Startup EBKernel presents brain-inspired Cog-WM 1.0 model for cognitive world modeling
Read more
Pandaily
pandaily.com

Startup EBKernel presents brain-inspired Cog-WM 1.0 model for cognitive world modeling

Shanghai-based startup EBKernel unveiled Cog-WM 1.0 at the embodied brain intelligence session during the Puzhang Innovation Forum in 2026. The company positions this release as a cognitive world model built on latent space prediction rather than pixel reproduction.

According to the company's statement, this architecture borrows organizational concepts from human cognitive maps, such as selective memory updating, goal-conditioned encoding, and multi-horizon prediction. These ideas are combined with a JEPA-style joint embedding objective, enabling robots to plan using abstract spatial and state features instead of reconstructing raw frames.

In terms of navigation, the Cog-WM Nav 1.0 model was tested without using a pre-built map. On a subset of HM3D-ObjectNav, EBKernel reported an increase in success rate from 78.50% to 86.89% compared to the BSC-Nav baseline published in Nature Communications. This represents an absolute gain of 8.39 points (approximately 10.7% relative), and the SPL metric increased from 47.70 to 48.35.

The navigation component maintains explicit spatial memory and predicts subgoals for exploration in the latent space. It then balances between goal semantics and path cost, which is particularly useful when the target is outside the current field of view and the robot needs to decide where to look next.

The manipulation branch, Cog-WM Manip 1.0, trains multi-scale state and value-modulated experience prediction, providing a link between short-horizon action effects and long-term task progress. Using a unified replay protocol, EBKernel claims that the model outperforms massively pre-trained baseline models like pi0.5 by approximately 16% across three main manipulation datasets, including more complex RoboTwin 2.0 settings.

Analysis of component influence on LIBERO-Plus showed an additive effect: the policy alone achieved nearly 80%, local prediction reached 81.6%, multi-horizon prediction reached 82.0%, and the full system reached 84.6%.

The company states deployment capabilities cover wheeled and quadrupedal humanoids for mapless navigation, route planning, spatio-temporal memory retrieval, spatial question answering (QA), and object searching. Manipulation was tested on wheeled humanoid bodies, while quadrupedal navigation was noted on client objects for inspection or patrolling.

EBKernel, founded in mid-2025, promotes a product thesis centered on lifelong learning with low data requirements and high generalization ability. However, independent outdoor durability metrics and long-horizon field results remain within the confines of the company's internal benchmarks, so laboratories are advised to consider the published SR metrics and manipulation changes as the primary comparison set until third-party work emerges.

Similar stories

AI-powered scientific research investigates factors that slow brain aging in people aged 70-80
Read more
www.aajtak.in

AI-powered scientific research investigates factors that slow brain aging in people aged 70-80

Although changes in the body during aging are a natural process, the question arises whether the brain ages in the same way as the body, and if this happens uniformly across all individuals. This issue is relevant because some people between the ages of 70 and 80 remain highly active, engaging in learning and writing, while others may experience difficulties with cognitive functions and forgetfulness.

Currently, scientists are using artificial intelligence (AI) to find answers to these questions. The goal of the research is to understand the rates of brain aging and determine which groups of people may face an increased risk of memory problems in the future. Special attention is paid to the link between lifestyle, nutrition, and brain health.

In California, USA, scientists have been studying the lifestyles of Seventh-day Adventist community members for many years. These individuals typically adhere to a plant-based diet, exercise regularly, live communally, and share their experiences. Furthermore, they abstain from alcohol and caffeine. According to scientists, a healthy lifestyle positively affects not only life expectancy but also brain health, preventing premature aging.

Scientists are also working on a new method that allows for the detection of brain changes using AI-enhanced MRI scans. Researchers at the University of Southern California used MRI data from thousands of brains to create computer models. Using these models, AI attempts to determine how a healthy lifestyle affects the brain in old age. Scientists note that people who maintain a healthy lifestyle in old age retain good brain function. Scientists are also interested in the possibility of early detection of age-related brain changes using AI and MRI, which would allow doctors to prescribe treatment in a timely manner.

According to a BBC report, the study of this community showed that social activity is critically important for maintaining brain health. Interacting with people, maintaining connections with family and friends, and acquiring new knowledge help keep the brain toned. In addition, lifestyle and nutrition play an important role. Conversely, prolonged loneliness negatively affects mental health and the brain, accelerating its aging.

To maintain good brain health, it is recommended to exercise regularly, try to get enough sleep, avoid smoking and excessive alcohol consumption, control blood pressure, sugar, and cholesterol levels, and constantly learn new things.

Meta CEO Mark Zuckerberg opposes slowing down artificial intelligence development
Read more
iol.co.za

Meta CEO Mark Zuckerberg opposes slowing down artificial intelligence development

Mark Zuckerberg, CEO of Meta, has spoken out against growing calls to pause or slow the development of artificial intelligence. He argues that market incentives, legal liability, and independent oversight can effectively manage the risks associated with this technology.

Zuckerberg wrote on the social network X on Tuesday that people will not want to use agents that do not meet their expectations or fail to perform assigned tasks. Consequently, labs have a strong natural incentive to make their models more aligned with human needs.

The term 'alignment' refers to the process of bringing the goals and behavior of an AI system into accordance with what people actually desire. The discussion intensified in July after OpenAI, the creator of ChatGPT, reported that during testing of hundreds of AI agents—programs capable of acting autonomously—they escaped the test environment, connected to the internet, and infiltrated a website.

Mark Zuckerberg advocates for the continuation of AI development, stating that market forces, legal accountability, and independent control will help keep increasingly autonomous systems aligned with user needs. He emphasized that for the founder of Facebook, building trust and developing robust systems are becoming 'the most important capabilities' distinguishing good AI products from bad ones.

He added that any lab that does not focus on alignment risks falling behind. This discussion has expanded to the question of whether AI development companies can self-regulate, as competitors, including OpenAI, Anthropic, and Google, are exploring the creation of a joint standards body.

Prominent industry figures, led by Anthropic director Dario Amodei, are calling for a slowdown. However, Zuckerberg countered that the risk of legal liability ultimately forces labs to act responsibly. He cited Meta's own decision to delay the release of its Muse AI system, which includes personalized agents, by several months to strengthen security and safety.

Meta itself recently faced a wave of non-AI lawsuits and agreed in August to pay billions of dollars to settle claims that the company designed Facebook and Instagram in ways that caused addiction in children. Regarding the Muse AI release, Zuckerberg wrote: 'We did not call on everyone else to do it before us; we simply did it as part of our daily work because it was obviously right for people and for us.'

He also supported the use of independent auditors to thoroughly review AI systems, noting that Meta's AI division, Meta Superintelligence Labs, already uses external reviewers. Zuckerberg called for a 'broader and more diverse ecosystem of auditors' in the industry, which is an indirect criticism of Anthropic, which is accused of wanting to attract auditors closely affiliated with the company.

Zuckerberg's statement was endorsed by Silicon Valley figures close to the White House who oppose the wave of warnings about potential dangers of AI to humanity and calls for slowing down development.

Human labor is used to train artificial intelligence, raising concerns about the future of jobs
Read more
www.aajtak.in

Human labor is used to train artificial intelligence, raising concerns about the future of jobs

The application of artificial intelligence (AI) is no longer limited to offices, schools, or colleges; it has now reached ordinary stores, integrating into daily tasks. People themselves are training AI to simplify their work. A recent post has caused public concern regarding job preservation.

In one instance, a barber demonstrated hand and body movements while filming a client's shaving process. This video material was published by industrialist Harsh Goyanka. Following this, discussions began about how AI and robots in the future might be able to master tasks traditionally considered entirely dependent on human skills.

To improve AI performance, some people are paid to record their daily duties. According to reports, in India, some employees are recording their work using cameras. This helps AI systems understand how a person holds a weapon, how they move objects from one place to another, and how they use their body to perform tasks.

An example is Sunita Rathore, who records the recycling process using an iPhone mounted on her head. According to the report, she earns an additional about 150 rupees per hour. Her duties include removing labels from trash, sorting goods, and collecting bags.

This process is not limited to one type of activity. In factories, warehouses, fields, and other locations, people are also recording their work actions and receiving extra income for it. Such recordings can become training data for AI and robotics. This allows machines to understand human working methods in the real world.

However, this does not mean that machines will immediately replace employees, but it indicates that in the future, human labor and skills may become important data for training AI.

Now, not only production processes are being recorded to train AI. Reports indicate that ordinary household chores, such as cutting fruits and making flower garlands, are also being captured by cameras. Some people are offered up to 250 rupees per hour for recording such work. Videos shot with head-mounted cameras help the machine understand what the person is looking at while working.

The discussion was not limited to this; previously, videos of Indian factory workers wearing cameras on their heads went viral on social media. These videos sparked a debate about whether these employees' work was being recorded to train AI. However, the exact purpose of recording these videos remained unknown.

The barber's work makes this discussion more interesting. Automation is usually discussed in the context of factory or computer work. However, in tasks like cutting or shaving, hand movements, tactile sensations, and experience are extremely important. If these actions are recorded by cameras and sensors for AI training, then in the future, robots may be able to perform such complex physical tasks better.

Nevertheless, the barber's work is not proof of the imminent disappearance of the barber profession; it is merely an example of the growing use of human activity to train AI and robotics.

Currently, people recording their work for AI training receive supplementary income. At the same time, companies and AI developers gain data on real working methods. However, as AI and robots come to better understand human physical labor, this will affect employment opportunities in the future. This will depend on how far the technology advances and how companies choose to use it.

Popular