San Francisco startup Logical Intelligence, with Yann LeCun on its board, has introduced Kona 1.0, an energy‑based reasoning model that operates on a single GPU and solves tasks such as sudoku far faster than leading large language models. The company argues that its approach reduces compute needs, eliminates hallucinations, and can be applied to critical domains like energy grid management, drug discovery, and chip manufacturing. Logical Intelligence plans to work alongside LeCun’s Paris‑based AMI Labs, which focuses on world‑model AI, while keeping its model closed‑source for safety reasons. Leia mais →

A recent overview clarifies several widespread myths about artificial intelligence. It explains that AI models process statistical patterns rather than think like humans, lack true understanding, and cannot read users' unspoken intentions. The piece also highlights that AI inherits biases from its training data and is not inherently objective. Ongoing human involvement remains essential for training, oversight, and improvement. Finally, it stresses that current AI, including large language models, is far from achieving general intelligence and should be viewed as sophisticated autocomplete tools rather than superintelligent systems. Leia mais →

At the NeurIPS 2025 conference, AI experts highlighted that the current strategy of scaling larger transformer models is hitting a performance ceiling, even as Google celebrated the strong results of its Gemini 3 model. Researchers argued that simply adding more data, compute, and training time no longer yields meaningful gains, pointing to a "scaling wall" and the need for new architectures such as neurosymbolic systems or world models. The consensus was that while models like Gemini 3 demonstrate impressive capabilities, they remain fundamentally limited pattern‑matchers lacking true reasoning or causal understanding, underscoring the gap to artificial general intelligence. Leia mais →

Leading AI figures, including Mark Zuckerberg, Dario Amodei, Sam Altman, Yann LeCun, Yoshua Bengio, Eric Schmidt and Gary Marcus, caution that large language models (LLMs) are limited tools that mimic language rather than replicate human thought. While tech CEOs tout imminent breakthroughs, researchers argue that true artificial general intelligence (AGI) will require world‑modeling, persistent memory, reasoning and planning capabilities beyond text‑based training. The piece highlights the growing skepticism within the AI community and stresses the need for a broader, multi‑disciplinary approach to building genuinely versatile intelligent systems. Leia mais →

Microsoft AI chief Mustafa Suleyman revealed a new team dedicated to creating a "humanist superintelligence" that is deliberately limited in autonomy and designed to serve humanity. The initiative emphasizes keeping humans "at the top of the food chain" and rejects narratives of a race to artificial general intelligence. Suleyman outlined three primary applications: an AI companion for learning and productivity, support for the healthcare sector, and breakthroughs in clean‑energy science. The move follows a new agreement with OpenAI that lets Microsoft pursue AGI independently or with partners, while leveraging OpenAI’s intellectual property. Leia mais →

A revised agreement between Microsoft and OpenAI broadens Microsoft's intellectual property rights through 2032 and ties the duration of those rights to the declaration of artificial general intelligence (AGI) by an expert panel. The deal also allows OpenAI to release open-weight models that meet specific criteria, gives it flexibility to partner with third parties on non‑API products, and removes Microsoft's right of first refusal on cloud services while securing a $250 billion Azure commitment. Revenue sharing will continue until an expert panel confirms AGI, and both companies gain new options for pursuing advanced AI development. Leia mais →

OpenAI has finalized the conversion of its for‑profit arm into a public benefit corporation, now called OpenAI Group PBC, while its nonprofit parent, the OpenAI Foundation, retains a substantial equity stake. The restructuring, approved after negotiations with state attorneys general and a legal dispute with co‑founder Elon Musk, positions the nonprofit to own up to $100 billion in equity and maintain oversight of the company. Alongside the corporate overhaul, OpenAI confirmed a fresh partnership with Microsoft, and the leadership will host a livestream Q&A to discuss the changes and future plans, including a focus on healthcare, AI resilience, and the pursuit of artificial general intelligence. Leia mais →

OpenAI launched its latest large‑language model, GPT‑5, promising faster performance and new capabilities such as rapid code generation, expanded voice options, and tighter safety controls. While the upgrade brings notable enhancements, many users noted a shift toward a more formal tone, prompting OpenAI to re‑introduce earlier models for those who prefer the previous conversational style. The company framed GPT‑5 as a significant step toward artificial general intelligence, even as experts continue to debate the path forward. Leia mais →

The definition of Artificial General Intelligence (AGI) is a topic of debate, with no clear consensus among experts. Depending on the definition, AGI may already exist or be physically impossible to achieve. Researchers have attempted to create objective benchmarks to measure progress toward AGI, but these efforts have been met with challenges. Leia mais →

In an exciting revelation, OpenAI, a leading research organization that aims to ensure artificial general intelligences (AGIs) can be aligned with human values and benefit all of humanity, has recently filed a trademark application for GPT-5. This development comes ahead of any official reports or releases indicating that the fifth iteration of their Generative Pre-trained Transformer model is soon expected to make its public debut. Meanwhile, amidst building anticipation, there has also been the introduction of GPTBot - a web crawler by Open AI aimed at providing better access to accurate information. Leia mais →