Good morning. It’s Wednesday, August 23rd.

Did you know: You can now fine-tune GPT-3.5 turbo to your own data?

In today’s email:

• Voice & Audio AI Technology

• Gaming, Rendering & Animation AI

• AI Translation & Language Models

• Healthcare & Drug Discovery AI

• Platform & Social Media AI Features

• Legacy & Code Translation

• Robotics & Automation

• AI Ethics, Regulation & Simulation

• AI Business & Market News

• AI Image & Design Technology

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think of this edition by replying to this email, or DM us on Twitter.

Today’s edition is brought to you by:

Billionaires wanted it, but 54,578 everyday investors got it first… and profited

When incredibly rare and valuable assets come up for sale, it's typically the wealthiest people that end up taking home an amazing investment. But not always…

One platform is taking on the billionaires at their own game, buying up and securitizing some of history’s most prized blue-chip artworks for its investors. In just the last few years, its investors have realized annualized net returns of 17.8%, 21.5%, 35% and more from these opportunities.

It's called Masterworks. Their nearly $1 billion collection includes works by greats like Banksy, Picasso, and Basquiat, all of which are collectively owned by everyday investors. When Masterworks sells a painting – like the 15 it's already sold – investors reap their portion of the net proceeds.

It's easy to get started but offerings can sell out in minutes. However, as a trusted partner, AI Breakfast readers can skip the waitlist to join with this exclusive link.

See important disclosures at masterworks.com/cd

Voice & Audio AI Technology

ElevenLabs launches AI voice cloning in 30 languages, offering voice replication for video games and audiobooks. Users input speech fragments to clone voices, with the new Multilingual v2 model supporting 22 additional languages. Despite controversies, the company implements measures to prevent misuse. The platform aims to serve media companies and potentially disrupt audiobook production. While AI narration could save time and money, challenges remain for human voice actors in the evolving landscape of voice technology.

India aims to overcome challenges in rural regions by introducing “conversational” payments through its Universal Payments Interface (UPI). With around 350 million UPI users, spotty internet, and low literacy in rural areas pose hurdles. The Reserve Bank of India’s plan involves AI-based speech recognition for voice-activated mobile payments, enabling users to make verbal transfer instructions on their phones. This move reflects India’s digital shopping trends, particularly among millennials and Generation Z. The technology aims to make digital payments more accessible across the country.

Gaming, Rendering & Animation AI

NVIDIA introduces DLSS 3.5, a breakthrough in AI neural rendering technology for gaming, enhancing ray-traced image quality and incorporating emotion into AI-powered non-playable characters (NPCs). The addition of Ray Reconstruction improves lighting effects and realism, while Reflex reduces system latency for more responsive gameplay. NVIDIA’s advancements also include AI-powered NPCs with enhanced emotive interactions and the deployment of RTX 4080-class servers in Ge Force NOW.

Chinese animation house puts AI capabilities to the test: Base Media, renowned for contributing to Hollywood hits like “Black Panther: Wakanda Forever” and Obi: Wan Kenobi,” collaborates with iQiyi's Wonderworks to harness generative AI potential. The Beijing animation firm addresses the resource-intensive task of generating 3D locations in animation production. iQiyi aims to leverage AI for the efficient creation of intricate settings, reducing costs and labor. Base Media’s exploration aligns with its mission to maintain industry leadership by integrating AI solutions into its animation production process, streamlining the creation of complex scenes.

AI Translation & Language Models

Introducing SeamlessM4T, a Multimodal AI Model for Speech and Text Translations: Meta introduces SeamlessM4T, an all-in-one multilingual multimodal AI translation model. It enables seamless communication through speech and text across languages, supporting speech-to-text, speech-to-speech, text-to-text, and text-to-speech translations for up to 100 languages. The model’s single-system approach improves translation efficiency and quality, bridging language barriers and promoting effective communication. Meta’s commitment to open science is reflected in the public release of SeamlessM4T and its associated metadata.

OpenAI has introduced fine-tuning capabilities for GPT-3.5 Turbo, allowing developers to customize the model for their specific use cases. This update enables better performance such as improved instruction-following, consistent-output formatting, and custom tone. Fine-tuned models can even match or exceed the capabilities of base GPT-4 on certain tasks. Developers can also reduce the prompt size by up to 90%, speeding up API calls. OpenAI ensures the safety of fine-tuning through moderation and GPT-4 powered systems. Pricing includes initial training and usage costs and updated GPT-3 models are available as replacements, including for fine-tuning.

Contextual AI and Google Cloud Partner to Bring Generative AI to the Enterprise: Contextual AI, a company focused on AI for enterprise applications, has entered a strategic partnership with Google Cloud. Contextual AI aims to build fully customizable, privacy-aware AI models for businesses. It will leverage Google Cloud’s infrastructure, including GU VMs and Tensor Processor Units (TPUs), to build and train its large language models (LLMs) for the enterprise. The partnership aims to address challenges in AI such as hallucinations, compliance, data privacy, and more. Contextual AI’s LLMs, known as Contextual Language Models (CLMs), aim to provide accurate and tailored responses based on a company’s data and institutional knowledge.

Healthcare & Drug Discovery AI

Microsoft and Epic expand AI collaboration to accelerate generative AI’s impact in healthcare, addressing the industry’s most pressing needs: Microsoft and Epic are expanding their collaboration to leverage generative AI in healthcare. The partnership aims to address critical needs in the healthcare industry by integrating conversational, ambient, and generative AI technologies into the Epic electronic health record ecosystem. This initiative intends to enhance patient care, streamline workflows, and support health systems globally. The collaboration focuses on deploying solutions that improve clinician productivity, reduce manual processes, and advance medical research using AI-driven insights. With healthcare facing challenges such as staffing shortages and burnout, this collaboration seeks to bring AI-powered solutions to the forefront of healthcare.

Genesis Therapeutics Closes $200M Series B to Advance AI-Discovered Drug Pipeline: Genesis Therapeutics secures $200 million in Series B funding to propel its AI-driven drug discovery efforts. Co-led by an undisclosed U.S. life sciences investor and returning investor Andreessen Horowitz, the oversubscribed round attracted new backers including Fidelity, BlackRock, and NVIDIA’s venture arm. The funding will drive clinical development of AI-enabled programs and expand initiatives targeting underserved patient populations. Genesis’ AI platform, GEMS, identifies unique molecules and properties for effective drug candidate identification, demonstrating the potential of AI in revolutionizing drug discovery. The company is poised to advance clinical candidates and transform therapeutic avenues.

Platform & Social Media AI Features

Subject line story

Meta confirms AI ‘off-switch’ incoming to Facebook, Instagram in Europe: Meta, the parent company of Facebook and Instagram, has confirmed that it will introduce non-personalized content feeds on its platforms in the European Union (EU) ahead of the August 25 deadline for compliance with the Digital Services Act (DSA). The DSA requires larger platforms to allow users to switch off AI-driven “personalization” features that track and display content based on individual user profiles. This move aims to provide users with more control and reduce the risk of addiction, filter bubbles, and manipulations. The new option will allow users to view content in chronological order or based on local popularity. TikTok also announced a similar change earlier this month.

Microsoft’s AI-powered design tool is now widely available in Edge: Microsoft Designer, an AI-powered design tool, is now available to Edge users in the US. The tool can be accessed from Edge’s sidebar, enabling users to generate designs without opening a separate tab or program. While Designer is still in preview, its integration into Edge provides convenience for creating designs and sharing them on platforms like email, Facebook, or Instagram. It’s a move by Microsoft to compete with other AI-powered design tools. The tool’s integration also aligns with updates to Bing Chat in the browser, allowing more actions like organizing tabs and obtaining passwords.

YouTube Sets Its “AI Music Principles” And Partners With Universal Music Group On New AI Incubator: YouTube unveils its AI music principles and launches the YouTube Music AI Incubator in collaboration with Universal Music Group. The principles emphasize responsible AI collaboration with the music industry, protection of copyright holders’ interests, and scaling safeguards for AI-generated content. The program includes artists, songwriters, and producers from UMG to inform YouTube’s generative AI approach to music. The announcement highlights the need for a balance between AI’s potential and human creativity in the music industry. The initiative aims to empower artists and leverage AI for musical expression responsibly and ethically.

Match Group is going steady with AI, appoints Zynga alum to lead AI-focused team: Match Groups, the company behind dating apps like Tinder and OkCupid, is strengthening its AI focus by appointing Zynga alum to lead an AI-focused team. CTO, Will Wu, highlighted the company’s focus on harnessing generative AI solutions to enhance user engagement, individuality expression, safety, and accessibility in dating apps. The company’s newly formed “Art x Science Lab” (ASL) team, aims to develop AI-powered features with Tinder already testing AI photo selection and is exploring the idea of AI-generated dating bios.

Robotics & Automation

Google Researchers Teach Robots New Skills With Just Natural Language: Google AI researchers have introduced a method enabling robots to learn new skills from natural language instructions. Bridging the gap between high-level natural language and low-level robotic control, the approach utilizes reward functions to create an interpretable intermediate representation. By converting natural language instructions into reward functions and optimizing them in real life, the system enables robots to understand and execute complex tasks. This method demonstrated on simulated legged and arm robots, achieved 90% task completion.

Robots Are Now Flying Planes Better Than Any Human: Scientists at the Korea Advanced Institute of Science & Technology (KAIST) have developed a humanoid robot capable of flying planes without major cockpit adjustments. Using external cameras to monitor flight parameters, the robot named “Pibot” can manipulate flight controls like a human pilot, memorize extensive navigation charts, and react to emergencies faster due to AI technology. Pibot can execute error-free flights and respond to in-flight issues. The robot is designed for various applications beyond aviation and is projected to be completed by 2026.

AI Ethics, Regulation & Simulation

Simulation lets AI agents decide whether to cooperate or deceive in nuclear code exchange: A GitHub project introduces an AI-driven simulation to test human interaction complexity in challenging situations. The project involves multiple AI-controlled players making decisions to cooperate or deceive each other regarding pieces of “atomic code.” The simulation’s aim is to analyze real-time AI decision-making in ethical dilemmas and establish a framework for further AI ethics experiments. The project utilizes OpenAI’s ChatGPT API for simulating AI behavior, and developers have created a Discord bot for human participation. The open-source project enables the exploration of AI-agent interactions in various scenarios.

This 'mind-reading' AI system can recreate what your brain is seeing: Researchers in Singapore have developed MinD-Vis, an AI system that deciphers brain wave patterns to generate images of what a person is looking at. The system matches brain scans with images to create individual AI models for participants, enabling computers to “read” thoughts and recreate visuals. The technology has potential applications in assisting individuals without motor abilities, communicating through thoughts, and integrating into virtual reality headsets. However, the researchers acknowledge challenges such as inter-individual differences in brain function and the need for strict privacy guidelines and ethical laws to protect data sharing and privacy concerns.

AI Business & Market News

Nvidia shares hit all-time high as chipmaker dominates AI market: Nvidia’s shares reached an all-time high as investors anticipate strong quarterly results from the chip designer, which has benefited greatly from the AI boom. The company’s stock has risen by about 19% from a recent two-month low, driven by expectations that its revenue target will exceed Wall Street estimates. Nvidia is a dominant player in the market for chips powering generative AI services. The company’s forecasted 110% growth in third-quarter revenue to $12.50 billion is eagerly awaited, as it has played a significant role in the tech stock rally this year.

Chinese firm launches WonderJourney satellite with AI-powered ‘brain’: Chinese company launches the WonderJourney satellite, touted as the first satellite with AI at its core. The satellite features an onboard intelligent processing unit that enables data processing without the need to transmit it back to ground control. The technology holds potential for applications such as monitoring environmental conditions and providing support during emergency relief efforts.

AI Image & Design Technology

MidJourney Adds Inpainting Feature: MidJourney introduces a new feature called Vary (Region) for image editing. Users can select specific image regions to modify, accompanied by a prompt editor. This tool works effectively for substantial image areas and offers a prompt remixing feature for customization. While it can serve as a general inpainting system, users are advised about potential challenges when deviating significantly from the original images. The feature enhances imaged editing experience, especially with larger regions or slight prompt adjustments according to MidJourney’s Discord message.

🎧 Did you know AI Breakfast has a podcast read by a human? Join AI Breakfast team member Luke (an actual AI researcher!) as he breaks down the week’s AI news, tools, and research: Listen here

Sponsored Tool:

Kypso introduces an AI copilot for managing and scaling team operations, ensuring project success. It facilitates clear decision-making, prevents scope confusion, and promotes transparent progress updates. Users can customize Kypso to align with their workflow, integrating popular tools for seamless functionality.

Alloy Unified API simplifies app experiences by offering a single API interface for building and scaling based on 3rd-party data. Developers can create integration features efficiently, benefiting from standardized API fields. This tool streamlines SaaS integration, enabling seamless connections across various platforms without data structure complexities.

PodStash transforms web links into concise podcast episodes through a Chrome extension. Using AI and tailored LLM prompts, it crafts summarized podcast scripts, with lifelike voiceovers. Episodes are available on compatible podcast platforms via custom RSS URLs. Earn $7 per referral.

EASY.DX empowers game developers with AI-generated voiceovers. Craft unique character voices, manage audio, and export seamlessly in an intuitive dashboard. Generate clips, export in multiple formats, and create lifelike placeholders for early development stages.

Luma AI introduces Flythroughs, an iPhone app using NeRF and 3D generative AI to create professional flythroughs. The easy three-step process - capture, add details, and generate - enables users to craft lifelike 3D experiences, enhancing their space presentations and listings with AI-driven technology.

📄 Giraffe: Adventures in Expanding Context Lengths in LLMs

📄 Using Large Language Models for Cybersecurity Capture-The-Flag Challenges and Certification Questions​​

📄 StableLLaVA: Enhanced Visual Instruction Tuning with Synthesized Image-Dialogue Data

📄 Tryage: Real-time, Intelligent Routing of User Prompts to Large Language Models

📄 ProAgent: Building Proactive Cooperative AI with Large Language Models

Thank you for reading today’s edition.

Your feedback is valuable.

Respond to this email and tell us how you think we could add more value to this newsletter.