varying degrees of autonomy. By addressing shortcomings such as high token usage, execution costs, and error propagation, organizations can improve their agents' decision-making and tool use. Effective data collection methods are essential, focusing on teaching models when and how to use specific tools, as well as providing context-driven reasoning instruction. Continued iterative improvements through execution data analysis are vital for aligning AI systems with organizational goals and ensuring their robustness.
03:04 - 03:26

Error propagation in agentic systems can lead to increased failure rates.

IBM Machine Learning Education
... 10:19
Running DeepSeek AI on a Raspberry Pi

Pi My Life Up

1+ months ago 2K views

DeepSeekAI models are tested on a Raspberry Pi. The full model can't run due to the device's computational limitations, so distilled versions are utilized. These smaller models are optimized to run efficiently on limited hardware. The installation process involves updating packages and using the Ollama tool. Three distilled models are tested, revealing a trade-off between model complexity and response time, with insights on performance metrics and the impact of model size on accuracy. The video concludes with guidance on potential performance improvements and system requirements for running larger models.
06:35 - 06:51

Larger 7 billion parameter models generate better responses but slower performance.

AI hardware Tech & Hardware
... 10:19
Running DeepSeek AI on a Raspberry Pi

Pi My Life Up

1+ months ago 2K views

DeepSeekAI models are tested on a Raspberry Pi. The full model can't run due to the device's computational limitations, so distilled versions are utilized. These smaller models are optimized to run efficiently on limited hardware. The installation process involves updating packages and using the Ollama tool. Three distilled models are tested, revealing a trade-off between model complexity and response time, with insights on performance metrics and the impact of model size on accuracy. The video concludes with guidance on potential performance improvements and system requirements for running larger models.
06:35 - 06:51

Larger 7 billion parameter models generate better responses but slower performance.

AI hardware Tech & Hardware