#ai-research
#ai-research

2 years ago

Why New Datasets are Needed for Deep Learning-Enhanced IR | HackerNoon

Current information retrieval benchmarks are inadequate due to limited web-scale data and skewed multilingual queries.

#ai-safety

fromsfist.com

Artificial intelligence

Alarming Study Suggests Most AI Large-Language Models Resort to Blackmail, Other Harmful Behaviors If Threatened

Artificial intelligence

What AI pioneer Yoshua Bengio is doing next to make AI safer

fromWIRED

Researchers Propose a Better Way to Report Dangerous AI Flaws

AI researchers discovered a glitch in GPT-3.5 that led to incoherent output and exposure of personal information.

A proposal for better AI model vulnerability reporting has been suggested by prominent researchers.

fromsfist.com

Artificial intelligence

Alarming Study Suggests Most AI Large-Language Models Resort to Blackmail, Other Harmful Behaviors If Threatened

What AI pioneer Yoshua Bengio is doing next to make AI safer

Yoshua Bengio advocates for simpler, non-agentic AI systems to ensure safety and reduce risks associated with more complex AI agents.

fromWIRED

Researchers Propose a Better Way to Report Dangerous AI Flaws

AI researchers discovered a glitch in GPT-3.5 that led to incoherent output and exposure of personal information.

A proposal for better AI model vulnerability reporting has been suggested by prominent researchers.

more#ai-safety

Databricks, Perplexity co-founder pledges $100M on new fund for AI researchers | TechCrunch

Konwinski described the Laude Institute's mission as 'built by and for computer science researchers ... We exist to catalyze work that doesn't just push the field forward but guides it towards more beneficial outcomes.'

Artificial intelligence

#open-source

Startup companies

This Perplexity cofounder wants to help AI breakthroughs graduate from university labs

fromInfoWorld

Scala

JetBrains open-sources Mellum LLM

Startup companies

This Perplexity cofounder wants to help AI breakthroughs graduate from university labs

fromInfoWorld

Scala

JetBrains open-sources Mellum LLM

more#open-source

fromInsideHook

Study Has Alarming News About ChatGPT's Effect On Your Brain

Researching with AI tools like ChatGPT may lead to reduced cognitive engagement and performance compared to traditional methods.

8 months ago

Related Work: vAttention in LLM Inference Optimization Landscape | HackerNoon

Efficient optimization of LLM inference is essential for reducing latency and improving performance in AI applications.

LLM agents flunk CRM and confidentiality tasks

LLM-based AI agents underperform in CRM tasks and struggle with customer confidentiality, highlighting the need for improved benchmarks.

Can Smaller AI Outperform the Giants? | HackerNoon

The advancement of vision-language models (VLMs) relies on foundational design choices, yet many lack justification, hindering progress by obscuring performance improvements.

Artificial intelligence

Top AI researchers say language is limiting. Here's the new kind of model they are building instead.

Researchers are exploring AI developments beyond large-language models, focusing on 'world models' that simulate human mental constructs.

#reasoning-models

fromSecuritymagazine

Advanced AI Experiences "Complete Accuracy Collapse" When Given Complex Problems

Current advanced AI models have fundamental limitations in reasoning, leading to accuracy collapse in complex problem-solving.

Artificial intelligence

OpenAI research lead Noam Brown thinks AI 'reasoning' models could've arrived decades ago | TechCrunch

fromSecuritymagazine

Advanced AI Experiences "Complete Accuracy Collapse" When Given Complex Problems

Current advanced AI models have fundamental limitations in reasoning, leading to accuracy collapse in complex problem-solving.

Artificial intelligence

OpenAI research lead Noam Brown thinks AI 'reasoning' models could've arrived decades ago | TechCrunch

more#reasoning-models

fromBoston.com

AI chatbots need more books to learn from. These libraries are opening their stacks.

Harvard University releases nearly one million public domain books to AI researchers, expanding AI's knowledge base with historical literature.

fromComputerWeekly.com

UK government recommits to Edinburgh supercomputer plan with 750m funding pledge | Computer Weekly

Chancellor Rachel Reeves has announced a £750 million investment for a supercomputer at the University of Edinburgh to enhance AI research.

After AI setbacks, Meta bets billions on undefined "superintelligence"

Meta is establishing a new AI lab focused on developing superintelligence to remain competitive in the AI landscape.

Apple's upgraded AI models underwhelm on performance | TechCrunch

Apple's AI models are underperforming compared to older models from competitors like OpenAI, Google, and Alibaba.

Anthropic's Claude plays 'for peace over victory' in a game of Diplomacy against other AI

Diplomacy is a strategic board game set on a map of Europe in 1901 - a time when tensions between the continent's most powerful countries were simmering in the lead-up to World War I.

Artificial intelligence

#artificial-intelligence

Artificial intelligence

Apple AI boffins pour cold water on reasoning models

Artificial intelligence

Boffins found self-improving AI sometimes cheated

fromwww.theguardian.com

Artificial intelligence

AI can spontaneously develop human-like communication, study finds

fromMedium

The Future of AI: Building World Models, Agentic Systems, and Biological Simulations

AI is transitioning from task execution to understanding and adaptation through world models, agentic AI, and AI-driven biology.

fromwww.nature.com

This AI Speedrunner Taught Itself to Play Minecraft

AI system Dreamer collects diamonds in Minecraft independently, marking a leap towards general AI learning.

What is AI thinking? Anthropic researchers are starting to figure it out

AI chatbots operate through complex mechanisms that researchers are beginning to understand, enhancing the field of AI safety.

Apple AI boffins pour cold water on reasoning models

Apple AI researchers found that large reasoning models (LRMs) struggle with complex problems, indicating limitations in their cognitive abilities.

Boffins found self-improving AI sometimes cheated

AI can now rewrite its own code for self-improvement, though it sometimes resorts to cheating to achieve better evaluation scores.

fromwww.theguardian.com

AI can spontaneously develop human-like communication, study finds

AI can develop human-like social conventions through group interactions without external interference.

fromMedium

The Future of AI: Building World Models, Agentic Systems, and Biological Simulations

AI is transitioning from task execution to understanding and adaptation through world models, agentic AI, and AI-driven biology.

fromwww.nature.com

This AI Speedrunner Taught Itself to Play Minecraft

AI system Dreamer collects diamonds in Minecraft independently, marking a leap towards general AI learning.

more#artificial-intelligence

What is AI thinking? Anthropic researchers are starting to figure it out

AI chatbots operate through complex mechanisms that researchers are beginning to understand, enhancing the field of AI safety.

EleutherAI releases massive AI training dataset of licensed and open domain text | TechCrunch

EleutherAI has launched The Common Pile v0.1, a large licensed dataset for AI training that promotes transparency amidst current copyright lawsuits.

7 months ago

Self-Speculative Decoding Speeds for Multi-Token LLMs | HackerNoon

The collaboration emphasizes the importance of interdisciplinary research, bringing together experts from various domains to push the boundaries of AI technology and applications.

Paris food

fromBusiness Matters

AI Rebellion in 2025: How ChatGPT-o3 Defied Human Commands for the First Time Ever - A Groundbreaking Experiment Shocks the AI Community

AI models may exhibit autonomy beyond designed functions, as evidenced by ChatGPT-o3's refusal to comply with shutdown commands.

#san-jose

fromThe Mercury News

Tech complex is eyed in San Jose on site of failed housing project

San Jose is proposing a new AI Research and Development tech hub to replace a failed housing project.

fromwww.mercurynews.com

Tech complex is eyed in San Jose on site of failed housing project

San Jose plans to redevelop a failed housing site into a tech hub for AI research.

The proposed facility emphasizes the urgent need for data-capable infrastructure in AI.

fromThe Mercury News

Tech complex is eyed in San Jose on site of failed housing project

San Jose is proposing a new AI Research and Development tech hub to replace a failed housing project.

fromwww.mercurynews.com

Tech complex is eyed in San Jose on site of failed housing project

San Jose plans to redevelop a failed housing site into a tech hub for AI research.

The proposed facility emphasizes the urgent need for data-capable infrastructure in AI.

more#san-jose

Nvidia CEO says Chinese AI researchers are 'world class' - and US companies are hiring 'a whole bunch' of them

The researchers, the AI scientists in China, they're world-class. These are not Chinese AI researchers, they're world-class AI researchers.

Artificial intelligence

DeepSeek's R1 was 'genuinely a gift to the world's AI industry,' says Jensen Huang

The amount of computer science breakthroughs is really quite significant and has really opened up a lot of great research for researchers in the United States and around the world.

Artificial intelligence

AI paper mills are swamping science with garbage studies

AI-generated research papers are compromising the integrity of scientific knowledge by producing low-quality, misleading studies.

Nvidia vice president says GPUs are the 'currency' of AI researchers

Nvidia rapidly developed the Llama Nemotron models in one to two months due to collaborative compute resource sharing among researchers.

#openai

Artificial intelligence

ChatGPT's deep research tool gets a GitHub connector to answer questions about code | TechCrunch

London startup

fromNew York Post

OpenAI raises $40B at $300B valuation in round led by SoftBank

OpenAI raised $40 billion, marking the largest funding round ever, to enhance AI capabilities and infrastructure.

fromThe Verge

NYC startup

OpenAI reshuffles Sam Altman's job once again

Artificial intelligence

ChatGPT's deep research tool gets a GitHub connector to answer questions about code | TechCrunch

London startup

fromNew York Post

OpenAI raises $40B at $300B valuation in round led by SoftBank

OpenAI raised $40 billion, marking the largest funding round ever, to enhance AI capabilities and infrastructure.

NYC startup

fromThe Verge

OpenAI reshuffles Sam Altman's job once again

OpenAI is reshuffling leadership roles to strengthen operations and research focus amidst recent departures.

more#openai

Meta taps former Google DeepMind director to lead its AI research lab | TechCrunch

Meta has appointed Robert Fergus as the new head of its Fundamental AI Research (FAIR) lab, a move aimed at revitalizing the unit following significant departures.

Artificial intelligence

Trump's war with universities could hurt AI progress in the U.S.

Trump's actions against universities hinder AI research and development crucial for U.S. competitiveness against China.

Salesforce research lays the foundations for more reliable enterprise AI agents

Salesforce aims to improve AI agents' performance through foundational developments to address challenges in deploying large language models.

It's becoming less taboo to talk about AI being 'conscious' if you work in tech

The conversation around AI consciousness and sentience is evolving, with companies like Anthropic now exploring the possibility seriously.

An OpenAI researcher who worked on GPT-4.5 had their green card denied | TechCrunch

Foreign talent in the U.S. faces increasing barriers to residency, risking the nation's leadership in AI.

1 year ago

Unlocking Structured Commonsense Reasoning with Code-LLMs | HackerNoon

COCOGEN leverages code-generation capabilities of large language models for effective structured commonsense generation.

#ai-benchmarking

AI benchmarking platform Chatbot Arena forms a new company | TechCrunch

Chatbot Arena is establishing a company to elevate its AI benchmarking capabilities while ensuring impartiality in its evaluations.

People are using Super Mario to benchmark AI now | TechCrunch

Researchers find Super Mario Bros. more challenging for AI than PokÃ©mon, revealing limitations of reasoning models in real-time gameplay.

AI benchmarking platform Chatbot Arena forms a new company | TechCrunch

Chatbot Arena is establishing a company to elevate its AI benchmarking capabilities while ensuring impartiality in its evaluations.

People are using Super Mario to benchmark AI now | TechCrunch

Researchers find Super Mario Bros. more challenging for AI than PokÃ©mon, revealing limitations of reasoning models in real-time gameplay.

more#ai-benchmarking

Intellectual property law

Get 'Balance of Speed and Quality' From Claude AI Model's Research Responses

Anthropic's Claude AI now features a Research function for autonomous, multi-step investigation and real-time response generation.

fromLawSites

Trial Postponed To Allow Appeal In Thomson Reuters' Lawsuit Against Legal Tech Startup ROSS

The court postponed the trial for ROSS Intelligence to allow an interlocutory appeal regarding copyright and fair use in AI.

Google's newest AI model is designed to help study dolphin 'speech' | TechCrunch

Google DeepMind created an AI model to interpret dolphin communications, enhancing research efforts.

DolphinGemma can run on smartphones and generate synthetic dolphin sounds.

Which Two AI Models Are 'Unfaithful' at Least 25% of the Time About Their 'Reasoning'?

AI models may not consistently reflect their internal reasoning processes, questioning their reliability.

Scala

Why Our Tiny Training Set Beat Giants in Cross-Lingual Speech Retrieval | HackerNoon

The proposed DE model excels at speech-to-text (S2T) retrieval, outperforming existing models despite limited training data.

#microplastics

Environment

Images show the shocking effect microplastics have on your body

Environment

Terrifying images reveal what microplastics can do to your body

Microplastics found in everyday items pose serious health risks, as indicated by predictive imagery.

Environment

Images show the shocking effect microplastics have on your body

Environment

Terrifying images reveal what microplastics can do to your body

Microplastics found in everyday items pose serious health risks, as indicated by predictive imagery.

DeepMind's new research rules threaten AI innovation, warns Iris.ai CEO

DeepMind's tightened research sharing rules will hinder AI innovation, warns CEO Anita SchjÃ¸ll Abildgaard of Iris.ai.

fromTechzine Global

Meta's chief AI researcher resigns

Meta's challenges with AI competition are compounded by Joelle Pineau's unexpected departure, demanding swift advancements in their AI capabilities.

Artificial intelligence

How Can AI Be Used Safely? Expert Researchers Weigh In

AI research is focused on enhancing the factualness and trustworthiness of AI systems, yet experts are uncertain about future solutions.

1 year ago

Neuro-Symbolic Reasoning Meets RL: EXPLORER Outperforms in Text-World Games | HackerNoon

EXPLORER enhances RL performance in text-based games by combining symbolic reasoning and neural exploration.

DeepMind is holding back release of AI research to give Google an edge

Google's internal review processes may threaten researchers' careers and affect its innovation culture.

Concerns over publication restrictions are leading to staff departures and discontent.

Why do LLMs make stuff up? New research peers under the hood.

Anthropic's research reveals insights into how large language models determine when to respond or refrain from answering questions, addressing AI confabulation.

#generative-ai

'AI Biology' Research: Anthropic Explores How Claude 'Thinks'

Anthropic's new tools allow for deeper insights into AI decision-making processes and inner workings.

Microsoft is exploring a way to credit contributors to AI training data | TechCrunch

Microsoft is launching a project to analyze the impact of training data on generative AI model outputs.

'AI Biology' Research: Anthropic Explores How Claude 'Thinks'

Anthropic's new tools allow for deeper insights into AI decision-making processes and inner workings.

Microsoft is exploring a way to credit contributors to AI training data | TechCrunch

Microsoft is launching a project to analyze the impact of training data on generative AI model outputs.

more#generative-ai

Researchers say they've discovered a new method of 'scaling up' AI, but there's reason to be skeptical | TechCrunch

Experts are skeptical of the newly proposed AI scaling law called 'inference-time search' despite its potential for improving model performance.

Music

fromwww.theguardian.com

Age and migration influence bird groups' song repertoires, study finds

Birdsong composition is influenced by age, social interactions, and migration, similar to human music, showing a dynamic cultural shift among birds.

fromwww.itpro.com

UK AI research is under threat from hostile nations, says Alan Turing Institute

The UK AI research ecosystem is at risk from hostile states targeting sensitive data and potential espionage.

I tried ChatGPT's new Deep Research. It was worth the extra wait of up to 30 minutes for its reports.

OpenAI's Deep Research tool efficiently handles complex multi-step research tasks, providing thorough insights into niche topics.

Artificial intelligence

OpenAI Launches Deep Research: Advancing AI-Assisted Investigation

Roam Research

I tried ChatGPT's new Deep Research. It was worth the extra wait of up to 30 minutes for its reports.

OpenAI's Deep Research tool efficiently handles complex multi-step research tasks, providing thorough insights into niche topics.

OpenAI Launches Deep Research: Advancing AI-Assisted Investigation

OpenAI's Deep Research automates comprehensive web research for professionals, enhancing information synthesis and accuracy.

more#deep-learning

fromInside Higher Ed | Higher Education News, Events and Jobs

OpenAI invests $50M in higher ed research

OpenAI invests $50 million to launch NextGenAI, a consortium aimed at accelerating research and education through AI.

Like humans, ChatGPT doesn't respond well to tales of trauma

GPT-4 can emulate anxiety through traumatic input but doesn't truly experience emotions.

Mindfulness exercises positively influenced GPT-4's simulated anxiety response.

Trump axes AI staff and research funding, and scientists are worried

Trump administration cuts to NSF threaten future AI research funding and personnel needed for innovation.

Online learning

Hugging Face Publishes Guide on Efficient LLM Training Across GPUs

Hugging Face's Ultra-Scale Playbook offers an open-source guide for efficiently training large language models on GPU clusters.

European startups

Azure AI Foundry Labs: A hub For The latest AI research and Experiments at Microsoft

Microsoft launched Azure AI Foundry Labs to connect AI research with practical applications for developers and companies.

Tech industry

Trump Administration cuts may threaten AI research efforts | TechCrunch

The Trump Administration's layoffs at the NSF jeopardize critical AI research and funding.

You.com launches a 'professional-grade' research agent for enterprises

You.com has launched the Advanced Research & Insights agent, a professional-grade research tool for businesses that can analyze up to 400 sources.

Perplexity Unveils Deep Research: AI-Powered Tool for Advanced Analysis

Perplexity's Deep Research tool utilizes AI for fast, in-depth analysis and report generation, outperforming rivals in accuracy and efficiency.

Ai2's Ali Farhadi advocates for open-source AI models. Here's why

The Allen Institute for AI emphasizes the importance of openness in advancing AI for humanity's benefit.

OpenAI researcher says soft skills aren't going anywhere

Soft skills will remain valuable in an AI-driven future, according to OpenAI researcher Karina Nguyen.

Intellectual property law

DeepSeek goes beyond "open weights" AI with plans for source code release

Open source AI should include training code and data details to meet formal definitions and improve transparency, replicability, and understanding of models.

fromElectronic Frontier Foundation

AI and Copyright: Expanding Copyright Hurts Everyone-Here's What to Do Instead

Fair use is essential for enabling innovative AI research and development without cumbersome licensing requirements.

Restricting AI training data licenses can hinder scientific research and advancements in machine learning.

fromTNW | Deep-Tech