In partnership with

Today:

My Favorite Newsletter: Stay ahead on the business of AI 

Have you heard of Prompts Daily newsletter? I recently came across it and absolutely love it.

AI news, insights, tools and workflows. If you want to keep up with the business of AI, you need to be subscribed to the newsletter (it’s free).

Read by executives from industry-leading companies like Google, Hubspot, Meta, and more.

Want to receive daily intel on the latest in business/AI?

In just ONE click, you can quickly sign up to the free Prompts Daily newsletter.

Billionaire Jeff Bezos and Cathie Wood Joined Microsoft, Nvidia, Amazon, and OpenAI in Funding 1 of 2024's Most Intriguing AI Start-Ups

In the bustling world of AI, big names like Jeff Bezos and Cathie Wood are backing Figure AI, a start-up aiming to revolutionize robotics. Led by Brett Adcock, Figure AI snagged $675 million in funding, drawing support from giants like Microsoft, Nvidia, and Amazon. 

The start-up's humanoid robot, Figure 01, targets labor shortages in sectors like warehousing and transportation. Although investing directly might be tricky, buying shares in companies like Microsoft offers a slice of the AI action. With tech titans betting big, Figure AI is poised to reshape the future of automation.

YAHOO!

Gemini now lets you tune and modify responses with a prompt

Google is enhancing Gemini's response tuning feature on the web app, allowing users to modify selected text with options like Regenerate, Shorter, Longer, and Remove. Users can adjust prompts but are limited by unsupported changes or policy violations. 

This feature aims to provide output closer to users' needs, giving more creative control and iterating on ideas. It complements existing features like seeing other drafts and adjusting tone. Rolling out in English initially, this update enhances Gemini's utility for users, aligning with Google's goal of refining its AI capabilities.

9TO5GOOGLE

Meta is building a giant AI model to power its ‘entire video ecosystem,’ exec says

Meta, formerly Facebook, is investing heavily in artificial intelligence (AI) to enhance its video recommendation engine. Tom Alison, head of Facebook, revealed plans for a unified AI model to power video recommendations across all Meta platforms by 2026. 

This AI initiative aims to optimize user engagement and relevance by leveraging GPU technology. Meta's phased approach involves transitioning recommendation systems to GPUs and developing a single model for all video content. If successful, this model could also enhance other features like digital assistants and group interactions, marking a significant advancement in AI integration within consumer environments.

CNBC

Salesforce aims to blaze new generative AI trail for developers with Einstein 1 Studio

Salesforce launches Einstein 1 Studio, a suite of tools for developers to customize their AI assistant, Einstein Copilot. This release coincides with the TrailblazerDX conference. Einstein 1 Studio includes Copilot Builder, Prompt Builder, and Model Builder, enabling tailored AI actions, prompts, and integration with various AI models. 

The move aims to empower developers for an AI-driven future. Alongside, Salesforce unveils research indicating growing enterprise demand for AI, with 87% of IT pros affirming AI's effectiveness. The pressure to implement AI swiftly persists, highlighting the need for customizable solutions like Einstein 1 Studio to meet diverse business needs.

VENTUREBEAT

IBM deploying Adobe Firefly generative AI tool for marketing, design teams

IBM is teaming up with Adobe to use Adobe Firefly AI tool, enhancing productivity for its design and marketing teams. Initially experimented with internally, IBM now deploys Firefly across its practitioners, leveraging its ideation and productivity capabilities. With features like style matching, Firefly aids in creating content at scale, facilitating personalized marketing campaigns. 

IBM also assists clients in adopting AI responsibly, ensuring brand consistency and quality. This partnership deepens IBM's collaboration with Adobe, aiming to transform marketing organizations and achieve personalized marketing at scale, a long-standing industry goal.

FOX BUSINESS

Accenture Makes $1B AI Power Play with Udacity Purchase

Accenture is making a bold move, investing $1 billion in acquiring Udacity to bolster its LearnVantage business, focusing on AI-driven training. LearnVantage aims to assist leaders in providing specialized tech, data, and AI training for their workforce. Accenture's CEO highlights the importance of investing in skills development amidst technological change. 

This acquisition adds expertise and scalable learning tech to Accenture's offerings, aiming to meet clients' evolving needs. With a previous successful partnership with Merck, Accenture is poised to further empower companies in their digital transformation journey, particularly in AI training and upskilling efforts.

INFORMATION WEEK

🧠RESEARCH

Design2Code: How Far Are We From Automating Front-End Engineering?

The paper discusses using generative AI for front-end development, specifically converting visual designs to code, a task dubbed Design2Code. It tests various models, including GPT-4V and an open-source Design2Code-18B model, on 484 webpages. GPT-4V outperforms others, accurately generating webpages that often surpass the originals in visual appeal. The study reveals that while models excel at text and color, they struggle with layout and recalling visual elements, though fine-tuning can significantly improve performance.

Scaling Rectified Flow Transformers for High-Resolution Image Synthesis

The paper introduces a new generative model approach for creating high-resolution images from text by improving noise sampling techniques. It presents a transformer-based architecture that separates weights for text and image modalities, allowing better interaction and understanding between the two. This results in enhanced text comprehension, typography, and human preference. Their largest models surpass current leading models in performance, promising to release data, code, and model weights for public use.

NaturalSpeech 3: Zero-Shot Speech Synthesis with Factorized Codec and Diffusion Models

NaturalSpeech 3 innovates text-to-speech (TTS) systems by breaking down speech into separate attributes like content and timbre, then individually generating them for improved quality and naturalness. By using a new type of model that disentangles speech, it achieves superior performance in clarity, likeness to human speech, and natural prosody compared to existing TTS technologies. This method's effectiveness is demonstrated through its exceptional outcomes in voice synthesis, particularly in creating speech that sounds more like a real person talking, with nuances in tone and expression.

Finetuned Multimodal Language Models Are High-Quality Image-Text Data Filters

The paper proposes a new method for filtering image-text data using fine-tuned Multimodal Language Models (MLMs). It outperforms previous methods like CLIPScore by integrating advancements in MLMs and establishing a new pipeline for creating high-quality instruction data. This approach generates more accurate and comprehensive scores for image-text data, enhancing the performance of pre-trained models on various tasks. The findings show significant improvements over CLIPScore, demonstrating the MLM filter's versatility and potential as a superior alternative for data quality assessment.

MathScale: Scaling Instruction Tuning for Mathematical Reasoning

MathScale introduces a scalable method for enhancing large language models' (LLMs) skills in solving math problems. It creates a vast dataset, MathScaleQA, with 2 million question-answer pairs by generating new questions from seed math questions using a concept graph, mirroring human learning mechanisms. Applied to open-source LLMs, MathScale significantly boosts mathematical reasoning, showcasing its effectiveness across various math problem levels. It sets new benchmarks, outperforming similar-sized models in accuracy and offering a promising approach to improving LLMs' capabilities in math reasoning.

🛠️TOP TOOLS

Reka -  a turbo-class, efficient multimodal language model that outdoes bigger models with its quick, high-quality performance.

3Daily - offers an AI-generated platform for creating high-quality 3D models tailored for the gaming industry, combining AI's efficiency with artists' creativity. 

HeadshotPro - AI headshot generator that creates professional business headshots without a physical photoshoot. 

Seoby -  free AI SEO tools aimed at helping websites rank faster on Google. 

Capeprivacy - Enterprise AI Platform that deploys directly into your infrastructure, optimized for both the security and privacy of your data

Typeset - AI-powered visual content design tool that streamlines the creation of presentations, documents, and social media content

Reclaimai - AI scheduling for the perfect workweek

Retool - Form to database, instantly

GoEnhance - Advanced Imaging and Video Editing

📲SOCIAL MEDIA

🗞️MORE NEWS

Microsoft AI engineer warns FTC about Copilot Designer safety concerns

A Microsoft AI engineer is sounding the alarm about the company's AI image generator, Copilot Designer, to the Federal Trade Commission. Shane Jones, who's been with Microsoft for six years, warned that the tool can create harmful images despite repeated alerts to Microsoft. Testing revealed disturbing scenes like demons, sexualized violence, and underage substance use. Disney characters were also depicted in inappropriate contexts, like Elsa from Frozen in war zones. Despite warnings since December, Microsoft hasn't taken adequate action. THE VERGE

Researchers tested leading AI models for copyright infringement using popular books, and GPT-4 performed worst

Patronus AI, led by ex-Meta researchers, found OpenAI's GPT-4 generated the most copyrighted text, with 44% of responses drawn from copyrighted books. Testing against popular titles, including Michelle Obama's "Becoming," showed widespread infringement across models. OpenAI's stance on training AI models with copyrighted material faces scrutiny amidst legal battles. CNBC

Google engineer indicted over allegedly stealing AI trade secrets for China

A Google engineer, Linwei Ding, got indicted for allegedly swiping secrets on Google’s AI chips for China. He allegedly filched over 500 confidential files before his arrest. These files detailed designs for Google's AI chips and data centers. The engineer faces up to ten years in prison and hefty fines if convicted. THE VERGE

AI start-up Lore Machine in funding talks, rolls out visual-storytelling platform

Lore Machine, an AI startup, unveils a visual storytelling platform converting stories into images, audio, and animation. It aims to charge $10 to $160 monthly, targeting enterprises with premium services. Seeking funds, the company hints at profitability and revenue from partnerships. Lore Machine sees potential in marketing and eyes partnerships with software platforms. REUTERS

Box posts first $1 billion fiscal year as AI sales pick up

Box Inc. celebrates its first $1 billion fiscal year, driving a 2% after-hours stock rise. CEO Aaron Levie attributes success to Box AI adoption, with financial service clients leading the charge. The company also integrates with Microsoft Azure OpenAI, reinforcing its AI capabilities for the Box Content Cloud. MARKETWATCH

Google is about to purge the AI spam littering its search Results

Google announces a crackdown on AI-generated content aimed at gaming search rankings. Changes to its algorithm will target automated content, including that produced by generative AI, to prioritize quality and user experience. The move aims to reduce low-quality content by 40% and combat tactics like using expired domains for spam. QUARTZ

How Amazon’s latest Climate Pledge Fund investment is transforming recycling through AI and robotics

Amazon's Climate Pledge Fund invests in Glacier, an AI and robotics company targeting recycling. Led by Rebecca Hu and Areeb Malik, Glacier uses AI-powered robots to improve recycling efficiency. The investment aligns with Amazon's commitment to support female-led climate tech companies and reduce waste. AMAZON

US Army tests AI chatbots as battle planners in a war game simulation

The US Army tests AI chatbots in simulated war games to enhance battle planning. OpenAI's technology is under scrutiny, but experts warn against real-world deployment due to potential risks. Military research explores AI's role beyond image analysis and target identification, focusing on generative AI's capabilities in simulated scenarios. NEWSCIENTIST

Overjet Raises $53 Million: The Largest Investment in the History of Dental AI

Overjet secures $53 million in Series C funding, marking the largest investment in dental AI to date. The influx will expand its AI platform, aiming to standardize dental care delivery for providers and payers, enhancing patient outcomes. Overjet's FDA-cleared technology aids in precise diagnoses and treatment recommendations, fostering industry-wide efficiency and quality. PR NEWSWIRE

AI won't replace your job — but the person who knows how to use it might

In the evolving workplace, AI integration is inevitable. Deloitte's survey reveals a shift in focus towards upskilling workers for AI adoption. Executives prioritize training, acquisition of AI-ready companies, and hiring for AI-centric roles. They emphasize AI literacy and ethical AI use. The key lies not in fearing AI but in mastering its utilization for job enhancement. QUARTZ