~hackernews | Bookmarks (11818)

DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss

hirundo.io

TL;DRBias in large language models (LLMs) is a growing concern, particularly in sensitive customer-facing industries where...
TL;DRBias in large language models (LLMs) is a growing concern, particularly in sensitive customer-facing industries where fairness and compliance are critical. With recent buzz around DeepSeek, we took the opportunity to showcase Hirundo’s bias unlearning capabilities on DeepSeek-R1-Distill-Llama-8B. Our results de...
1
Ötzi the Lyme Mummy (2019)

illinois.edu

Brrrr! It’s a chilly day here in Illinois with all the wind. I’m not complaining however....
Brrrr! It’s a chilly day here in Illinois with all the wind. I’m not complaining however. Can you imagine being frozen in a glacier for thousands of years and suffer from Lyme disease?! Whew, I can’t image that ever actually happening to anyone. Doctors examining the body of Ötzi, the Iceman with Lyme disease bacter...
1
Adding iodine to salt played a role in cognitive improvements: research (2013)

discovermagazine.com

Iodized salt is so commonplace in the U.S. today that you may never have given the...
Iodized salt is so commonplace in the U.S. today that you may never have given the additive a second thought. But new research finds that humble iodine has played a substantial role in cognitive improvements seen across the American population in the 20th century.Iodine is a critical micronutrient in the human diet ...
1
The widely cited studies on mouse vs. keyboard efficiency are completely bogus

danluu.com
1
Effective AI code suggestions: less is more

qodo.ai
1
Case Study: ByteDance Uses eBPF to Enhance Networking Performance

ebpf.foundation

Download this case study in PDF format Overview Bytedance, a global technology company operating a wide...
Download this case study in PDF format Overview Bytedance, a global technology company operating a wide range of content platforms around the world at massive scale, faced significant challenges in ensuring performance and stability across its data centers. With over a million servers running containerized applicati...
1
Frigade (YC W23) is hiring engineer #2

ycombinator.com

Growth platform for modern software teams$130K - $180K / 0.50% - 1.50%San Francisco, CA, USConnect directly...
Growth platform for modern software teams$130K - $180K / 0.50% - 1.50%San Francisco, CA, USConnect directly with founders of the best YC-funded startups.Apply to role ›We’re looking for an experienced software engineer who embraces the breadth of responsibilities that come with a senior full-stack role. You’ll colle...
1
How to run DeepSeek R1 locally

workos.com

DeepSeek R1 is an open-source LLM for conversational AI, coding, and problem-solving. Here's how to run...
DeepSeek R1 is an open-source LLM for conversational AI, coding, and problem-solving. Here's how to run it locally.People who want full control over data, security, and performance run LLMs locally.DeepSeek R1 is an open-source LLM for conversational AI, coding, and problem-solving that recently outperformed OpenAI’...
1
An Analysis of DeepSeek's R1-Zero and R1

arcprize.org

Special thanks to Tuhin and Abu from Baseten and Yuchen from Hyperbolic Labs for hosting r1-zero...
Special thanks to Tuhin and Abu from Baseten and Yuchen from Hyperbolic Labs for hosting r1-zero for us. Hardly any providers are hosting this model variant, and its availability is important for research purposes. ARC Prize Foundation’s goal is to define, measure, and inspire new ideas towards AGI. To this end, we ...
1
OpenAI says it has evidence China's DeepSeek used its model to train competitor

ft.com
1
Society for Technical Communication to permanently close its doors

stc.org

Dear Members, Stakeholders, and Friends, Despite significant efforts over the past several years by the various...
Dear Members, Stakeholders, and Friends, Despite significant efforts over the past several years by the various Boards of Directors and STC leadership, financial liabilities coupled with falling membership numbers have created a situation where we are not able to offer the education, resources, and outreach that thi...
1
US children fall further behind in reading

cnn.com

America’s children have continued to lose ground on reading skills in the wake of the COVID-19...
America’s children have continued to lose ground on reading skills in the wake of the COVID-19 pandemic and have made little improvement in math, according to the latest results of an exam known as the nation’s report card. The findings are yet another setback for U.S. schools and reflect the myriad challenges that ...

hnk
1
Comcast is rolling out 'ultra-low lag' tech that could fix the internet

theverge.com

If you use Comcast Xfinity internet, your FaceTime calls might be about to get better. Instead...
If you use Comcast Xfinity internet, your FaceTime calls might be about to get better. Instead of bumping up the amount of data that your internet connection can send or receive at one time (usually called bandwidth or throughput), a new upgrade is coming to reduce the amount of time it takes for each packet of info...
1
Only 1 in 10 Oracle Java users want to stay with Big Red

theregister.com

Only around one in ten Oracle Java customers are likely to stick around following costly licensing...
Only around one in ten Oracle Java customers are likely to stick around following costly licensing changes Big Red made to its development and runtime environments in January 2023, according to research. A report from Dimensional Research found the percentage of Oracle Java users considering switching to alternative...
1
Show HN: DeepSeek vs. ChatGPT – The Clash of the AI Generations

sigmabrowser.com

AI Сhatbots are the trend of the last few years. They help us write, research, and...
AI Сhatbots are the trend of the last few years. They help us write, research, and automate everyday tasks. But when it comes to choosing the best of them, recent debate has centered on DeepSeek and ChatGPT.Both are powerful artificial intelligence models, but they are designed for different needs. So which one is t...
1
On DeepSeek and Export Controls

darioamodei.com
1
X partners with Visa to enable money transfers on the platform

washingtonpost.com
1
Meta's Mark Zuckerberg explores purchase of Washington DC property

ft.com
1
The ancient discovery that put a Silk Road city back on the map

bbc.com

Matt Stirn(Credit: Matt Stirn)After unearthing a rare collection of highly prized Silk Road artefacts, a family...
Matt Stirn(Credit: Matt Stirn)After unearthing a rare collection of highly prized Silk Road artefacts, a family in the Himalayas has opened a museum dedicated to one of the fabled route's final traders.Muzzamil Hussain was in grade school when the first bombs fell on the playground outside of his classroom in Kargil...
1
Google Pixel 4a old firmware is gone, trapping users on the buggy battery update

androidcentral.com

(Image credit: Android Central) Google’s Pixel 4a battery update is a disaster, causing crazy battery drain...
(Image credit: Android Central) Google’s Pixel 4a battery update is a disaster, causing crazy battery drain for many users.The update also wiped older firmware, leaving no way to roll back to the previous version.Google has acknowledged the mess and offered compensation: a free battery replacement, $50 cash, or $100...
1
Italy's privacy regulator goes after DeepSeek

politico.eu

algorithms artificial intelligence big data china data breaches
1
A story about restoring and upgrading a Commodore Amiga 1000

celso.io
2
DeepSeek proves the future of LLMs is open-source

getlago.com

By now, you’ve heard of DeepSeek. It’s the Chinese AI lab that trained R1, an open-source...
By now, you’ve heard of DeepSeek. It’s the Chinese AI lab that trained R1, an open-source reasoning model as good as OpenAI’s o1, but trained on inferior hardware for a fraction of the price. ‍They achieved this with novel training methods that are more efficient than the ones OpenAI, Anthropic and other well-funded...
1
The Debian Publicity Team will no longer post on X/Twitter

debian.org
1