Open Source News & Articles on Techminis

A naukri.com initiative

New

Home

Open Source News

Marktechpost

211

Yandex Develops and Open-Sources Perforator: An Open-Source Tool that can Save Businesses Billions of Dollars a Year on Server Infrastructure

Yandex has developed and open-sourced Perforator, a tool for real-time monitoring and analysis of servers and applications.
Perforator identifies code inefficiencies and supports profile-guided optimization, which can lower infrastructure costs by up to 20% for companies.
The tool helps companies optimize their code to reduce server load, lowering energy and equipment costs as a result.
Perforator can be deployed on company servers, providing organizations with control over their data and making it a strong fit for firms with stringent data security requirements.
Perforator is designed to support native programming languages such as C, C++, Go, Rust, Python, and Java and has essential features tailored to large organizations, such as A/B testing capabilities.
Open-sourcing Perforator reflects Yandex's commitment to fostering community collaboration in developing system technologies.
Perforator’s source code is now available on GitHub, alongside other Yandex open-source solutions, such as YaFSDP.
Yandex is a global technology company, building intelligent products and services with the aim of helping consumers and businesses navigate the online and offline world.
Perforator can help companies reduce CPU resource usage by up to 20%.
Perforator is accessible for free on GitHub and will be enhanced with additional capabilities in the future, such as more precise analysis of events.

Read Full Article

12 Likes

Insider

330

Image Credit: Insider

OpenAI is reaping what it sowed with DeepSeek. What's that old saying about karma?

OpenAI said that Chinese lab DeepSeek may have used OpenAI model outputs 'inappropriately.'
OpenAI has been accused of using copyrighted content to train their own AI models.
Some AI experts believe that the complaints against DeepSeek are hypocritical considering OpenAI's success in using content they did not have permission to use.
OpenAI is being sued by authors who claim this is breaking copyright law by using their books to train their AI models.
DeepSeek may be using OpenAI outputs in the process of distillation, a technique used to extract intelligence and create new models.
Some AI researchers believe DeepSeek used outputs from OpenAI's new reasoning model to create synthetic data to improve its own models.
The rise of DeepSeek and the 'training data free-for-all' have resulted in backfiring on OpenAI.
The use of copyrighted and original content without permission is being fought in court with a dispute over fair use.
'Fair use' doctrine may apply to both OpenAI and DeepSeek in these situations.
This reckoning may encourage data creators to demand appropriate credit and compensation for their content.

Read Full Article

19 Likes

TechCrunch

Image Credit: TechCrunch

Ai2 says its new AI model beats one of DeepSeek’s best

Ai2, a nonprofit AI research institute based in Seattle, has released a new AI model called Tulu3-405B.
Tulu3-405B outperforms DeepSeek V3, a leading system by Chinese AI company DeepSeek, and beats OpenAI's GPT-4o on certain AI benchmarks.
Unlike GPT-4o and DeepSeek V3, Tulu3-405B is open source and freely available for replication from scratch.
The model, containing 405 billion parameters, achieved competitive performance through reinforcement learning with verifiable rewards.

Read Full Article

5 Likes

Hackernoon

215

Image Credit: Hackernoon

Could China’s DeepSeek Shatter the ‘More GPUs, More Power’ Theory?

Chinese AI startup DeepSeek has released its open-source reasoning model, R1; it claims the tech matches or outperforms OpenAI's ChatGPT on benchmarks for a fraction of the cost
The company said its approach included using heterogeneous infrastructure, reducing numerical precision and applying a technique called an intelligent mixture of experts, which it said had cut training costs from $100m to $5m and reduced hardware needs from 100,000 graphics processing units to 2,000
DeepSeek is part of a group of Chinese companies developing their own graphics processing units that can rival Nvidia and other international chip firms
OpenAI, Oracle, Softbank and MGX's Stargate Project, which President Trump has called 'the largest AI infrastructure project in history', aims to invest $500bn in US AI infrastructure before 2029
Nvidia currently holds a market lead in graphics processing units, though rival AMD has launched some promising products which use application-specific integrated circuits
Edge and embedded AI represent a significant area of opportunity for either disrupting or partnering with leading chip firms to create specialized products, as different requirements and specifications arise, according to Chris Kachris, founder and CEO of InAccel
Kachris added there are still chances for upstarts to find innovation in areas too small for larger companies
In-memory computing, which combines storage and processing, and neuromorphic computing, which enables an AI system to adapt to new situations, have been put forward as areas to watch
Kachris also noted that although FPGAs may not be the most powerful solution for AI training, they may be the most energy-efficient solution for inference
The software layer for AI solutions is also important in terms of driving adoption: Nvidia's CUDA stack provides an advantage to the leading provider on the software front, although AMD's ROCm software competes with it

Read Full Article

12 Likes

Marktechpost

433

YuE: An Open-Source Music Generation AI Model Family Capable of Creating Full-Length Songs with Coherent Vocals, Instrumental Harmony, and Multi-Genre Creativity

YuE is an open-source music generation AI model family capable of creating full-length songs with coherent vocals, instrumental harmony, and multi-genre creativity.
YuE models are designed to generate full-length songs lasting several minutes, incorporating lyrics, vocals, and instrumental accompaniment.
YuE employs advanced techniques such as dual-token modeling and Lyrics-Chain-of-Thoughts (Lyrics-CoT) to ensure harmonious and consistent song generation.
YuE has the potential to revolutionize AI-driven music production, aiding musicians, composers, and content creators in generating full-length compositions across various styles and languages.

Read Full Article

25 Likes

Marktechpost

239

NVIDIA AI Releases Eagle2 Series Vision-Language Model: Achieving SOTA Results Across Various Multimodal Benchmarks

NVIDIA AI introduces Eagle 2, a transparent vision-language model (VLM) framework.
Eagle 2 prioritizes openness, providing details on its data strategy and model training.
Key innovations in Eagle 2 include a refined data strategy, multi-phase training approach, and vision-centric architecture.
Eagle 2 performs well on multiple benchmarks, surpassing other models in accuracy and efficiency.

Read Full Article

14 Likes

Insider

Image Credit: Insider

Zuckerberg says it's too soon to tell what impact DeepSeek will have on AI spending

Mark Zuckerberg says it's too soon to gauge DeepSeek's impact on Meta's AI spending.
DeepSeek, a Chinese AI startup, says it can build powerful models at a fraction of US costs.
Zuckerberg calls for an open-source AI standard that is 'American.'
Zuckerberg believes major AI infrastructure will still be needed despite DeepSeek's advancements.

Read Full Article

2 Likes

Pymnts

363

Image Credit: Pymnts

PayPal-Backed Formance Raises $21 Million to Grow Open-Source Ledger

Open-source financial infrastructure provider Formance raised $21 million in a Series A funding round to grow its Ledger and other solutions.
The funding round was co-led by PayPal Ventures and Portage, and Formance aims to expand across the U.S. and Europe.
Formance's open-source, programmable Ledger simplifies the process of building financial infrastructure, providing detailed visibility into transactions.
The company's solution is used by large-scale SaaS platforms, FinTechs, financial institutions, and other global customers.

Read Full Article

21 Likes

Mjtsai

216

Google Open Sources PebbleOS

The source code that powered Pebble smartwatches is now available for download.
Google is open sourcing the Pebble OS to support volunteers maintaining functionality for Pebble watches.
The repository includes the entire OS with smartwatch features like notifications, media controls, fitness tracking, and support for custom apps.
Pebble, known for its lightweight system and long battery life, was acquired by Fitbit and subsequently by Google.

Read Full Article

13 Likes

Siliconangle

409

Image Credit: Siliconangle

OpenAI finds DeepSeek used its data to train R1 reasoning model

OpenAI believes its data was used to train DeepSeek’s R1 large language model.
DeepSeek, a Chinese AI provider, developed R1 using large amounts of training data.
OpenAI suspects that some of its data was used in breach of its terms of service.
Microsoft identified potential data misuse by individuals linked to DeepSeek and informed OpenAI.

Read Full Article

24 Likes

Moneyweb

179

DeepSeek AI arrived over the weekend out of China

DeepSeek AI, a new AI technology from China, arrives as open-source software that offers efficient inference due to its approach to software despite being allegedly trained without H100s from Nvidia due to sanctions.
There is a significant improvement in efficiency by DeepSeek (90%), resulting in a direct impact in the inferences market along with predictions that the cost of inference market could be ten times less.
DeepSeek's technology advancement has reportedly changed people's perception of the chip market's demand.
Nvidia and ASML seem to have been under pressure from the new technology despite being key leaders.
As there is a significant improvement in DeepSeek's AI models, companies like Meta and Microsoft could benefit from using AI devices even more, which in turn could help in running businesses at a better profit margin.
Microsoft's long-term outlook of AI use means that there is a higher potentiality for Azure to be used to run smarter and more efficient programs.
DeepSeek's AI technology helps reduce deployment costs while increasing efficiency in performance.
Efficient models are attractive because computer hardware in data centres is expensive, and efficient models are cheaper to operate in the long term.
DeepSeek's approach is more inclined towards software than chips, which offers good performance, thereby changing the position of businesses.
The DeepSeek team includes about 200 people, illustrating that companies can invest in AI technology even with a small team; it is not necessary to have a large team.

Read Full Article

10 Likes

Marktechpost

354

Qwen AI Releases Qwen2.5-VL: A Powerful Vision-Language Model for Seamless Computer Interaction

Qwen AI has introduced Qwen2.5-VL, a vision-language model designed for seamless computer interaction.
Qwen2.5-VL offers improved visual understanding and reasoning capabilities and can recognize a wide range of objects and visual elements.
The model incorporates advancements in architecture, training methods, and video processing to enhance its performance.
Qwen2.5-VL demonstrates strong results in various tasks, making it a practical tool for real-world applications.

Read Full Article

21 Likes

Medium

299

Image Credit: Medium

DeepSeek: The Scrappy AI Revolutionizing the Game

DeepSeek R1 is an affordable open-source language model from China that rivals major players like OpenAI's GPT-4.
Built for under $6 million, DeepSeek disrupts the market and lowers the cost of entry for AI development.
DeepSeek uses distillation and has some limitations, but it paves the way for a more accessible and democratized AI landscape.
Its affordability raises questions about possible state-level funding and offers a glimpse into a future with lightweight and adaptable AI tools.

Read Full Article

17 Likes

VentureBeat

234

Image Credit: VentureBeat

Jack Dorsey is back with Goose, a new, ultra simple open source AI agent building platform from his startup Block

Jack Dorsey's Block has launched Goose, a free, open-source framework that helps build AI agents.
Goose allows developers to work with a range of large language models as backend intelligence.
Goose automates time-consuming coding tasks and can refine outputs and handle additional tasks with minimal oversight.
It is modular, allowing developers to integrate it into their systems and customise user interfaces.
Goose's focus is primarily on software development, but it can also work across multiple software vendors.
It features an open architecture, offering a range of large language models to choose from, including open-source options.
Block envisions broader applications such as streamlining music composition or enhancing e-commerce experiences.
Goose was built in collaboration with Anthropic, with the MCP tool enabling Goose to interface with different tools and systems.
The open-source nature of Goose fosters community-driven innovation, enabling new integrations to come from outside the company.
As a financial company, Block is interested in open-source models as it can bring models to where data is already hosted, reducing risk.

Read Full Article

14 Likes

TechCrunch

446

Image Credit: TechCrunch

Hugging Face researchers are trying to build a more open version of DeepSeek’s AI ‘reasoning’ model

Researchers at Hugging Face have launched Open-R1, a project that seeks to build a duplicate of R1 and open source all of its components, including the data used to train it.
Hugging Face is trying to replicate the model from scratch in what they’re calling a pursuit of “open knowledge”.
DeepSeek released the R1 model last week, a model that matches and even surpasses performance of OpenAI’s o1 reasoning model.
The Open-R1 project is less concerned about U.S. AI dominance than “fully opening the black box of model training”.
DeepSeek is reluctant to reveal the secret of their model, making full replication hard.
The goal of the Open-R1 project is to replicate R1 in a few weeks, relying in part on Hugging Face’s Science Cluster, a dedicated research server with 768 Nvidia H100 GPUs.
If the Open-R1 project is successful, AI researchers will be able to build on top of the training pipeline and work on developing the next generation of open-source reasoning models.
Open source development immediately benefits everyone, including the frontier labs and the model providers, as they can all use the same innovations.
There is a lot of interest already, with the Open-R1 project having racked up 10,000 GitHub stars in just three days.
Hugging Face is planning to tap its Science Cluster to generate data sets similar to those DeepSeek used to create R1.

Read Full Article

26 Likes

For uninterrupted reading, download the app

Open Source News

Discover more