🚀 Medial Secures Investment on Shark Tank India - Fueling the Future of Professional Social Networking. 🔥

Startup Showcase

Premium Content

Try our Valuation Calculator →

News on Medial

Apple’s new research paper says AI reasoning isn’t all it’s cracked up to be.

The Verge

The Verge · 1m ago

Apple’s new research paper says AI reasoning isn’t all it’s cracked up to be.

Apple's new research paper, "The Illusion of Thinking," highlights the limitations of AI reasoning, especially when faced with unfamiliar complexities. The study found that AI models from OpenAI, Anthropic, and DeepSeek demonstrate increasing reasoning abilities with problem complexity to a point, after which performance declines. This research suggests that current AI models may struggle beyond certain complexity levels, particularly with novel tasks, challenging the perception of AI's reasoning capabilities.

1

Related News

Apple says it uses no Nvidia GPUs to train its AI models

Economic Times

Economic Times · 12m ago

Apple says it uses no Nvidia GPUs to train its AI models

Apple has used chips designed by Google instead of Nvidia for its new artificial intelligence (AI) software infrastructure, according to a research paper published by Apple. This decision is notable because Nvidia is a leader in AI processors. Apple did not explicitly confirm the absence of Nvidia chips in its AI tools, but its research paper did not mention Nvidia hardware. Instead, Apple relied on Google's tensor processing unit (TPU) chips, with 2,048 TPUv5p chips used for its mobile AI model and 8,192 TPUv4 chips for its server AI model. Apple plans to roll out the AI features to beta users this week.

Apple’s MM1 AI Model Shows a Sleeping Giant Is Waking Up

Wired · 1y ago

Apple’s MM1 AI Model Shows a Sleeping Giant Is Waking Up

Apple, known for its secrecy around AI, has quietly published a research paper detailing its new generative AI model called MM1. The model is trained on both text and images, making it multimodal, and is capable of answering questions about photos and displaying general knowledge skills. Apple engineers describe the model as the "beginning" and are already working on the next generation of models. The company's move into generative AI could potentially enhance its virtual assistant, Siri, and compete with the AI tools offered by rivals Samsung and Google. It remains to be seen how Apple will incorporate its AI models into its products.

Apple Paper questions path to AGI, sparks division in GenAI group

Economic Times

Economic Times · 1m ago

Apple Paper questions path to AGI, sparks division in GenAI group

Apple's research paper critiques large AI models for lacking true reasoning ability, revealing that they excel in pattern recognition but fail with complex problems. Apple's findings suggest caution in the pursuit of artificial general intelligence (AGI) compared to competitors like Google. Critics from Anthropic and Open Philanthropy contest the study's design, arguing models perform well with code-based tasks. The study highlights debates on AI's limitations and spurs discussions about its trajectory towards AGI.

ETtech Explainer: Is Apple’s ReALM better than OpenAI’s GPT-4?

Economic Times

Economic Times · 1y ago

ETtech Explainer: Is Apple’s ReALM better than OpenAI’s GPT-4?

Apple has revealed some details about its artificial intelligence plans in a recent research paper. The company discussed its large language model (LLM) called Reference Resolution As Language Modeling (ReALM) and how it surpasses OpenAI's GPT-4. ReALM focuses on resolving references, such as ambiguous or contextual words, to create a better understanding in AI chatbots. Apple's models have shown improvements over existing systems, including a 5% gain for on-screen references. While ReALM performs well in specific benchmarks, it is not yet clear if it outperforms GPT-4 overall.

Fractal rolls out Fathom-R1-14B, an open-source AI model for reasoning

YourStory

YourStory · 2m ago

Fractal rolls out Fathom-R1-14B, an open-source AI model for reasoning

Fractal has launched Fathom-R1-14B, an open-source AI model with 14 billion parameters, aimed at enhancing mathematical and general reasoning capabilities. Developed by Fractal’s AI Research team, it is built on Deepseek-R1-Distilled-Qwen-14B and supports a 16K context length for better comprehension of extensive inputs. This model, part of IndiaAI Mission’s proposal for India’s first Large Reasoning Model, outperforms other open-source models and excels in reasoning and math benchmarks.

OpenAI and Meta are on the verge of releasing AI models capable of reasoning like humans, report says

Business Insider

Business Insider · 1y ago

OpenAI and Meta are on the verge of releasing AI models capable of reasoning like humans, report says

AI companies OpenAI and Meta are gearing up to release advanced AI models that can tackle complex problem-solving tasks. OpenAI's next version of GPT aims to make progress in solving "hard problems" like reasoning, while Meta's Llama 3 model aims to exhibit capabilities in talking, reasoning, planning, and memory. Achieving artificial general intelligence (AGI) has been a shared goal for both companies and is seen as a trillion-dollar industry opportunity. Nevertheless, concerns about the safety and ethical implications of surpassing human intelligence continue to be raised by experts in the field.

Google Quantum AI Breakthrough Makes 2048-bit RSA Encryption 20 Times Easier to Break in One Week, Threatens Bitcoin Within Five Years | DeepNewz Markets

Business Bytes

Business Bytes · 2m ago

Google Quantum AI Breakthrough Makes 2048-bit RSA Encryption 20 Times Easier to Break in One Week, Threatens Bitcoin Within Five Years | DeepNewz Markets

The Google Quantum AI team has achieved a significant breakthrough, as detailed in a new paper. They demonstrate that a quantum computer with 1 million qubits, operating for a week, can break 2048-bit RSA encryption. This advancement represents a twentyfold improvement over research from five years ago.

Mistral's Le Chat chatbot gets a productivity push with new ‘deep research' mode | TechCrunch

TechCrunch · 9d ago

Mistral's Le Chat chatbot gets a productivity push with new ‘deep research' mode | TechCrunch

Mistral has enhanced its Le Chat chatbot with features like "deep research" mode, improved multilingual reasoning, and advanced image editing. The deep research function helps users search and synthesize data, useful for both consumer and enterprise tasks. Le Chat also integrates with enterprise data on-premises, avoiding cloud dependency, and aims to seamlessly work with office suites. Additionally, Le Chat supports multilingual reasoning and projects organization, marking its position as a competitive AI tool.

Wingify’s Paras Chopra eyes India's new tech subsidy to bootstrap new AI venture

Livemint

Livemint · 5m ago

Wingify’s Paras Chopra eyes India's new tech subsidy to bootstrap new AI venture

Wingify founder Paras Chopra is seeking government support in accessing subsidized GPUs for developing a foundational AI model through his new venture, Lossfunk. Chopra, using his $200 million from Wingify’s sale, aims to innovate in AI with a focus on creativity, reasoning, and efficiency. He emphasizes strategic investment for breakthrough research. Meanwhile, India’s government is committed to AI growth, balancing policy and innovation, amidst privacy concerns and regulatory challenges in the global AI landscape.

OpenAI working on new reasoning technology under code name ‘Strawberry’

Livemint

Livemint · 1y ago

OpenAI working on new reasoning technology under code name ‘Strawberry’

OpenAI is reportedly working on a new reasoning technology called Strawberry to enhance the reasoning capabilities of its language models. The project aims to enable AI models to navigate the internet autonomously and perform advanced research. Strawberry involves post-training OpenAI's generative models to improve their performance in specific ways. OpenAI intends to use this technology for long-horizon tasks and develop a computer-using agent capable of taking action based on its findings. The project, previously known as Q*, has shown promising results in answering complex science and math questions.

Trackers

Active Indian VC’s

OG Capital Email

With a hands-on approach, OG Capital aims to invest in over 20 promising...

Accel Partners Email

Early and growth-stage investments in disruptive technology companies with...

Early-stage venture capital firm investing in technology startups in India. Focus on...

Access All Trackers

Startup Showcase Winners

June 2025

Helping your parents when you are miles away

The Pit Stop Your Cravings Deserve

The next generation E-commerce platform

Enter Ongoing Startup Showcase

Top Users

Trending News on Medial

Download the medial app to read full posts, comements and news.