📰 News 📰 

AI Marilyn Monroe: Digital Resurrection Without Consent

Soul Machines introduced "Digital Marilyn," an AI chatbot resembling Marilyn Monroe, sparking debate on ethical issues surrounding the use of celebrities' digital likenesses without consent. This AI, debuted at SXSW and created in partnership with Authentic Brands Group (which holds rights to Monroe's image), showcases "emotions and nuanced expressions" and can engage in lengthy conversations. Unlike their other AI personalities who are based on living individuals capable of consent, Monroe's posthumous representation raises questions about respect and autonomy in digital afterlives​

Read More At: Engadget

Breakthrough in AI-Driven Drug Discovery

Insilico Medicine, a biotech startup, announced a significant milestone with the publication of a paper detailing the first AI-generated and AI-discovered drug, INS018_055, now in Phase II clinical trials, aimed at treating idiopathic pulmonary fibrosis. This milestone is significant for using generative AI across the entire drug discovery process, from identifying drug targets with their PandaOmics tool to designing drug compounds via the Chemistry42 engine. Founder Alex Zhavoronkov highlights this achievement as a proof-of-concept for their Pharma.AI platform and as a precedent for generative AI's potential in accelerating drug discovery. Remarkably, the AI-driven approach enabled reaching clinical trials in just two and a half years, drastically reducing time and cost compared to traditional methods.

Read More At: VentureBeat

AI's New Frontier: Understanding Life

Groundbreaking research where AI models, after being fed with extensive biological data, made significant discoveries, such as identifying new cell types like the Norn cells in mere weeks, a process that previously took humans over a century. These AI systems, trained similarly to ChatGPT but on biological data, are beginning to unravel complex biological mysteries, hinting at future capabilities that could redefine our understanding of life and disease 

Read More At: NYT

AI Mishaps Surge, New Tracking Database Emerges

The AI Incident Database, a new initiative by the Partnership On AI turned non-profit, is tracking rising AI failures, cataloging over 600 unique incidents to date. It aims to improve transparency between AI developers and the public, inspired by similar models in cyber security and transportation safety. Trends show generative AI accounting for about half of the incidents, highlighting serious societal impacts beyond mere software bugs. The platform, supported by Underwriters Laboratories, seeks volunteers and aims to educate and improve AI safety measures.

Read More At: The Register

Revolutionizing AI with Grayskull

Tenstorrent, led by Jim Keller, has released its first hardware, Grayskull, a RISC-V alternative to traditional GPUs designed for enhanced programmability and scalability in AI tasks. The company also launched two Grayskull-powered DevKits, the Grayskull e75 and the e150, priced at $599 and $799 respectivly, these devices are set to democratize AI development with their advanced features and affordability.

Read More At: TechRadar

💬 Large Language Models 🗨️ 

Inflection AI Elevates Chatbot Capabilities

Inflection AI, co-founded by DeepMind's Mustafa Suleyman and LinkedIn's Reid Hoffman, has introduced Inflection-2.5. This new model, powering their Pi assistant, rivals OpenAI's GPT-4, showing exceptional performance in STEM fields. Inflection-2.5 is notable for its empathetic interactions and efficient training, achieving 94% of GPT-4's performance with less computational effort. It also features real-time web search capabilities. This launch signifies a substantial advancement in making AI more accessible and tailored to individual needs. The Pi chatbot demonstrates significant user engagement, marking a new era in conversational AI.

Read More At: VentureBeat

🧠 Research 🧠 

Open Foundation Models by 01.AI

The authors introduce the Yi model family, encompassing language and multimodal models based on pretrained models extended into various domains such as chat, long context, and vision-language models. Key findings include strong performance on benchmarks like MMLU, and human preference rates in AlpacaEval and Chatbot Arena. The paper highlights the importance of high-quality data engineering in model performance. Limitations include lesser focus on mathematical and coding tasks. Future research could explore further scaling and optimizing data for enhanced performance in specialized domains. The paper emphasizes data quality over quantity and continual pretraining for capability extension.

Read the paper: ArXiv

PixArt-Σ: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation

The paper presents the PixArt-Σ model, advancing the PixArt-α with enhanced 4K image generation through weak-to-strong training. Key improvements include high-quality training data and efficient token compression, enabling significant advancements in image quality and alignment with text prompts at lower computational costs. The paper addresses limitations in prior models with less focus on high resolution and efficiency. Future research could further refine data quality and model scalability to push boundaries in text-to-image synthesis.

Read the paper: ArXiv

🔨 AI Powered Tools 🔨 

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

• Dora AI - With Dora AI, you can generate, customize, and deploy websites with natural language — all within Dora's powerful no-code editor.

• Dubbing AI - Change your voice to any character you want in real-time in gaming, streaming, or meetings, while retaining your emotions, intonations, and delivery. 

• Rask AI - effortlessly translate, dub, and repurpose your videos into engaging Shorts and tweets, reaching a global audience and connecting with diverse cultures and languages.

🗞️ Other News Rollup 🗞️ 

Tech Conference Protest - Google engineer fired for protesting technology supporting genocide at Israeli tech event in New York.

AI at the Edge - Lenovo pioneers AI solutions at the edge, focusing on public safety, energy efficiency, and transformative applications.

Biden's AI Ban - Biden pushes for AI voice impersonation ban, emphasizes US semiconductor industry boost and US-China competition.

Altman's Board Return - Fired CEO Altman returns to OpenAI's board after investigation. New female executives join. Governance improvements planned.

Pixel 8 AI Issue - Gemini Nano AI model incompatible with Pixel 8 due to hardware limitations, raising concerns about Google's communication.

AI-Generated Nude Scandal - Florida teens arrested for creating deepfake nude images of classmates, exposing gaps in laws on AI-generated content.

Pixel 8 AI Limitation - Pixel 8 misses out on Nano AI model due to hardware limitations. Google plans more AI features next year.

Cultural Bias in LLMs - Georgia Tech study finds Western cultural bias in large language models, impacting non-Western users. CAMeL benchmark introduced.

AI Peripheral Vision - MIT researchers create dataset to simulate peripheral vision in AI models, improving object detection and recognition.

🎶 Prompts 🎶 

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

Thank You for Subscribing

Enjoying what you’re reading? Help me get better so I can continue to provide you with the most relevant content.