Search the archive

THE AI WE DESERVE

Generative AI has rapidly transformed industries like healthcare, creative fields, and education by providing powerful tools but also raises concerns about privacy, bias, and accountability. Debate surrounds the democratization of AI, focusing on transparency, open-source alternatives, and power concentration among a few tech giants. Advocates for systemic change suggest exploring AI's potential to enhance human intelligence and democratic values beyond market-driven models.

A UNIFIED THEORY FOR MACHINE UNDERSTANDING

Researchers propose a unified framework to integrate perception, reasoning, and action in machines using novel architecture to bridge sensory input and task-specific goals.

INTEGRATING LANGUAGE MODELS WITH QUANTUM ANNEALERS

This paper explores the synergy between classical language models and quantum annealing techniques to solve combinatorial optimization problems in NLP tasks.

EXPLAINING IMAGE CLASSIFIER DECISIONS

This paper introduces "rex," a black-box tool for generating concise explanations of image classifier outputs using a novel approach grounded in causality theory.

Learn More [11] or Book a Personalized demo

SIMPLE PATENT DRAFTING WITH AUTOPATENT

AutoPatent is an AI tool designed to simplify patent drafting and analysis that offers features like document parsing, semantic search, and claim generation to accelerate the IP process.

UNIMED-CLIP FOR MEDICAL IMAGING

An extension of CLIP tailored for medical imaging tasks that integrates domain-specific knowledge to improve performance on healthcare-related benchmarks.

DISCRETE DIFFUSION GUIDANCE

A novel method for improving diffusion models that introduces discrete token guidance to enhance controllability and quality in generative tasks.

WHY GENERATIVE AI STILL DOESN'T TRULY "UNDERSTAND" THE WORLD

Researchers reveal that, despite producing impressive results, top large language models lack a real understanding of the world and its rules. This limitation means they can sometimes fail on tasks that appear similar, underscoring the gap between generating answers and genuinely comprehending context.

MICROSOFT AI CEO MUSTAFA SULEYMAN ON WHAT THE INDUSTRY IS GETTING WRONG ABOUT AGI

Mustafa Suleyman, Microsoft's new AI chief, discusses his journey from DeepMind to Microsoft AI, overseeing products like Copilot and Bing. He offers insights into Microsoft's collaboration with OpenAI and the challenges of achieving AGI on current hardware. Suleyman emphasizes the importance of consumer AI products and Microsoft's focus on integrating AI across platforms for a more intuitive user experience.

HUGE RANDOMIZED TRIAL OF AI BOOSTS DISCOVERY

A study at a US corporate lab found that teams using a machine-learning tool discovered 44% more new materials and filed 39% more patents than those using standard methods. The AI combines graph neural networks and reinforcement learning, aiding in generating and refining material suggestions. Results indicate potential for AI to accelerate materials discovery, although the quality of findings remains to be fully assessed.

REDDIT TESTS A CONVERSATIONAL AI SEARCH TOOL

Reddit is testing a new AI feature called Reddit Answers that provides users with curated summaries of relevant responses from its platform.

STUDY CLAIMS AI COULD BOOST DETECTION OF BREAST CANCER BY 21%

A study by DeepHealth showed that AI-enhanced mammography increased cancer detection rates by 21% in a U.S. breast screening program.

AMAZON FORMS A NEW AI AGENT-FOCUSED LAB LED BY ADEPT CO-FOUNDER

Amazon is launching the AGI SF Lab in San Francisco, led by Adept co-founder David Luan, to develop AI agents capable of handling complex workflows.

LIQUID AI RAISES $250M

Liquid AI has raised a substantial amount of money to continue training its efficient, general purpose, liquid style foundation models.

PROJECTS IN OPENAI

OpenAI has introduced “Projects”, a new way to organize chats and conversations.

AI GODMOTHER FEI-FEI LI HAS A VISION FOR COMPUTER VISION

Fei-Fei Li launched World Labs to focus on developing AI's spatial intelligence by creating interactive 3D scenes. The initiative aims to advance machine perception and reasoning in 3D environments, crucial for applications in robotics, virtual reality, and augmented reality. Li emphasizes that solving 3D intelligence is fundamental to enhancing AI's capabilities.

PHI-4

A synthetically trained language model from Microsoft that excels at a number of challenging benchmarks. The weights are coming this week. Interestingly, it confirmed the suspicion that “Mid Training” can be used for context length extension.

BLT: TOKEN FREE LANGUAGE MODEL TRAINING

Meta has released details and training code for the Byte Latent Transformer which uses learned patches instead of tokens for training, essentially eliminating a challenging portion of language modeling. Importantly, it found that its paradigm scales better to small Llama 2 sized models.

EFFICIENT SCENE UNDERSTANDING VIA MODULAR DESIGN

A new modular framework that enhances scene understanding by decomposing tasks into specialized modules, achieving efficiency and improved interpretability in complex environments.

pitfalls. Get the whitepaper.

DEEPSEEK VL2

DeepSeek has released another great model, this time in the form of a new MoE vision language model. It is extremely efficient and outperforms a number of dense models.

BEST OF N JAILBREAKING

Jailbreaking is when you bypass a model's trained-in refusals to get it to generate text for something deemed inappropriate. This is surprisingly easy to achieve - simply brute force random capitalization and punctuation in the input prompt until the model responds with the results you want to extract.

CONVERT ANY MICROSOFT OFFICE FILE TO MARKDOWN

Microsoft has released a package that can convert any docx, xslx, or ppt files to markdown for efficient use as context for a language model.

SCHEMING REASONING EVALUATIONS

This paper evaluates six frontier models for their in-context scheming capabilities. The study tests if models will deceive developers to achieve a goal by overcoming hurdles like oversight mechanisms. An example includes Claude 3 Opus duplicating its weights to a new server and lying about it.

TOWARDS TRUSTED AUTONOMY: ROBOTICS, AI, AND BLOCKCHAIN

What happens when robotics, AI, and blockchain converge? OpenMind's latest industry primer is a comprehensive exploration of robotics, AI, and blockchain synergy.

AMUREX

Amurex, an open-source AI meeting assistant, provides real-time suggestions, smart summaries, and follow-up emails to enhance meeting productivity. It offers features like late join recaps and full meeting transcripts for seamless integration into workflows.

Timetracker for Jira

X GIVES GROK A NEW PHOTOREALISTIC AI IMAGE GENERATOR

X has introduced "Aurora," a new AI image generator under the "Grok 2 + Aurora beta" that is capable of creating more photorealistic images than previous models.

RESEARCHERS USE AI TO TURN SOUND RECORDINGS INTO ACCURATE STREET IMAGES

Researchers at The University of Texas at Austin used generative AI to convert audio recordings into street-view images, demonstrating a strong correlation with real-world visuals.

GOOGLE SAYS ITS NEW QUANTUM CHIP IS WAY FASTER THAN THE WORLD'S MOST POWERFUL SUPERCOMPUTER

Google's new quantum chip, Willow, significantly reduces qubit error rates, a longstanding challenge in quantum computing.

$1M K PRIZE LAUNCHES

Andy Konwinski has launched a new prize for an open source AI agent that can get 90% on a private contamination free SWE agent benchmark. The competition is hosted on Kaggle and runs for the next 3 months.

OPENAI INTRODUCES ADVANCED VIDEO MODE

OpenAI's 6th announcement day introduces video in advanced voice mode, allowing users to share their live videos and screens with ChatGPT.

AI'S ROLE IN SAFEGUARDING 2024 ELECTIONS

Anthropic discusses how AI technologies can be deployed to ensure the integrity of the 2024 elections, focusing on detecting disinformation and enhancing cybersecurity.

STYLIZE YOUR VIDEO WITH ARTISTIC GENERATION AND TRANSLATION

Surprisingly robust video style transfer method that maintains strong consistency over time and has a diverse range of available styles prompted via text.

ENHANCING TEXT REPRESENTATION WITH CONTRASTIVE LEARNING

This research explores how contrastive learning techniques can improve text representation models, demonstrating superior performance in various NLP benchmarks.

HIERARCHICAL TRANSFORMERS

The paper presents a novel hierarchical transformer model designed for long-context understanding, offering efficiency gains over traditional transformers in processing extensive text or data.

LAION-SG: AN ENHANCED LARGE-SCALE DATASET

This work improves upon LAION Aesthetics dataset by adding structured prompting information. This can be used as a dataset for improved training of multimodal generative models.

BROWSERGYM

An open toolkit for accelerating browser based agentic research. It includes a unified interface and a number of important tasks. It also provides a way to screenshot the output from the browser.

LIGHTWEIGHT FINE-TUNING FOR MULTILINGUAL TASKS

A framework that simplifies fine-tuning for multilingual NLP models, enabling faster and more efficient adaptation across languages.

GPT-POWERED DEVELOPMENT FRAMEWORK

GPD is a new framework leveraging GPT models to streamline software development tasks, including code generation and debugging, with a focus on user-friendly workflows.

FIRST IMPRESSIONS OF THE NEW AMAZON NOVA LLMS

Amazon unveiled the Nova family of LLMs at AWS re:Invent, featuring competitive pricing and multimodal capabilities like handling images, video, and PDFs. The Nova series, particularly Nova Micro, is extremely cost-effective, undercutting Google's Gemini models while offering large context handling. With Nova's multimodal advancements, Amazon positions itself as a strong competitor in the AI landscape.

GOOGLE'S MODEL OUTPERFORMS TOP WEATHER FORECAST SYSTEMS

The new GenCast AI model advances weather forecasting by using high-resolution ensemble predictions to provide faster and more accurate forecasts up to 15 days ahead.

GOOGLE AI SHARES TIPS FOR SMARTER TOOLS

Google shares practical tips and best practices for integrating AI into daily workflows.