Among those advocating a different direction is Yann LeCun, who said existing AI systems such as ChatGPT, Claude and Gemini are effective for tasks including coding, mathematics and text generation but are not designed to achieve human-like or even animal-level understanding of real-world environments.

Speaking during the VivaTech conference, LeCun said these systems primarily learn statistical relationships from vast datasets rather than developing an understanding of how the physical world behaves. He argued that this limitation makes them unsuitable for many robotics applications that require flexible reasoning and adaptation.

LeCun, who previously served as chief AI scientist at Meta before leaving in 2025, now leads Paris-based Advanced Machine Intelligence Labs (AMI Labs), where researchers are developing an alternative AI architecture known as Joint Embedding Predictive Architecture (JEPA).

Earlier this year, AMI Labs announced it had raised more than $1 billion in seed funding. The investment included backing from Nvidia and a fund managing the private wealth of Jeff Bezos, making it one of Europe’s largest seed funding rounds.

Why Researchers Want AI Beyond Language Models

LeCun argues that LLMs are highly capable within structured and predictable domains because they generate responses based on patterns learned during training. However, he said those systems do not possess an underlying model of physical reality.

To illustrate the difference, he described balancing a pen on its tip. While a person instinctively knows the pen will fall without needing to predict the exact direction, an LLM may attempt to generate a statistically likely outcome rather than reasoning about the uncertainty of the situation itself.

According to LeCun, JEPA seeks to address this challenge by creating abstract representations of the physical world. Rather than predicting every possible outcome, the system is intended to identify which information is meaningful while ignoring unnecessary detail, allowing it to reason more efficiently about real-world situations.

He said this capability could eventually make AI systems better suited for robotics, where machines must continually interpret changing environments instead of responding to fixed prompts.

Robotics Continues to Drive AI Research

Improving AI reasoning has become a significant objective for robotics developers, who have invested billions of dollars in humanoid machines capable of operating in human environments.

Although robotic hardware has advanced rapidly, teaching robots to perform everyday household activities such as loading dishwashers or ironing clothing safely remains technically difficult and expensive.

LeCun said current LLM-based approaches are unlikely to solve these challenges effectively because they are not built to interpret complex physical interactions.

World Models Gain Momentum Across the Industry

Other researchers share the view that future AI systems will require more sophisticated reasoning than today’s language models provide.

Ingmar Posner, who directs the Applied AI Lab at the University of Oxford and also serves as an Amazon Scholar, said future AI systems should be capable of explaining cause and effect, identifying what matters in a situation and evaluating alternative actions.

His research group has spent several years developing what he describes as a mechanistic world model, designed to organize knowledge so that information can be efficiently retrieved, combined and modified when solving problems.

World models have existed as a research concept for decades but received renewed attention following influential work published in 2018 by David Ha and Jürgen Schmidhuber. Their research proposed that AI systems could learn by building internal simulations of the world rather than relying solely on memorized patterns.

Since then, several organizations have expanded work in the field. Google has developed Dreamer world models, including a version that learned to collect diamonds in the video game Minecraft by imagining future scenarios during decision-making.

Additional research includes Genie from Google DeepMind, Gaia from Wayve, and work at World Labs, founded in 2023 by Fei-Fei Li.

Commercial Deployment Remains Ahead

Posner said it remains difficult to predict how quickly these newer AI architectures will mature, noting that the rapid arrival of generative AI systems surprised many researchers who had expected such capabilities to take decades longer.

LeCun said AMI Labs plans to continue refining its JEPA-based system through the remainder of this year, with the goal of introducing initial industrial deployments next year if development progresses as planned.

Looking further ahead, he said broader-purpose AI systems capable of performing many tasks with limited additional training remain the long-term objective. Even if those systems eventually exceed human capabilities in certain areas, LeCun argued that people will continue to play the central role in defining goals, asking questions and directing how AI is applied.

Tags: Artificial Intelligence, World Models, Yann LeCun, AMI Labs, Large Language Models, Robotics, Machine Learning, ChatGPT, Google DeepMind, Nvidia, JEPA, AI Research

The post AI Researchers Shift Focus Beyond Large Language Models Toward World Models for Real-World Intelligence appeared first on Journos News - Breaking News, World News, Top Stories, Todays Headlines and Flash Reports.

Despite its impressive performance, Nvidia’s stock remained flat in extended trading. However, the company’s guidance for Q1 2025 signals sustained momentum, projecting $43 billion in revenue—higher than analysts’ expectations.

Key Financial Highlights (Compared to Analyst Estimates)

Revenue: $39.33 billion (vs. $38.05 billion expected)
Earnings Per Share (EPS): $0.89 adjusted (vs. $0.84 expected)
Net Income: $22.09 billion (up from $12.29 billion a year ago)
Gross Margin: 73% (down slightly due to higher costs in new data center products)

Nvidia’s full-year revenue surged 114% to $130.5 billion, marking a historic rise driven by its dominance in AI accelerator chips.

AI Chip Business: The Backbone of Nvidia’s Growth

Data Center Revenue: $35.6 billion in Q4, a 93% annual increase, exceeding estimates of $33.65 billion.
AI Chip Sales (Blackwell & Hopper): Account for 91% of total sales, up from 83% a year ago.
Blackwell AI Chips: $11 billion in Q4 sales, with CEO Jensen Huang calling demand “amazing.”

Why It Matters:

• Nvidia’s Blackwell AI processors are experiencing the fastest product ramp in company history, with cloud giants like Amazon, Microsoft, and Google driving sales.
• AI inference (running AI applications) is expected to require exponentially more computing power, boosting future demand for Nvidia’s chips.

Challenges & Areas of Concern

Networking Sales: $3 billion in Q4, down 9% year-over-year, despite being a key growth focus.
Gaming Revenue: $2.5 billion, below the $3.04 billion estimate, marking an 11% annual decline.
Automotive & Robotics Chips: While still a small segment, revenue hit $570 million, a 103% increase year-over-year.

Nvidia also addressed concerns about custom AI chips being developed by tech giants like Amazon, Microsoft, and Google. CEO Huang reassured investors, saying, “Just because a chip is designed doesn’t mean it gets deployed.”

Looking Ahead: Can Nvidia Maintain Its Growth?

• Q1 2025 Forecast: $43 billion revenue, implying a 65% year-over-year growth—a slowdown from the 262% growth in Q1 2024.
• AI Demand Continues: Nvidia expects new AI models to require up to 100 times more processing power, fueling future chip sales.
• Stock Buybacks: Nvidia spent $33.7 billion on share repurchases in fiscal 2025.

Bottom Line: Nvidia remains the dominant force in AI hardware, but growth may slow as the company scales. The focus now shifts to how quickly it can deliver next-generation AI processors and expand into new markets like robotics and automotive AI.

Source

The post Nvidia’s Revenue Soars 78% Thanks to AI Demand appeared first on Journos News - Breaking News, World News, Top Stories, Todays Headlines and Flash Reports.

White House Concerns Over AI Distillation

Top White House advisers are raising alarms over China’s DeepSeek and its potential use of a controversial AI training method known as “distillation.“ This technique allows one AI system to learn from another, potentially giving DeepSeek an advantage by leveraging the advancements of U.S. rivals—without the massive costs and computing power investments.

Despite concerns, stopping this practice may prove challenging, according to Silicon Valley executives and investors.

DeepSeek’s Breakthrough Shakes Up the AI Industry

DeepSeek made headlines this month by unveiling an AI model that rivals top U.S. technologies, such as OpenAI’s ChatGPT, but at a significantly lower cost. Even more surprising, the China-based company released its model for free, sparking debate over how it achieved such rapid advancements.

Some experts suspect that DeepSeek may have used distillation to learn from U.S. models, allowing it to bypass the costly and time-consuming process of developing AI from scratch.

How AI Distillation Works

AI distillation involves training a newer model by having it interact with an older, more powerful AI system. The established model evaluates the quality of responses from the newer system, effectively transferring its knowledge.

This means that companies like DeepSeek could benefit from the extensive resources spent by U.S. firms—without directly accessing or copying proprietary data.

While AI distillation is widely used in the industry, it violates the terms of service of several major U.S. AI firms, including OpenAI.

OpenAI and U.S. Firms Investigating DeepSeek

A spokesperson for OpenAI confirmed that the company is aware of groups in China actively working to replicate U.S. AI models through distillation. OpenAI is now investigating whether DeepSeek improperly used this method to develop its latest model.

Industry Experts: Learning from Rivals is Common

Despite ethical and legal concerns, some industry leaders argue that learning from competitors is standard practice in AI development.

Naveen Rao, vice president of AI at San Francisco-based Databricks, compared AI distillation to automakers reverse-engineering each other’s engines to gain insights.

“To be completely fair, this happens in every industry. Competition is real, and when information is extractable, companies will try to use it to gain an advantage,” Rao said. “We all try to be good citizens, but we’re also competing at the same time.”

Why Stopping DeepSeek May Be Difficult

Blocking DeepSeek or similar companies from leveraging U.S. AI advancements is complicated for several reasons:

1. AI distillation doesn’t require direct access to U.S. systems – Instead of stealing data, the method allows models to learn indirectly, making enforcement tricky.
2. The practice is widely used in AI research – Even though some U.S. firms prohibit distillation in their terms of service, monitoring and proving violations can be difficult.
3. AI innovation moves at a rapid pace – By the time regulations catch up, new methods may emerge to bypass restrictions.

The Bigger Picture: Global AI Competition

The DeepSeek case highlights the increasingly competitive race for AI dominance between the U.S. and China. With open-source models and indirect learning techniques making AI more accessible, preventing knowledge transfer between global rivals is becoming a significant challenge for policymakers and tech companies alike.

For now, DeepSeek’s rapid rise is a reminder that in the world of AI, staying ahead means more than just innovation—it also means navigating the complex realities of global competition.

Source

The post Why Stopping China’s DeepSeek from Using U.S. AI is a Challenge appeared first on Journos News - Breaking News, World News, Top Stories, Todays Headlines and Flash Reports.

Neysa, an Indian startup led by seasoned tech entrepreneur Sharad Sanghi, aims to leverage this growth opportunity by offering its AI solutions to local and multinational businesses in the country.

The Mumbai-based startup provides AI and machine learning infrastructure and platform as a service to enterprise customers based on their requirements. It also includes dedicated machine learning operations and infrastructure consulting teams to help customers find the relevant size for their infrastructure, and to fine-tune or customize the models they choose.

Before founding Neysa with his former colleague Anindya Das in 2023, Sanghi spent over 27 years at his previous venture and data center provider, Netmagic, which Japan’s NTT Data acquired in 2016. He told TechCrunch that he intended to focus on cloud infrastructure and AI in 2022 but was unable to do so. He resigned as the managing director and CEO of Netmagic in June 2023 to start fresh with Neysa.

“I started at Neysa with a view of providing infrastructure as a service, platform as a service, inference as a service, the services layer around ML, as well as the platforms that we need for developers,” he said in an interview.

Neysa initially started as an infrastructure service provider and launched its flagship platform, Velocis, in July to provide on-demand access to computing infrastructure. However, it plans to expand the product lineup by launching its developer platform and inference-as-a-service before the year-end. The startup is also working on developing an “observability for better management” of its infrastructure and securing AI workloads, Sanghi said.

With its entire suite of offerings getting ready, Neysa is looking to compete with global hyperscalers, including the typical cloud service providers such as AWS, Google Cloud Platform, and Microsoft Azure, as well as the new-age contenders like CoreWeave and Lambda Labs. Sanghi asserted that the startup differentiates from the existing players by offering “flexibility” in its models.

“We can offer both public cloud and private clusters. It’s also the open-source nature of our offering. All our platforms are built on open-source platforms… so there’s no lock-in for clients,” he stated.

The startup’s consultation service also aims to attract local businesses, which often find it challenging to get the appropriate infrastructure without spending thousands of dollars.

“Very often, clients come to us and say that they want so many GPUs… and when we really look at the requirement, they don’t need half the amount they had asked,” Sanghi said.

Neysa has raised $30 million in an all-equity Series A round co-led by its existing investors NTTVC, Z47 (formerly called Matrix Partners India), and Nexus Venture Partners. This follows up the startup’s $20 million seed round earlier this year.

The fresh funding, Sanghi said, will augment Neysa’s infrastructure, enhance its R&D, and broaden go-to-market. The funds will also set the base for the startup to launch its integrated Gen AI acceleration cloud service.

The startup currently has a headcount of 55 people, which it will grow by adding more engineers and staff to expand direct and indirect sales.

Neysa currently has around 12 paying customers and runs about six large proof-of-concepts. As much as 70 percent of its entire customer base has opted for the private cluster, while the remaining 30 percent is on a public cloud, Sanghi said.

While Sanghi didn’t disclose the names of Neysa’s customers, he said the startup caters to broadly three categories: research institutes, AI-native startups, and enterprise customers, initially in the banking, manufacturing, and media sectors.

Neysa’s current customer base is in India, though Sanghi said the startup does plan to enter global markets with its next round of funding — talks for which have already started, and it is expected to close in the next six to nine months.

He did not reveal the exact amount Neysa seeks to raise in its next round, though he stated that it would be “in an order of magnitude more than what we’ve currently raised.” The startup also plans to raise debt to fulfil the growing GPU and other infrastructure requirements.

Source

The post India’s Neysa Raises $30 Million to Rival Global AI Hyperscalers appeared first on Journos News - Breaking News, World News, Top Stories, Todays Headlines and Flash Reports.