AI & Startups July 29th - August 4th

Gen AI Product News

Meta Unveils SAM 2 for Video AI

Meta has just introduced Segment Anything Model 2 (SAM 2), an advanced AI model that can identify and track objects across video frames in real-time, marking a significant leap in video AI.

? The Details:

?? Advanced Capabilities: SAM 2 extends Meta's previous image segmentation capabilities to video, addressing challenges like fast movement and object occlusion.

?? Ease of Use: The model can segment any object in a video and create cutouts in a few clicks, with a free demo available to try.

?? Open Source: Meta is open-sourcing the model and releasing a large, annotated database of 50,000 videos used for training.

?? Potential Applications:

• Video Editing
• Mixed Reality Experiences
• Scientific Research

?? Why It Matters: SAM 2’s ability to track objects in real-time could make complex video editing tasks like object removal or replacement as simple as a single click. With Llama 3.1 last week and now SAM 2, Meta is continuing its strategy of developing massive AI breakthroughs while making everything open and free to use.

??? OpenAI Begins ChatGPT Voice Rollout

OpenAI has begun a limited rollout of its highly anticipated ‘Advanced Voice Mode’ for paying ChatGPT Plus users, offering natural, real-time conversations and the ability for the AI to detect and respond to emotions.

? The Details:

?? Initial Access: The feature will initially be available to a small group of ChatGPT Plus users, with plans to give all Plus users access by fall 2024.

?? Advanced Capabilities: Advanced Voice Mode uses GPT-4o and can sense emotions in users' voices, including sadness, excitement, or singing.

?? Future Features: Video and screen-sharing capabilities, previously showcased in OpenAI’s early demo, will launch at a ‘later’ date.

?? Early Access: OpenAI has sent email instructions to the initial ‘Alpha‘ group selected for early access.

?? Why It Matters: AI is slowly shifting from a tool we text/prompt with, to an intelligence that we collaborate, learn, and grow with. Advanced Voice Mode’s ability to understand and respond to emotions in real-time convos could also have huge use cases in everything from customer service to mental health support.

Perplexity??’s Publisher Revenue-Sharing Program

Perplexity just introduced a "Publishers' Program" to share ad revenue with media partners, following recent plagiarism accusations and aiming to support quality journalism in the age of AI-powered search.

? The Details:

?? Revenue Sharing: The program includes cash advances on future revenue as Perplexity builds its advertising model, set to launch in September.

?? Initial Partners: Time, Der Spiegel, Fortune, WordPress.com, and more will receive a "double-digit percentage" of ad revenue.

?? Added Benefits: Partners also get free access to Perplexity's Enterprise Pro tier, developer tools, and insights through Scalepost AI.

?? Why It Matters: Despite constant pushback on AI firms and their training data, media companies are finding few available paths forward other than accepting partnership deals. Perplexity's initiative is a good step toward fairness, but it likely won’t be the end of the growing pains with publishers.

Google Gemini 1.5 Pro Tops Chatbot Leaderboard

For the first time ever, Google DeepMind's experimental Gemini 1.5 Pro has claimed the top spot on the AI Chatbot Arena leaderboard, surpassing OpenAI's GPT-4o and Anthropic's Claude-3.5 with an impressive score of 1300.

? The Details:

?? Community Votes: Gemini 1.5 Pro (experimental 0801) gathered over 12K community votes during a week of testing on the LMSYS Chatbot Arena.

?? Top Rankings: The new experimental model achieved the #1 position on both the overall and vision leaderboards.

?? Early Access: The experimental version is available for early testing in Google AI Studio, the Gemini API, and the LMSYS Chatbot Arena.

?? Future Updates: Google DeepMind hasn't disclosed specific improvements, but promises more updates soon.

?? Why It Matters: Without any announcement, Gemini 1.5 Pro unexpectedly rose to the top of the overall AI chatbot leaderboard — by a whopping 14 points. This leap means that either Google just quietly established itself as the new leader in the LLM space, or we’re on the cusp of major competitive responses from industry rivals.

Google’s Tiny AI Model Bests GPT-3.5

Google just unveiled Gemma 2 2B, a lightweight AI model with just 2B parameters that outperforms much larger models like GPT-3.5 and Mixtral 8x7B on key benchmarks.

? The Details:

?? Compact Power: Gemma 2 2B boasts just 2.6B parameters but was trained on a massive 2 trillion token dataset.

?? Benchmark Performance: It scores 1130 on the LMSYS Chatbot Arena, matching GPT-3.5-Turbo-0613 (1117) and Mixtral-8x7b (1114) — models 10x its size. Other notable scores include 56.1 on MMLU and 36.6 on MBPP, beating its predecessor by over 10%.

?? Open Source: The model is open-source, and developers can download the model’s weights from Google’s announcement page.

?? Why It Matters: As we enter a new era of on-device, local AI, lightweight and efficient models are crucial for running AI directly on our phones and laptops. With Gemma 2 beating GPT-3.5 Turbo at just 1/10th the size, Google isn't just showing what's possible — they're cementing their position as the leader in the small model space.

Friend ’s AI Companion Necklace

Avi Schiffmann , a Harvard dropout and Webby Award winner, just unveiled Friend, a $99 AI-powered wearable device designed to combat loneliness by providing constant companionship.

? The Details:

?? Companionship Focus: Friend is a pendant-like device that hangs around the neck and uses AI to engage in conversations and offer emotional support.

?? Always Listening: The device is always listening and can proactively send messages based on the wearer's context.

?? Funding Success: Friend has raised $2.5 million in funding at a $50 million valuation from notable investors.

?? Preorders Open: Preorders are now open for the basic white version, with shipping expected in January 2025.

?? Why It Matters: Friend takes a different approach compared to other AI wearables by focusing on companionship rather than productivity. However, with a market flooded with overpromises, privacy implications, and concerns about real human connection, the pendant is fighting an uphill battle when it comes to trust.

Black Forest Labs Announces FLUX.1

Black Forest Labs just launched FLUX.1, a suite of state-of-the-art AI image generation models that rival current leaders like MidJourney and DALL-E 3 — coming in three variants: [pro], [dev], and [schnell].

? The Details:

?? FLUX.1 [pro]: The highest-quality model, offering state-of-the-art performance, available via API and for free on Replicate.

?? FLUX.1 [dev]: An open-weight, non-commercial version matching [pro]'s quality while outperforming competitors in efficiency at the same size.

? FLUX.1 [schnell]: An ultra-efficient, 4-step model tailored for local development or personal use.

?? Future Tease: Black Forest Labs also teased its upcoming text-to-video generation model, which appears to rival Sora in quality.

?? Why It Matters: FLUX.1's incredible quality and open-source options are set to democratize high-quality AI image generation. With the teased AI video model on the way, we might see an open-sourced Sora-level video generator hit the market before Sora itself.

?? Stability AI 's Instant 3D Asset Generator

Stability AI just introduced Stable Fast 3D, an AI model that generates high-quality 3D assets from a single image in just 0.5 seconds — potentially reshaping industries from gaming to e-commerce.

? The Details:

?? High-Quality 3D Assets: The model creates complete 3D assets, including UV unwrapped mesh, material parameters, and albedo colors with reduced illumination bake-in.

? Speed and Efficiency: It outperforms previous models, reducing generation time from 10 minutes to 0.5 seconds while maintaining high-quality output.

?? Availability: Stable Fast 3D is available on Hugging Face and through Stability AI's API, under Stability AI's Community License.

?? Why It Matters: The leap from 10 minutes to 0.5 seconds for high-quality 3D asset generation is nothing short of insane. We're entering a world where video games will soon feature infinite, dynamically generated assets, e-commerce will have instant 3D product previews, architects will see designs in real-time, and so much more.

Video Summary:

Capital Watch: This Week's Funding Highlights

AI Startup EMA Raises $36M, Launches Persona Builder for Custom AI Agents

Ema, an enterprise AI company, has successfully raised an additional $36 million in Series A funding, bringing its total capital raised to over $61 million. The latest funding round was led by Accel Partners and Section 32, with contributions from Prosus Ventures, Hitachi Ventures, Sozo Ventures, Wipro Ventures, SCB10X, and Frontier.

? The Details:

?? Persona Builder: Ema has launched Persona Builder for creating custom AI agents designed to perform complex workflows.

?? Proprietary Technology: The AI agents, known as Ema’s Personas, are supported by its proprietary Generative Workflow Engine? and the EmaFusion? model.

?? Funding Purpose: The additional capital will be used to enhance the company's AI capabilities and expand its market reach.

?? Why It Matters: Ema’s innovative approach to building custom AI agents for complex workflows positions it as a leader in the enterprise AI space. The significant funding will enable further development of its proprietary technologies and expansion into new markets, potentially transforming how businesses leverage AI for operational efficiency.

intelmatix Raises $20M Series A to Enable MENA Businesses to Tap AI for Decision-Making

Intelmatix, a deep tech B2B startup targeting businesses in the MENA region, has closed a $20 million Series A funding round to expand its AI-powered enterprise decision intelligence platform, EDIX.

The Details:

?? Focus Areas: Intelmatix aims to help businesses in retail, logistics, and workforce sectors unlock intelligence on operational and strategic issues like demand and supply, recruitment, staff planning, fleet management, and marketing.

?? EDIX Platform: Their platform can provide recommendations with high accuracy, like predicting the best location for a new branch and its projected revenue.

?? Regional Customization: Unlike competitors like o9 and Palantir, Intelmatix's EDIX is tailored for the MENA region, addressing local data and knowledge gaps.

?? Future Plans: The funding will be used to expand its platform's capabilities and target larger and medium businesses, as well as public entities in MENA.

?? Investors: The round was led by Shorooq , with participation from Olayan Financing Company, Rua Growth Fund , Saudi Technology Ventures , Saudi Venture Capital Investment Company ny, Sultan Holdings, and Zain-Ventures .

Why It Matters:

Intelmatix's EDIX platform is designed to democratize access to AI in the MENA region, providing businesses with the tools to make informed decisions and improve efficiency, even without a dedicated AI team. The startup's success in securing one of the largest Series A rounds for a regional company underscores the growing demand for AI solutions in MENA's business landscape.

Bengaluru-based Conversational AI Startup gnani.ai Raises $4M

Gnani AI, India’s first voice-first SLM for vernacular languages, has raised $4 million from tech holding company Info Edge to support its expansion plans.

? The Details:

?? Customer Base: Gnani AI serves over 100 customers in industries including banking and financial services, insurance, telecom, automotive, and healthcare.

?? Innovative Technology: The startup is pioneering a fusion of voice and text models with its multimodal model, focusing on vernacular languages.

?? Funding Purpose: The investment from Info Edge will be used to accelerate the startup's expansion plans.

?? Why It Matters: Gnani AI’s innovative approach to conversational AI, particularly in vernacular languages, positions it as a key player in the Indian market. The funding will enable them to expand their reach and enhance their technology, potentially transforming customer interactions across various industries.

AI in Global Events

AI Revolutionizes the 2024 Olympics

The Paris 2024 Summer Olympic Games are showcasing an extensive use of AI, transforming experiences for athletes, spectators, and organizers, signaling a new era in sports.

? AI Innovations:

?? AthleteGPT: An AI chatbot providing 24/7 assistance to athletes through the Athlete365 mobile app.

?? 3D Athlete Tracking (3DAT): AI-powered technology offering detailed biomechanical insights for performance enhancement.

?? Talent Scouting: AI used in a recent IOC pilot program in Senegal.

?? Personalized Highlights: NBC using AI to provide personalized highlights and enhanced real-time statistics for viewers.

?? Why It Matters:

The integration of AI at a major sporting event like the Olympics marks a significant shift towards embracing technology, potentially paving the way for a new era in sports viewing and management.

?? Athlete Training:

Belgian runner John Heymans successfully used AI to create a training plan via ChatGPT. AI technologies are also being utilized to protect athletes from online abuse and to monitor energy consumption at venues.

Technogym’s AI-Powered Innovations:

?? Technogym Checkup: Customizes training programs by analyzing strength, cardio performance, balance, mobility, and cognitive abilities.

???♂? Technogym Biostrength: Guides strength training with AI, ensuring correct weights, posture, and speed, and adjusts load automatically to optimize results.

?? Future of AI in Olympics:

The IOC's chief technology officer, Ilario Corna, emphasizes that pioneering AI concepts will enhance the Games and make them future-ready. The IOC has also launched an Olympic AI Agenda to guide AI application at future events.

"AI tools have become crucial allies for today’s Olympic athletes," Technogym said. The adoption of AI marks a shift towards smarter, more efficient sports training and management.