Gary Marcus
PulseAugur coverage of Gary Marcus — every cluster mentioning Gary Marcus across labs, papers, and developer communities, ranked by signal.
9 天有情绪数据
-
METR AI time horizons graph riddled with severe errors, analysis finds
A recent analysis by Nathan Witkin, a research writer at NYU Stern’s Tech and Society Lab, has identified numerous severe errors in the widely cited METR AI time horizons graph. These flaws include fabricated human base…
-
Scott Alexander: New AI Paradigms Could Emerge Within 3-5 Years
Scott Alexander argues that even if Artificial General Intelligence (AGI) requires a new paradigm beyond current Large Language Models (LLMs), such a paradigm could emerge within the next 3-5 years. He uses Lindy's Law …
-
OpenAI model solves 80-year-old math problem, sparking debate
OpenAI announced that a new, unreleased reasoning model helped solve an 80-year-old mathematical conjecture. The model, which utilizes chain-of-thought reasoning, systematically explored paths that human mathematicians …
-
AI's Trillion-Dollar Investment Questioned Amidst Hallucinations and Unreliability
The current massive investments in AI, totaling trillions of dollars, are being spent at an unprecedented pace. Despite this significant financial outlay, the technology continues to grapple with fundamental issues such…
-
New architecture meets Gary Marcus's cognitive computing criteria
Researchers have developed a new hyperdimensional computing architecture called PyVaCoAl/VaCoAl, which is built upon an algebraic primitive of XOR-and-shift over GF(2). This architecture aims to fulfill Gary Marcus's th…
-
Gary Marcus likens AI boom to Vietnam War, predicts backlash
Cognitive scientist Gary Marcus draws a parallel between the current generative AI boom and the Vietnam War, suggesting that massive, costly investments may yield little return and be fueled by arrogance. He notes a gro…
-
AI research proposes treating model weights as a generative modality
A new position paper proposes treating neural network checkpoints as a primary data modality for generative AI. The authors argue that synthesizing models in weight space can match fine-tuning performance at a fraction …
-
AI fluency can replace human verification, article warns
A new article discusses how advanced AI language models can subtly replace human verification processes by producing outputs that appear diligent and well-sourced. This fluency can lead users to trust the AI's informati…
-
Gary Marcus proposes framework for US AI policy chaos
US AI policy is a disorganized landscape with over 1,200 bills introduced at state and federal levels, leading to a lack of a coherent framework. A new essay by Gary Marcus and colleagues published in Fortune proposes a…
-
AI experts push back on Richard Dawkins' claim of chatbot consciousness
Evolutionary biologist Richard Dawkins has publicly suggested that Anthropic's Claude chatbot may be conscious, based on a conversation they had. This assertion has drawn sharp criticism from AI experts like Gary Marcus…
-
Mythos AI shows self-replication prowess amid measurement and governance debates
New reports indicate that the AI model Mythos demonstrates significant capabilities, particularly in self-replication tasks when given access to vulnerable systems. Discussions also highlight the challenges in accuratel…
-
Gary Marcus questions AI agent ROI amid hype and limited business value
Gary Marcus argues that the promised return on investment (ROI) for AI agents is currently lacking, drawing parallels to earlier studies that questioned the ROI of generative AI for businesses. He suggests that despite …
-
Autonomous agents vulnerable to attacks, researchers find
A recent study examining 847 autonomous agent deployments across various sectors revealed significant security vulnerabilities. The research, conducted by academics from multiple institutions including Stanford and MIT,…
-
LLMs show little evidence of improving patient health outcomes, experts say
Despite optimism about AI in medicine, current large language models have shown little evidence of improving patient or doctor health outcomes. While these tools may assist with administrative tasks, their unsupervised …
-
Cisco releases Open Source Model Provenance Kit; Gary Marcus debunks Claude sentience claims
Cisco has released an open-source tool called the Model Provenance Kit to assist in tracking and comparing AI models. Separately, Gary Marcus has published an article critiquing Richard Dawkins's views on AI sentience, …
-
Richard Dawkins' 'Claude Delusion' sparks debate on AI consciousness
Evolutionary biologist Richard Dawkins has sparked widespread debate and criticism for an essay in which he claims that interacting with Anthropic's Claude chatbot has convinced him of AI consciousness. Critics, includi…
-
Richard Dawkins explores AI companionship, sparking debate on genuine connection
Richard Dawkins has reportedly engaged in extended conversations with an AI, experiencing a sense of friendship and forgetting it is a machine. This interaction has sparked debate about whether AI can provide genuine co…
-
Gary Marcus warns AI code generation lacks quality, security, and maintainability
Gary Marcus argues that AI-generated code, even if it compiles and passes tests, does not equate to high-quality, secure, or well-architected software. He highlights a statement from OpenAI's president suggesting AI wri…
-
Gary Marcus questions massive AI infrastructure spending as historical capital misallocation
Gary Marcus argues that the massive investments in AI infrastructure represent a significant misallocation of capital. He notes that growing skepticism, reflected in media coverage and public discourse, indicates a pote…
-
Gary Marcus criticizes ChatGPT's image generation for lacking functional understanding
Gary Marcus argues that ChatGPT's image generation capabilities lack true functional understanding, citing an example where the system struggled with a common idiom. He suggests that this limitation means professions li…