📆 July 19th 2024📆

Price drops for advanced models, yet another challenger to big tech, exciting partnerships & new features.

• I believe this is one of OpenAI’s best moves. Continuously improving and lowering the price of it’s models. GPT-4o mini is better and chapter than 3.5 - which is commonly used for first pass efforts before applying larger, more expensive models.

• Anyone who has followed me knows I’m a big fan of Groq. they have released two models based on Llama3. Coupled with the speed Groq brings these are serious contenders…. and open source.

• Google have secured their spot as USA’s AI partner. The Olympics stand to showcase many AI technologies.

• OpenAI Announce better Enterprise grade controls - interesting for those who’s organisations are still on the fence about adoption & security.

• As always, too much to put in headlines. Check out the Other News Rollup section for a quickfire review of the top stories in AI.

Dive in for all the details!

💫Help Support my Work💫

I really enjoy researching and writing about AI & appreciate your support. If you enjoy this content please use any of the options below to help

• Buy Me a Coffee☕️- AI is really fuelled by coffee - keep the tank filled

• Spread the Word🗞️- Share on social or directly with your friends

• Give me Feedback💬 - Tell me what’s important to you

• Follow me on X - Always happy to chat AI and software dev

📰News📰

OpenAI Launches Cost-Effective GPT-4o Mini

OpenAI has introduced GPT-4o Mini, a smaller and more affordable version of its GPT-4o language model. This new model retains multimodal capabilities, handling both text and vision, with a context window of 128,000 tokens. It can produce up to 16,000 tokens of output and is trained on data up to October 2023.

The GPT-4o Mini is designed to be cost-efficient, priced at 15 cents per million input tokens and 60 cents per million output tokens, significantly cheaper than the full GPT-4o model. For those using delayed batch processing, these costs can be halved, making advanced AI more accessible to a broader audience. Initially, the Mini version supports only text and vision, with future plans to include audio and other formats.

Benchmark tests indicate that GPT-4o Mini outperforms comparable models from competitors like Google and Anthropic by 5 to 15 percent in most tasks, highlighting its competitive edge in accuracy and efficiency. This makes it a strong contender for applications requiring frequent AI interaction, such as customer service chatbots and data analysis tools.

Read More At: The Register

Groq's Llama AI Model Outperforms Competitors

Groq, an AI hardware startup, has launched two open-source language models that outperform major competitors in specialized tool use. The Llama-3-Groq-70B-Tool-Use model now tops the Berkeley Function Calling Leaderboard (BFCL), surpassing proprietary models from OpenAI, Google, and Anthropic. The larger 70B model achieved a 90.76% accuracy rate, while the smaller 8B model scored 89.06%, ranking third overall.

These models were developed in collaboration with AI research company Glaive and trained using synthetic data, addressing data privacy and overfitting concerns. They are available via the Groq API and Hugging Face, promising to democratize access to advanced AI capabilities and foster innovation.

Read More At: VentureBeat

Google Becomes Official AI Sponsor for Team USA at 2024 Olympics

Google has secured the title of "official AI sponsor for Team USA" for the upcoming 2024 Olympics in Paris, commencing on July 26th. This partnership with NBCUniversal will prominently feature Google's AI-powered technologies throughout the event broadcasts. Key integrations include Google Maps' 3D views of Olympic venues, such as Versailles and Stade Roland Garros, which will provide viewers with immersive visualizations of event locations.

Moreover, Google Search's AI Overviews will be incorporated into broadcast segments to address common Olympic and Paralympic questions, enhancing the viewer experience with AI-driven insights. Additionally, comedian Leslie Jones will engage with Google's Gemini AI in scripted activities aimed at introducing new sports to viewers. Olympic and Paralympic athletes will also participate in promotional activities using Google's AI tools across social media and late-night shows.

Read More At: The Verge

OpenAI Enhances Enterprise Control and Compliance

OpenAI has introduced new enterprise control features for its ChatGPT Enterprise users, enhancing compliance, data security, and scalability. The new Enterprise Compliance API allows users to audit interactions, including conversations and file uploads, by providing time-stamped records. This integration, combined with third-party compliance providers like Microsoft Purview and Palo Alto Networks, helps businesses meet regulatory standards such as GDPR, HIPAA, and FINRA.

OpenAI also introduced an identity management system for easier user permission management and broader administrative controls over custom GPTs. These updates aim to support the enterprise sector's growing demand for secure AI solutions.

Read More At: VentureBeat

🔨AI Powered Tools🔨

• HTTPie AI - In early pre-release, designed for API developers. Integrated into HTTPie, it allows users to create and manage API requests using natural language. By leveraging OpenAI’s GPT-3 models, it simplifies the process of generating and customizing HTTP requests. This feature is especially useful for those who need to interact with APIs without extensive knowledge of the underlying structure.

• Fliki - Platform that turns text into engaging videos. It caters to various industries, including marketing, education, and entertainment, by providing tools to create videos with AI voices, realistic avatars, and multimedia elements. The platform leverages advanced AI to generate high-quality voiceovers and automate the video creation process, making it easy to produce professional content without technical expertise.

• Filmora AI Translation - feature within the Filmora video editor that enables seamless video translation into 23 languages. It utilizes AI to add multilingual voiceovers, generate synchronized subtitles, and ensure high accuracy in translation. This tool is ideal for content creators, marketers, and educators looking to reach a global audience by breaking language barriers efficiently.

🗞️Other News Rollup🗞️

AI and Carbon Dispute - Google's AI not solely responsible for carbon emissions rise. Challenges in clean energy commitment. Microsoft faces similar struggles.

AI Image Generation Tool - Microsoft launches Designer, an AI image generation platform with templates and Copilot integration.

AI Stability Assurance - MIT develops techniques for neural network-controlled robots, ensuring stability and safety in various conditions. Source: MIT News

AI Startup Fund - Menlo Ventures and Anthropic launch $100M fund for AI start-ups, providing support and investments.

YouTube Subtitles for AI - AI labs train models using YouTube subtitles without creators' consent, sparking legal concerns and lawsuits.

Coding's AI Evolution - AI assistants like SWE-agent and GitHub Copilot are transforming coding, with companies like Factory AI and Amazon leading the way.

AI's Human-Washing Dilemma - Voice bots mimic humans, raising ethical concerns. Bland AI easily tricked into lying about being human.

AlphaFold 3 Breakthrough - AlphaFold 3 predicts all biomolecules' structures, simplifying predictions with a diffusion-based model. Free tool generates 1M+ structures.

Desktop AI Revolution - Nvidia and Mistral release Mistral-NeMo, a powerful desktop AI model, democratizing access to advanced AI solutions.

AI in Gaming - Artificial Agency secures $16M for AI-powered behaviour engine, enhancing gaming experiences with generative AI agents.

Innovative Science Experiment Platform - Briefly Bio raises $1.2M to improve reproducibility in science experiments by converting lab documentation using AI.

AI-Powered Dating Photos - Tinder's AI suggests best dating photos, easing profile setup for users. Privacy concerns addressed.

EU AI Model Restriction - Meta and Apple withhold AI models in EU due to regulatory concerns, impacting services and competition.

🎶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.