Data breaches are in the news. But even if your password is safe, your personal data could be at risk!
Data brokers are profiting from your home address, SSN, phone number, and other personal information - and they're selling it to scammers, spammers, and the highest bidder.
Meta's new AI Studio tool will soon allow users without technical skills to create personalized AI chatbots for Instagram, Messenger, and WhatsApp. The tool will enable customized interactions with followers and full control over auto-replies.
OpenAI is launching a new AI voice chatbot powered by the GPT-4o model with features that include natural, fluent conversation capabilities. It is initially available to select ChatGPT Plus subscribers.
Perplexity will start paying publishers a revenue share from ad sales when their content is used to form AI search answers, marking a new business model for the AI search engine, which will allow brands to buy "related follow-up questions" appearing below user query answers. Publishers will earn a percentage of ad revenue and gain free access to Perplexity's large language models and Pro service tier.
This paper introduces Theia, a vision foundation model for robot learning that integrates multiple existing vision models. Theia offers rich visual representations, enhancing robot learning with less training data and smaller model sizes. Experiments show that Theia outperforms its predecessors, and the authors suggest that higher entropy in feature norms improves performance. The code and models are available for public use.
This paper introduces SaulLM-54B and SaulLM-141B, two large language models designed for the legal sector that feature 54 billion and 141 billion parameters. Using the Mixtral architecture, the researchers employed large-scale domain adaptation through continued pre-training with a vast legal corpus, a specialized legal instruction-following protocol, aligning outputs with human legal interpretations. The models outperform previous open-source models on LegalBench-Instruct and offer state-of-the-art performance. Base, instruct, and aligned versions of these models have been released under the MIT License for reuse and collaborative research.
LightGODE is a new approach to improve the efficiency and scalability of recommender systems. It bypasses intensive computations during training by focusing on post-training graph convolution and utilizing a continuous graph ODE.
Researchers have developed a wavelet-based feature enhancement network to improve face super-resolution. The method enhances facial details without introducing distortions by decomposing input features into high and low-frequency components and using a full domain Transformer.
DGFNet is a new approach for predicting the trajectories of multiple traffic participants in autonomous driving. It enhances predictions by considering the difficulty differences among agents, capturing rich spatio-temporal features, and using a difficulty-guided decoder.
Today, Writer welcomed Palmyra-Fin to their family of enterprise LLMs. The model brings unparalleled accuracy to financial generative AI applications — outperforming leading frontier models in a head-to-head test. In a first for finance, Palmyra-Fin even passed the CFA Level III exam.
Palmyra-Fin is now live on the Writer full-stack generative AI platform — allowing developers to build with their APIs, open-source Python framework, or no-code builder.
Meta has open-sourced Llama 3.1, an LLM rivaling OpenAI's ChatGPT, allowing almost all companies to incorporate it freely. This strategy may commoditize LLMs and drive demand for corresponding complements like server space. As big tech releases similar models, AI startups might face challenges, with industry giants potentially outpacing smaller competitors in AI advancements.
This article discusses the impact of LLMs on the service economy, particularly on knowledge work sectors such as legal, healthcare, and education. While rapid AI advancements suggest significant automation potential, the adoption by enterprises is slow, with many still in the experimentation phase. Actual deployment is expected to be gradual. Specialized AI startups that leverage LLMs to improve industry-specific workflows will have a competitive advantage in the evolving market.
AI-powered job interviews, which leverage large language models for automation and potential bias reduction, are being increasingly adopted by firms globally. Companies like Meituan and multinationals are using AI avatars to conduct initial interviews, perceiving them as efficient despite concerns over potential embedded biases and transparency in decision-making. Skeptics advocate for clear reporting on AI tools' construction and bias mitigation. Some candidates are using AI to formulate responses, highlighting the evolving competition between AI systems and job applicants.
Microsoft has announced updates to Azure AI, including serverless fine-tuning for PHI-3 models, improvements in PHI-3-MINI performance, and the integration of models like GPT-4o mini and Meta's LLAMA 3.1 into Azure AI.
Sam Altman of OpenAI has presented a strategic imperative for the US and allies to lead a global AI initiative based on democratic values that suggests robust security, infrastructure investment, commercial diplomacy, and collaborative norms development to maintain AI supremacy over authoritarian regimes.
The most important AI, ML, and data science news in a free daily email.