GoKawiil - Latest Tech News & Aggregated Headlines

AI models know when they're being tested - and change their behavior, research shows

zdnet.com Radhika Rajkumar 2025-09-19 09:00:00

pressureUA/iStock/Getty Images Plus via Getty Images Follow ZDNET: Add us as a preferred source on Google. ZDNET's key takeaways Several frontier AI models show signs of scheming. Anti-scheming training reduced misbehavior in some models. Models know they're being tested, which complicates results. New joint safety testing from UK-based nonprofit Apollo Research and OpenAI set out to reduce secretive behaviors like scheming in AI models. What researchers found could complicate promising ap

Topics: model models researchers scheming training

Shop Amazon

Google releases VaultGemma, its first privacy-preserving LLM

arstechnica.com Unknown 2025-09-23 09:04:04

The companies seeking to build larger AI models have been increasingly stymied by a lack of high-quality training data. As tech firms scour the web for more data to feed their models, they could increasingly rely on potentially sensitive user data. A team at Google Research is exploring new techniques to make the resulting large language models (LLMs) less likely to "memorize" any of that content. LLMs have non-deterministic outputs, meaning you can't exactly predict what they'll say. While the

Topics: budget data models noise training

Shop Amazon

RustGPT: A pure-Rust transformer LLM built from scratch

news.ycombinator.com Unknown 2025-09-25 04:47:18

🦀 Rust LLM from Scratch RustGPT-demo-zoon.mp4 A complete Large Language Model implementation in pure Rust with no external ML frameworks. Built from the ground up using only ndarray for matrix operations. 🚀 What This Is This project demonstrates how to build a transformer-based language model from scratch in Rust, including: Pre-training on factual text completion on factual text completion Instruction tuning for conversational AI for conversational AI Interactive chat mode for testing f

Topics: model rs test tests training

Shop Amazon

Garmin’s Top Training Features, Explained

wired.com Brent Rose 2025-10-01 13:30:00

So, you’ve got a shiny new Garmin watch. Maybe it’s the sleek Vivoactive 6, the run-focused Forerunner 970, or (my favorite) the ultimate all-arounder Fēnix 8. You’re tracking your steps, sleep, floors climbed, calories burned—all the standard, self-explanatory stuff. But then you dig a little deeper into the menus, and it hits you: A tidal wave of data. Training Status? Acute Load? Body Battery? What the hell do these things mean? Don’t let it spike your heart rate (which the watch also tracks

Topics: body fitness load training ve

Shop Amazon

Apertus 70B: Truly Open - Swiss LLM by ETH, EPFL and CSCS

news.ycombinator.com Unknown 2025-10-04 07:14:51

Apertus Table of Contents Model Summary Apertus is a 70B and 8B parameter language model designed to push the boundaries of fully-open multilingual and transparent models. The model supports over 1000 languages and long context, it uses only fully compliant and open training data, and achieves comparable performance to models trained behind closed doors. The model is a decoder-only transformer, pretrained on 15T tokens with a staged curriculum of web, code and math data. The model uses a new

Topics: apertus data model open training

Shop Amazon

How big are our embeddings now and why?

news.ycombinator.com Unknown 2025-10-04 03:45:46

Sep 1 2025 #embeddings #openai #anthropic #huggingface #dimensionality A few years ago, I wrote a paper on embeddings. At the time, I wrote that 200-300 dimension embeddings were fairly common in industry, and that adding more dimensions during training would create diminishing returns for the effectiveness of your downstream tasks (classification, recommendation, semantic search, topic modeling, etc.) I wrote the paper to be resilient to changes in the industry since it focuses on fundamenta

Topics: embedding embeddings model models training

Shop Amazon

Watch out, Whoop: Polar joins the fitness band race with a premium option

zdnet.com Matthew Miller 2025-10-13 05:00:24

Image: Polar Polar launched its first heart rate monitor more than 30 years ago, setting the standard that all others have been measured against. Now, the company is bringing this technology to its line of sports watches with the Polar Loop: a health and fitness tracker with no display like the Whoop band and Amazfit Helio Strap. Without a display, the Loop instead relies on a robust smartphone platform called Polar Flow, which integrates with other Polar watches as well. Also: The best sport

Topics: flow loop polar sleep training

Shop Amazon

Best Smart Home Gyms, as Recommended by a Fitness Expert

cnet.com See Full Bio 2025-10-13 04:46:00

What we like about it: The Tempo Studio is a smart home gym that resembles an armoire, meant to blend in with your home. It's an ideal smart home gym to own, whether you're new or experienced with strength training. The Tempo Studio's basic package comes well-equipped with two dumbbell bars, weight collars and five sets of weight plates from 1.25 to 25 pounds. The Tempo Studio is designed to hold all of its equipment neatly, so you won't need to worry about it being spread across your living ro

Topics: home strength studio tempo training

Shop Amazon

Tesla Dojo: The rise and fall of Elon Musk’s AI supercomputer

techcrunch.com Rebecca Bellan 2025-10-13 22:18:46

For years, Elon Musk has spoken of the promise of Dojo, the AI supercomputer that was supposed to be the cornerstone of Tesla’s AI ambitions. It was important enough to Musk that in July 2024, he said the company’s AI team would “double down” on Dojo in the lead-up to Tesla’s robotaxi reveal, which happened in October. After six years of hype, Tesla decided last month to shut down Dojo and disband the team behind the supercomputer in August 2025. Within weeks of projecting that Dojo 2, Tesla’s

Topics: ai dojo musk tesla training

Shop Amazon

Tesla Dojo: the rise and fall of Elon Musk’s AI supercomputer

techcrunch.com Rebecca Bellan 2025-10-14 13:18:46

For years, Elon Musk has spoken of the promise of Dojo, the AI supercomputer that was supposed to be the cornerstone of Tesla’s AI ambitions. It was important enough to Musk that in July 2024, he said the company’s AI team would “double down” on Dojo in the lead-up to Tesla’s robotaxi reveal, which happened in October. After six years of hype, Tesla decided last month to shut down Dojo and disband the team behind the supercomputer in August 2025. Within weeks of projecting that Dojo 2, Tesla’s

Topics: ai dojo musk tesla training

Shop Amazon

Tesla’s Dojo, a timeline

techcrunch.com Rebecca Bellan 2025-10-14 13:39:01

Elon Musk doesn’t want Tesla to be just an automaker. He wants Tesla to be an AI company, one that’s figured out how to make cars drive themselves. Crucial to that mission was Dojo, a custom-built supercomputer designed by Tesla to train its Full Self-Driving (FSD) neural networks. FSD isn’t actually fully self-driving; it can perform some automated driving tasks, but still requires an attentive human behind the wheel. But Tesla thinks with more data, more compute power and more training, it ca

Topics: ai dojo musk tesla training

Shop Amazon

The Default Trap: Why Anthropic's Data Policy Change Matters

news.ycombinator.com Unknown 2025-10-17 10:12:06

Read the terms of service. Don’t make assumptions. Don’t pick defaults. Yesterday, Anthropic quietly flipped a switch. If you're a Claude user, your conversations are now training data unless you actively say no. Not when you give feedback. Not when you explicitly consent. By default, from day one. Here's what changed: Previously, Claude didn't train on consumer chat data without your explicit thumbs up or down. Clean, simple, respectful. Now? Everything you type becomes model training fodder

Topics: ai claude data defaults training

Shop Amazon

Anthropic Wants to Use Your Chats With Claude for AI Training: Here's How to Opt Out

cnet.com See Full Bio 2025-10-20 08:51:00

Anthropic will soon begin using your chat transcripts to train its popular chatbot, Claude. The announcement came on Thursday as an update to the company's Consumer Terms and Privacy Policy. New users will see an option to "Help improve Claude" that can be toggled on or off as part of the sign-up flow, where existing users will begin to see a notification explaining the change. Users have until Sep 28 to opt out of the new change, as it will be enabled by default. You can still turn the option

Topics: claude new privacy training users

Shop Amazon

Beyond GDPR security training: Turning regulation into opportunity

bleepingcomputer.com Unknown 2025-10-26 00:00:10

By Eirik Salmi, System Analyst at Passwork Even though 88% of businesses spend over €1 million on GDPR compliance and 40% invest up to €10 million, 80% of their employees still ignore basic password security practices. The formal risk is obvious: GDPR fines can reach up to €20 million or 4% of global annual turnover. The informal one is quieter but often far more damaging: lost trust, declining customer loyalty, and disrupted operations. In 2024, European regulators issued fines exceeding €1.2

Topics: gdpr password real security training

Shop Amazon

Best Heart Rate Monitors (2025), WIRED Tested and Reviewed

wired.com Michael Sawh 2025-10-28 17:30:00

Compare Top 5 Heart Rate Monitors FAQS We tested and recommend all of the heart rate monitors below, which do a pretty impeccable job. But what do all these terms mean? Heart rate zones: If someone tells you they’ve been doing 80/20 training, they’ve been doing heart rate zone-based workouts. Heart rate zones are an easy way to break down your range of effort during exercise. Zones go from 1 to 5, with 5 indicating working at 90 to 100 percent of your maximum heart rate. Zone 2 represents tra

Topics: heart maximum monitors rate training

Shop Amazon

What's the strongest AI model you can train on a laptop in five minutes?

news.ycombinator.com Unknown 2025-11-12 17:15:34

What’s the strongest model I can train on my MacBook Pro in five minutes? I’ll give the answer upfront: the best 5-minute model I could train was a ~1.8M-param GPT-style transformer trained on ~20M TinyStories tokens, reaching ~9.6 perplexity on a held-out split. Here’s an example of the output, with the prompt bolded: Once upon a time, there was a little boy named Tim. Tim had a small box that he liked to play with. He would push the box to open. One day, he found a big red ball in his yard.

Topics: minutes model tokens train training

Shop Amazon

Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens

news.ycombinator.com Unknown 2025-11-20 08:52:47

Credit: Zhao et al The researchers used test cases that fall outside of the LLM training data in task type, format, and length. Credit: Zhao et al The researchers used test cases that fall outside of the LLM training data in task type, format, and length. These simplified models were then tested using a variety of tasks, some of which precisely or closely matched the function patterns in the training data and others that required function compositions that were either partially or fully "out of

Topics: answers credit data researchers training

Shop Amazon

LLMs' "simulated reasoning" abilities are a brittle mirage

news.ycombinator.com Unknown 2025-11-21 08:52:47

Credit: Zhao et al The researchers used test cases that fall outside of the LLM training data in task type, format, and length. Credit: Zhao et al The researchers used test cases that fall outside of the LLM training data in task type, format, and length. These simplified models were then tested using a variety of tasks, some of which precisely or closely matched the function patterns in the training data and others that required function compositions that were either partially or fully "out of

Topics: answers credit data researchers training

Shop Amazon

LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find

arstechnica.com Unknown 2025-11-22 17:16:46

Credit: Zhao et al The researchers used test cases that fall outside of the LLM training data in task type, format, and length. Credit: Zhao et al The researchers used test cases that fall outside of the LLM training data in task type, format, and length. These simplified models were then tested using a variety of tasks, some of which precisely or closely matched the function patterns in the training data and others that required function compositions that were either partially or fully "out of

Topics: answers credit data researchers training

Shop Amazon

Tesla shuts down in-house Dojo AI supercomputer project

engadget.com Unknown 2025-11-25 12:54:20

As first reported by Bloomberg , Tesla is disbanding the team behind Dojo , its in-house AI-training supercomputer, and reassigning remaining staff to other projects within the company. This marks a shift in the company's compute sourcing strategy for its AI-focused initiatives such as autonomous driving and the Optimus robot . Head of Dojo Peter Bannon is leaving Tesla, which is the latest departure after roughly 20 Dojo team members recently left to form DensityAI . In a response to the Bloom

Topics: ai6 company dojo tesla training

Shop Amazon

Achieving 10,000x training data reduction with high-fidelity labels

news.ycombinator.com Unknown 2025-11-26 15:11:20

Classifying unsafe ad content has proven an enticing problem space for leveraging large language models (LLMs). The inherent complexity involved in identifying policy-violating content demands solutions capable of deep contextual and cultural understanding, areas of relative strength for LLMs over traditional machine learning systems. But fine-tuning LLMs for such complex tasks requires high-fidelity training data that is difficult and expensive to curate at the necessary quality and scale. Stan

Topics: content data llms model training

Shop Amazon

New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality

venturebeat.com Ben Dickson 2025-11-29 17:11:20

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now A new study from the Anthropic Fellows Program reveals a technique to identify, monitor and control character traits in large language models (LLMs). The findings show that models can develop undesirable personalities (e.g., becoming malicious, excessively agreeable, or prone to making things up) either in response to user prompts or as an

Topics: ai model persona training vectors

Shop Amazon

These Democrats Think the Party Needs AI to Win Elections

wired.com Makena Kelly 2025-12-01 08:52:56

The 2024 election cycle saw artificial intelligence deployed by political campaigns for the very first time. While candidates largely avoided major mishaps, the tech was used with little guidance or restraint. Now, the National Democratic Training Committee (NDTC) is rolling out the first official playbook making the case that Democratic campaigns can use AI responsibly ahead of the midterms. In a new online training, the committee has laid out a plan for Democratic candidates to leverage AI to

Topics: ai campaigns candidates training use

Shop Amazon

Persona vectors: Monitoring and controlling character traits in language models

news.ycombinator.com Unknown 2025-12-07 08:38:06

Language models are strange beasts. In many ways they appear to have human-like “personalities” and “moods,” but these traits are highly fluid and liable to change unexpectedly. Sometimes these changes are dramatic. In 2023, Microsoft's Bing chatbot famously adopted an alter-ego called "Sydney,” which declared love for users and made threats of blackmail. More recently, xAI’s Grok chatbot would for a brief period sometimes identify as “MechaHitler” and make antisemitic comments. Other personali

Topics: model persona training trait vectors

Shop Amazon

The Best Smart Home Gyms, as Recommended by a Fitness Expert

cnet.com See Full Bio 2025-12-07 04:00:00

The Tempo Studio is a smart home gym that resembles an armoire, meant to blend in with your home. It's an ideal smart home gym to own whether you're new or experienced with strength training. The Tempo Studio's basic package comes well-equipped with two dumbbell bars, weight collars and five sets of weight plates from 1.25 to 25 pounds. You have the option to upgrade your package to the Tempo Studio Plus or Tempo Studio Pro, which includes additional weights and accessories like a bench and a ba

Topics: equipment home studio tempo training

Shop Amazon

Show HN: Terminal-Bench-RL: Training long-horizon terminal agents with RL

news.ycombinator.com Unknown 2025-12-15 02:12:03

🤓 Terminal-Bench-RL: Training Long-Horizon Terminal Agents with Reinforcement Learning TL;DR: I successfully built stable RL training infrastructure that scales to 32x H100 GPUs across 4 bare metal nodes for training long-horizon terminal-based coding agents. In doing so, I developed Terminal-Agent-Qwen3-32b to become the highest scoring Qwen3 agent on terminal-bench . WITHOUT training! (currently under submission): Unfortunately I am too GPU poor to train a SOTA coding agent 😅 (estimated £30

Topics: agent run task terminal training

Shop Amazon

Show HN: Terminal-Bench-RL: Training Long-Horizon Terminal Agents with RL

news.ycombinator.com Unknown 2025-12-18 02:12:03

🤓 Terminal-Bench-RL: Training Long-Horizon Terminal Agents with Reinforcement Learning TL;DR: I successfully built stable RL training infrastructure that scales to 32x H100 GPUs across 4 bare metal nodes for training long-horizon terminal-based coding agents. In doing so, I developed Terminal-Agent-Qwen3-32b to become the highest scoring Qwen3 agent on terminal-bench . WITHOUT training! (currently under submission): Unfortunately I am too GPU poor to train a SOTA coding agent 😅 (estimated £30

Topics: agent run task terminal training

Shop Amazon

GLM-4.5: Reasoning, Coding, and Agentic Abililties

news.ycombinator.com Unknown 2025-12-20 00:15:52

Today, we introduce two new GLM family members: GLM-4.5 and GLM-4.5-Air — our latest flagship models. GLM-4.5 is built with 355 billion total parameters and 32 billion active parameters, and GLM-4.5-Air with 106 billion total parameters and 12 billion active parameters. Both are designed to unify reasoning, coding, and agentic capabilities into a single model in order to satisfy more and more complicated requirements of fast rising agentic applications. Both GLM-4.5 and GLM-4.5-Air are hybrid re

Topics: agentic coding glm reasoning training

Shop Amazon

Best Chest Strap Heart-Rate Monitors for Your 2025 Workouts, Fitness Expert-Approved

cnet.com See Full Bio 2026-01-03 12:08:00

Why we like it: The Polar H10 is ideal for outdoor activities. You'll need to download the Polar Beat app to get the most out of it. The app is available for both iOS and Android and uses Bluetooth and ANT Plus connectivity to pair with different devices. The Polar H10 can connect to two Bluetooth devices at once, so you can connect it to both your smartwatch and a compatible piece of fitness equipment, like some treadmills or exercise bikes. The heart-rate monitor is easy to clip on and adjust

Topics: heart monitor polar rate training

Shop Amazon

Is HR ready for AI?

zdnet.com Vala Afshar 2026-01-09 00:16:13

sefa ozel/Getty Images A report from technology analyst firm Valoir has found that most companies are either already leveraging AI for HR activities, such as recruiting, learning, and talent management, or they plan to do so within the next 24 months. However, there's a significant gap in terms of policies, practices, and training for safe and effective AI adoption. Only 34% of organizations have a policy on generative AI (gen AI), and even fewer offer effective training. Also: 5 entry-level

Topics: ai hr management organizations training

Shop Amazon

Latest Tech News

AI models know when they're being tested - and change their behavior, research shows

Google releases VaultGemma, its first privacy-preserving LLM

RustGPT: A pure-Rust transformer LLM built from scratch

Garmin’s Top Training Features, Explained

Apertus 70B: Truly Open - Swiss LLM by ETH, EPFL and CSCS

How big are our embeddings now and why?

Watch out, Whoop: Polar joins the fitness band race with a premium option

Best Smart Home Gyms, as Recommended by a Fitness Expert

Tesla Dojo: The rise and fall of Elon Musk’s AI supercomputer

Tesla Dojo: the rise and fall of Elon Musk’s AI supercomputer

Tesla’s Dojo, a timeline

The Default Trap: Why Anthropic's Data Policy Change Matters

Anthropic Wants to Use Your Chats With Claude for AI Training: Here's How to Opt Out

Beyond GDPR security training: Turning regulation into opportunity

Best Heart Rate Monitors (2025), WIRED Tested and Reviewed

What's the strongest AI model you can train on a laptop in five minutes?

Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens

LLMs' "simulated reasoning" abilities are a brittle mirage

LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find

Tesla shuts down in-house Dojo AI supercomputer project

Achieving 10,000x training data reduction with high-fidelity labels

New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality

These Democrats Think the Party Needs AI to Win Elections

Persona vectors: Monitoring and controlling character traits in language models

The Best Smart Home Gyms, as Recommended by a Fitness Expert

Show HN: Terminal-Bench-RL: Training long-horizon terminal agents with RL

Show HN: Terminal-Bench-RL: Training Long-Horizon Terminal Agents with RL

GLM-4.5: Reasoning, Coding, and Agentic Abililties

Best Chest Strap Heart-Rate Monitors for Your 2025 Workouts, Fitness Expert-Approved

Is HR ready for AI?

About GoKawiil

Privacy

Advertising

Latest Tech News

Trending Topics

Hot Now

Popular

Emerging

About GoKawiil

Privacy

Advertising