Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Google will be scraping and indexing all publicly available online content, including forum posts and social media, to train its AI models. This move raises concerns about privacy and the potential for Google to have access to vast amounts of personal information.
Some of the biggest companies in Europe have taken collective action to criticize the European Union’s recently approved artificial intelligence regulations, claiming that the Artificial Intelligence Act is ineffective and could negatively impact competition.
Shared more for the insightful take on Casetext, the overnight success that took a decade to build, this article gives some perspective on how transformative AI can be and how when the opportunity arises, you have to seize it.
State space models have made a splash recently. They arose as a potentially useful model alternative by sidestepping some challenging aspects of the Transformer while maintaining performance. In this work, the Hyena model is applied to DNA sequence modeling and improves in 23 tasks.
CLIPA-v2 improves the efficiency of image-text matching models like CLIP by using shorter sequences, achieving a zero-shot ImageNet accuracy of 81.1% for just $10,000.
This work allows you to take a single image and a pre-trained generative model to produce a panorama or multiple views of the image.
GPT-Migrate enables you to easily migrate your codebase from one framework or language to another.
This study introduces 'Volume Fusion' (VF), a new way to train 3D medical imaging models without needing large amounts of pre-labelled data. The research showed that VF, combined with a new network structure, improved the accuracy of medical image interpretation for different body parts.
This research presents a method called Filtered Guided Diffusion (FGD) that enhances Image-to-Image (I2I) translation without requiring extra training or changes to the underlying model. FGD adjusts the diffusion process based on previous steps and offers significant improvements in structural guidance and performance.
While we are all waiting for our 22k H100s to be delivered, we should probably practice training large models. Sasha Rush has another set of puzzles, but instead of GPUs or tensor shapes, he walks us through model parallelism, gradient accumulation, and more!
The article discusses the rise of AI engineering as a specialized field and the skills required to thrive in this domain, highlighting the need for a strong understanding of machine learning algorithms, data manipulation, and programming languages, as well as the ability to bridge the gap between research and implementation in order to create practical AI solutions.
It is well known that Europe has stronger privacy requirements for many digital applications. This includes AI and language models. Unfortunately, many of the model providers today don't fully comply with these privacy requirements. This post explores where they fall short and which ones are the closest to full compliance.
Matt Rickard questions whether the symbiotic relationship between AI companies and hardware providers is a positive one.
The Walrus examines the rise of AI religion and explores the concept of artificial general intelligence (AGI) as a potential deity. It discusses the beliefs and practices of those who see AGI as a religious force, the ethical implications of AI worship, and the motivations behind this emerging phenomenon, highlighting the role of technology in shaping our spiritual experiences and challenging traditional notions of religion.
This article is about the author's experience using AI tools, which were often dissatisfying or simply did not live up to the hype they were given, and reflects on the few AI tools the author found most unpleasant.
Sponsored
Storybird.ai
Inflection, a one-year-old startup founded by ex DeepMind co-founder and head of applied AI, has raised a significant amount of money. The company offers a personalized AI assistant called Pi. This announcement comes on the heels of the recent MLPerf results from Inflection’s newly built Nvidia + CoreWeave supercomputer. Most of the money raised will go to compute, with 22k H100s announced.
Rep. Ted Lieu has criticized new House guidelines restricting AI tool ChatGPT's use to research only, advocating for each office's autonomy to use it as they see fit. These advisory guidelines arrive as lawmakers focus on formulating AI regulation.
a16z lays out the future of computing and the paradigm shift from a tool to a companion. The article walks through a bunch of potential interesting angles for what’s next with AI.
This delightful article discusses two novel concepts: integration and coverage. These refer to two axes of capabilities in language models - generalization and memorization. The idea posed is that both are useful for tasks and it is hard to tell which is being used for those tasks. Additionally, small models might be good on benchmarks because of increased coverage at the expense of true integration.
Instruction tuning is good at getting models to do what we want. People misunderstand the importance of FLAN because it gives short responses and humans typically prefer long responses. This article also shows other models and when you should consider using them.
This study introduces ChatLaw, a new open-source language model designed specifically for the Chinese legal language, aiming to assist the digital transformation of the field.
Audiocraft presents MusicGen, a model that simplifies and enhances music generation. This single-stage, auto-regressive Transformer model doesn't need self-supervised semantic representation like prior methods and it generates music more efficiently by predicting all codebooks in parallel, resulting in fewer steps per audio second.
XrayGPT is a pioneering AI model that can interpret chest X-rays and answer questions about them in a conversational manner. It merges visual understanding with textual knowledge and is enhanced by summaries of real-world radiology reports. XrayGPT offers new possibilities for automated X-ray analysis.
Tool for chatting with your codebase and docs using OpenAI, LlamaCpp, and GPT-4-All.
Can we get models to take images as output without retraining them? Amazingly, if you use small, image-specific modules, you can provide context to a fully frozen language model that allows it to answer questions about images. Paper, code, and demo is included.
The rise of generative AI demands startups to increase their pace of innovation. This piece suggests a mindset shift, including detaching ego from product, seeing operations as experiments, and overcoming self-sabotage, to fully capitalize on the opportunities AI presents.
An interesting case study on how to build an LLM-powered generative application.
Casemark’s chrome extension allows legal professionals to summarize and create documents, perform legal research, and improve general workflow management all in the browser. They are working now to build LLMs that run in secure environments.
Microsoft today announced a slew of new AI-powered shopping tools for its new Bing search engine and the Bing AI chatbot in the Edge sidebar.
Stability AI, the artificial intelligence startup that has been struggling to raise funding at a $4 billion valuation, has secured a $25M convertible note from investors.
Sponsored
You don’t need complicated software to access custom charts, data and automation — just ask Pluto
MosaicML has one of the industry leading toolkits for training large AI models while Databricks has the customer's data. The combination of the two seems sensible. Congrats to the team!
Google Colab is an online, VM-backed, Jupyter Notebook interface. It has continually been the entry point for many programmers and AI developers. Google is now rolling out its AI coding assistant to Colab for Pro+ subscribers.
Stability AI, the closely watched artificial intelligence startup, has lost at least two top executives in recent weeks, including its Head of Research and Chief Operating Officer. David Ha, Head of Research for the London-based startup, resigned this month. Chief Operating Officer Ren Ito left in June.
Researchers have developed "SHGNN", a tool that smartly organizes city maps, accounting for significant variations between different locations. This new system effectively groups nearby similar locations, resulting in a more accurate representation of urban diversity.
"PromptIR" is a new method that uses prompts to restore high-quality images from degraded versions, regardless of the type or severity of degradation. This advanced technique adapts to a variety of image issues like noise, blur, and haze.
An extensible, convenient, and efficient toolbox for fine tuning large machine learning models. Designed to be user-friendly, speedy, reliable, and accessible to the entire community.
DragGAN allows you to edit photos by dragging parts of the image. Its development team has released the code and a live HuggingFace demo.
Kor enables you to extract structured data from text using LLMs.
Giskard is an open-source testing framework dedicated to ML models, from tabular models to LLMs.
Sponsored
Done tinkering? Build and ship robust AI apps with Vellum
A survey of AI experts from 2016 on their predictions for the future of the industry.
OpenAI plans to market a “personalized assistant for work,” a move that would seemingly put the buzzy startup in competition with its chief backer, Microsoft.
A poll of 2,000 people about how they’re using AI, what they want it to do, and what scares them about it the most.
Microsoft's Jared Spataro believes that AI will change the nature of work, including meetings and creative collaboration. He argues that we need to adopt a new approach to work and management in order to take advantage of the potential of AI. This includes using AI to automate tasks, personalize experiences, and make better decisions.
Generative AI models are changing the economy of the web, making it cheaper to generate lower-quality content that is flooding the web.
The age old question around if you should switch to AI from crypto. The real magic would be regulatory clarity for blockchain and AI - that will be the big winner.
Sponsored