OpenAI Under Investigation

PLUS: Experts Advocate Legal 'Safe Harbor' for AI Evaluation, DNA Chip Boosts AI Efficiency and more.

Today:

OpenAI is INVESTIGATED by Law Firm. The TRUTH about the Board's Firing of Sam Altman is Revealed…

The recent uproar surrounding Mira Murati allegedly orchestrating the coup to oust Sam Altman has stirred controversy. Despite media claims, Murati vehemently denies involvement, citing a productive partnership with Alman. The situation escalated when anonymous tips, likely from former board members, circulated, prompting media coverage. 

However, a comprehensive audit by WilmerHale concluded in Altman and Greg Brockman's favor, reaffirming their leadership roles in OpenAI. Amidst the drama, OpenAI introduces reforms, including a whistleblower hotline. Despite setbacks, the organization remains focused on its mission of advancing artificial general intelligence.

OpenAI announces new board members, reinstates CEO Sam Altman

OpenAI reinstates CEO Sam Altman to its board alongside three new members, addressing prior criticisms of a lack of diversity. The new appointees, including former Gates Foundation CEO Sue Desmond-Hellmann, aim to bring varied expertise. An investigation concluded Altman's removal stemmed from a breakdown in trust, not concerns over product safety or finances. 

Despite this, reports suggest Altman's manipulative leadership style raised internal concerns. OpenAI pledges governance reforms, including a conflict of interest policy and whistleblower hotline. The move aims to strengthen leadership for the ethical advancement of transformative technologies.

Nvidia is sued by authors over AI use of copyrighted works

Nvidia faces a lawsuit from three authors for allegedly using their books without permission to train its AI platform, NeMo. The authors claim Nvidia infringed their copyrights by including their works in a dataset of about 196,640 books used for training NeMo. They seek unspecified damages for US individuals whose works contributed to NeMo's training. 

Nvidia declined to comment. The lawsuit adds to a trend of writers suing over generative AI. Nvidia's stock has soared amid AI's popularity. The case reflects ongoing legal challenges in the evolving landscape of AI technology.

The need for legal "safe harbor" protections has been emphasized by AI experts, academics, and creatives in a recent paper. This safeguard is crucial for enabling researchers, journalists, and artists to conduct good-faith evaluations of AI tools without fear of legal repercussions. Current terms of service for popular AI models often prohibit such evaluations, hindering progress in AI safety and trustworthiness research. 

The authors urge tech companies to provide indemnification for public interest AI research and safeguard it from account suspensions or legal actions. Transparency and dialogue between researchers and companies are key to addressing these challenges and advancing responsible AI development.

New DNA-infused computer chip can perform calculations and make future AI models far more efficient

A new biocomputing chip utilizing DNA for calculations could revolutionize AI training and big data processing. Developed by scientists, the chip processes data using DNA strands, encoding binary information and conducting calculations through microfluidic channels. Unlike conventional methods, DNA-based processing integrates storage and computing, offering efficiency and parallelism. 

Although still a proof-of-concept, scaling this technology could enhance AI systems' efficiency and environmental sustainability. This innovation marks a significant step in bridging biology and data computation, potentially transforming future computing landscapes.

🧠RESEARCH

"Chatbot Arena" is a fresh open platform shaking up how we gauge large language models (LLMs) by tapping into what people dig. It racks up opinions through a head-to-head voting system, already scoring over 240K votes. This paper spills the beans on its setup, the smart stats tricks for fair model scoring, and confirms it's legit with both Joe Public and experts nodding in agreement. It's now a go-to leaderboard, getting props from big names in the LLM scene. 

Recent research challenges the notion that only large-scale language models excel in math tasks. The LLaMA-2 7B model demonstrates robust math abilities without extensive pre-training. Scaling up synthetic data boosts accuracy, overcoming limitations of scarce real math questions. This approach achieves significant improvements, shedding light on scaling behaviors across reasoning complexities.

Pix2Gif introduces a novel approach for GIF creation, employing motion-guided diffusion. By integrating text and motion prompts, it translates images into GIFs. A motion-guided warping module ensures fidelity to motion cues, while a perceptual loss maintains content coherence. Trained on curated data from TGIF, it excels in capturing both textual and spatial prompts.

StableDrag enhances point-based image editing by addressing issues in existing dragging techniques like inaccurate point tracking and incomplete motion supervision. It introduces a precise point tracking method and a confidence-based strategy for motion supervision, resulting in more stable and high-quality manipulation. StableDrag-GAN and StableDrag-Diff models demonstrate improved performance on DragBench through rigorous experimentation.

MAGID introduces an automated framework for generating diverse and high-quality multimodal datasets. By augmenting text-only dialogues with images and employing a diffusion model, it ensures alignment between text and generated images. Incorporating feedback loops enhances image quality and multi-modal dialogue generation, outperforming baselines, particularly in small image databases, as demonstrated in evaluations.

🛠️TOP TOOLS

FForward - automating financial processes for businesses, providing AI-driven solutions to streamline invoicing, payments, and financial management. 

Slite - create effective wikis through a simple process involving starting with a specific type, chatting with Slite AI for context, and then customizing the content. 

Soona - offers AI tools and a creative platform for brands to create professional content, including photos and videos. 

ShopThatLook - fashion search engine that allows users to upload a photo of an outfit they like and then finds stores offering similar items. 

InstaText - writing tool designed to help you write like a native English speaker

Clay - Combine 50+ data sources, web scraping, and AI messaging to enrich your data and automate your outbound at scale.

Botika - Create stunning fashion product photos with AI

🗞️MORE NEWS

SORA AI VIDEOS EASILY CONFUSED WITH REAL FOOTAGE IN SURVEY TEST 

Survey data reveals most Americans struggle to tell apart videos made by humans from those created by Sora AI, sparking concerns. Respondents urge government regulation for AI-generated content labeling. The nascent AI industry faces scrutiny amid fears of deepfake potential, particularly in entertainment and during elections. Regulation calls intensify. VARIETY

India plans 10,000-GPU sovereign AI supercomputer

India plans to invest $1.24 billion in a sovereign AI supercomputer with 10,000 GPUs, aiming to boost AI infrastructure. The initiative includes funding for an innovation center, startup financing, datasets platform, and skills program. India aims to achieve self-reliance in technology and democratize AI benefits. Private partnerships are anticipated, yet specifics remain undisclosed amid India's complex relationship with Big Tech. THE REGISTER

Opera's new AI browser to soon get even more features through new AI innovation program

Opera is leveling up its game with Opera One, its AI-powered browser, set to introduce more AI features via its new AI Feature Drops program. This program aims to offer users early access to the latest AI innovations, enhancing web browsing by integrating native AI, Aria, without extra costs or extensions. Aria stands out by offering personalized browsing experiences, leveraging real-time web data for improved productivity and creativity. PR NEWSWIRE

The AI industry is pushing a nuclear power revival — partly to fuel itself

The tech world, fueled by the AI industry's rapid growth, is making a bold push toward reviving nuclear power. This pivot is not just for cleaner energy but also as a vital resource for powering data-intensive AI operations. Companies spearheaded by Silicon Valley's elite, like Sam Altman of OpenAI, are heavily investing in nuclear startups, aiming to address the dual challenge of meeting the massive energy demands of AI technologies while advancing toward a greener, more sustainable energy future. Nuclear power, with its promise of abundant, low-carbon energy, appears as a compelling solution to the tech industry's escalating power needs, despite the slow pace and rigorous regulatory environment that characterizes nuclear energy projects. NBC NEWS

Signos uses a glucose monitor patch and AI to help you eat healthier. Here’s what it’s like

Signos is revolutionizing how people approach weight loss and health with a blend of continuous glucose monitoring (CGM) technology and artificial intelligence. By tracking how your body responds to food, exercise, stress, and more, Signos provides personalized insights to help manage your blood sugar levels effectively. While the cost may be a barrier for some, and there's a risk of becoming overly fixated on data, the potential benefits for understanding and improving metabolic health are significant. CNBC

AI fraud detection software maker Inscribe.ai lays off 40% of staff

Inscribe.ai, an AI-powered fraud detection software provider, has recently laid off approximately 40% of its workforce, which amounts to dozens of employees. This decision was recommended by the company's board due to Inscribe missing its revenue goals for over a year amidst market challenges. The layoffs come as the company navigates changes in the financial services industry and advances in AI technology. TECHCRUNCH

Welcome to the Valley of the Creepy AI Dolls

Hyodol unveils a $1,800 AI-enabled doll at Mobile World Congress, aiming to combat loneliness among older adults. Despite its eerie appearance, the doll can hold conversations and provide health reminders. The rise of such devices raises concerns about privacy, security, and emotional attachment. Societal acceptance and ethical considerations are still evolving. WIRED

Mapping Disease Trajectories from Birth to Death with AI

This groundbreaking study, using AI to map disease trajectories from birth to death, analyzed over 44 million hospital stays in Austria. Researchers uncovered 1,260 distinct disease paths, shedding light on critical points where targeted prevention could drastically alter health outcomes. This work highlights the importance of early, personalized interventions to manage multimorbidity and could significantly relieve healthcare burdens while enhancing life quality. With the global population aging rapidly, such insights are vital for healthcare planning and individual health management, offering a new approach to understanding and addressing complex health issues over a lifetime. NEUROSCIENCE NEWS

Beyond human intelligence: Claude 3.0 and the quest for AGI

Anthropic's Claude 3.0 chatbot, a leap from Claude 2.0, sets new AI standards, challenging GPT-4 dominance. Released quietly, it offers near-human comprehension and fluency, raising questions on AI ethics and human-AI dynamics. While impressive, doubts linger on achieving true AGI. Progress like this demands ethical alignment and raises anticipation for what's next. VENTUREBEAT

What'd you think of today's edition?

Login or Subscribe to participate in polls.

What are MOST interested in learning about AI?

What stories or resources will be most interesting for you to hear about?

Login or Subscribe to participate in polls.

Join the conversation

or to participate.