🔨 AI Powered Tools 🔨 

A run down of the latest SaaS products and services which leverage AI to help you take back some time.

• Wondershare - for video content generators. Diverse array of solutions for content creation, diagramming, prototyping, file repair, data recovery, and more

• Artisse - for image editors. Easily turn your everyday images re-imagined photos

• Reiki - for AI enthusiasts. Agent creation and monetization platform

📰 News 📰 

Brilliant Labs Introduces $349 Smart Glasses with AI Superpowers

Brilliant Labs' latest innovation, Frame smart glasses, priced around $349, melds style with cutting-edge AI to redefine smart eyewear. With capabilities like visual analysis, real-time translation, and web searches, these glasses are a leap towards interactive learning and digital assistance. Integrated with the Noa app, Frame is not just a gadget but a personalized learning companion, set to ship in mid-April. This venture into open-source, AI-powered eyewear promises a unique blend of technology and personalization.

Sources: The Verge, TechRadar

Sam Altman’s Lofty Ambition Raises Eyebrows

Sam Altman, CEO of OpenAI, is spearheading an ambitious initiative to secure $5-$7 trillion funding aimed at boosting global AI chip manufacturing capabilities. This venture seeks to mitigate the current GPU shortage by establishing chip foundries in collaboration with key investors, such as the UAE, and industry giants like SoftBank and TSMC. The project, while showcasing the potential for significant advancements in semiconductor production, also brings to light concerns over geopolitical implications and the practicality of creating a saturated market amidst fluctuating demand cycles.

Read More At: Ars Technica, The Register

Google’s Gemini Assistant: A Glimpse of the AI Future

Google's Gemini, evolving from Bard, showcases a significant stride towards integrating AI into daily tasks through its services ecosystem. While Gemini excels in tasks like summarizing emails and creating drafts by leveraging Google's suite, including Gmail and Docs, it faces challenges in context understanding and lacks calendar functionalities. Despite these hurdles, Gemini's introduction of AI tools, including a powerful language model and subscription options, marks a pivotal move towards more sophisticated voice and chat assistants, potentially setting the stage for the future of digital assistance.

Read More At: The Verge, TechRadar

🧠 Research 🧠 

The Unlocking Spell on Base LLMs: Rethinking Alignment via In-Context Learning

This research investigates whether large language models can self-align to human intentions without explicit tuning, through a novel approach named URIAL. It discovers that alignment predominantly influences style rather than the core knowledge, suggesting that the essential knowledge for response accuracy is innately present in LLMs. This work challenges existing alignment paradigms, advocating for more resource-efficient utilization of base LLM capabilities.

For the detailed paper, visit ArXiv.

More Agents Is All You Need

This study introduces a straightforward yet potent approach to boost large language models' (LLMs) performance by leveraging the scaling property of agents through a simple sampling-and-voting method. It reveals that increasing the number of agents generally enhances LLM effectiveness across various tasks. The method's efficacy correlates with task difficulty, with notable improvements in complex problem-solving with more agents.

For the detailed paper, please visit ArXiv.

Question Aware Vision Transformer for Multimodal Reasoning

The study presents QA-ViT, a novel Question Aware Vision Transformer approach designed to enhance multimodal reasoning by embedding question awareness directly within the vision encoder, dynamically focusing visual features on image aspects relevant to user queries. The approach is model-agnostic with extensive experiments demonstrating QA-ViT's effectiveness in improving visual and scene-text understanding, outperforming existing methods. Future research aims to seek advancements in AI's ability to understand and interact with complex visual and textual information seamlessly.

For the detailed paper, please visit ArXiv.

💬 Large Language Models 🗨️ 

Exploring NeMo Guardrails: An Open-Source LLM Security Toolkit

NeMo Guardrails, developed by NVIDIA, is an open-source toolkit designed to enhance security in large language models (LLMs). Contrasting with Llama Guard, NeMo Guardrails offers a more comprehensive set of guardrails for LLM-based conversational systems, including content moderation, topic guidance, hallucination prevention, and response shaping. The toolkit provides developers with the flexibility to control and guide LLM inputs and outputs effectively.

Read More At: Towards Data Science which delves into the implementation details of integrating NeMo Guardrails into an RAG pipeline for enhanced security measures.

Meet the Pranksters Behind Goody-2, the World’s ‘Most Responsible’ AI Chatbot

Goody-2, a chatbot designed to take AI safety to an extreme, refuses every request citing potential harm or ethical breaches. Created by artists, it satirizes the overzealous safety measures in AI models like ChatGPT. Despite its absurd responses, Goody-2 raises serious questions about responsible AI development and the challenges of aligning AI with moral values. The project highlights the ongoing safety issues in large language models and the difficulty in achieving neutrality. The chatbot's creators prioritize caution over intelligence, emphasizing safety above all else.

Read More At: Wired

🗞️ Other News Rollup 🗞️ 

Meta's Artemis AI Chip - Meta to deploy Artemis chip to to reduce reliance on Nvidia's H100 chips in data centers, saving costs and boosting efficiency. Despite this, Meta will still use Nvidia's GPUs for now, aiming to save costs and boost efficiency in running AI workloads.

AMD's AI GPU Pricing - AMD sells MI300 AI accelerators to Microsoft at a third less, posing a challenge to Nvidia's pricier H100.

Notepad Getting Copilot AI - Microsoft has introduced Copilot AI into its Notepad application for Windows Insiders in the Canary and Dev Channels on Windows 11. This integration allows users to get explanations for highlighted text by using Ctrl + E or selecting 'Explain with Copilot.'

Perplexity & Vercel partner on AI Search for Devs - Perplexity AI has partnered with web development platform Vercel to allow developers to integrate Perplexity's large language models into their applications. Enhancing knowledge support provides real-time access to online LLMs.

Amazon Bedrock - a managed service offering high-performing foundation models for generative AI applications. The service allows IT teams to provide access to models while ensuring centralized governance, cost tracking, and usage controls.

California AI Regulation - Proposed legislation aiming to establish safety standards for large-scale AI systems. The proposed law includes building a public AI research cluster, CalCompute, to align AI systems with California's values.

AI Boosts ANZ Bank Productivity - GitHub Copilot enhanced productivity and code quality at ANZ Bank, particularly benefiting expert Python programmers.

🎶 Prompts 🎶 

Prompts used to generate some of this issues images. Unless otherwise stated all images are using Dall-E and ChatGPT Plus.

That’s all for today.

So what did you think?