• AI KATANA
  • Posts
  • Meta’s free Code Llama AI programming tool closes the gap with GPT-4

Meta’s free Code Llama AI programming tool closes the gap with GPT-4

Also: A Camera-Wearing Baby Taught an AI to Learn Words

Author

AI KATANA
February 01, 2024

Hi!

Recent developments highlight the rapidly evolving field of AI and its integration into various sectors. Meta's introduction of Code Llama 70B, an AI-powered coding tool, represents a significant leap towards improving programming efficiency, nearly matching the capabilities of GPT-4. Concurrently, innovative research has enabled an AI to learn language from a toddler's perspective, demonstrating a novel approach to machine learning. Meta's deployment of custom chips underscores a move towards self-reliance and cost reduction in AI operations, while Amazon's AI assistant, Rufus, aims to revolutionize online shopping with intuitive natural language processing. Furthermore, ASEAN ministers are actively addressing AI-driven cyber threats, emphasizing the need for collaborative governance and cybersecurity measures. Mastercard's AI model, Decision Intelligence Pro, sets a new standard in fraud detection, promising substantial financial savings for banks. These developments, coupled with venture capital investments in AI and cybersecurity startups, reflect a dynamic shift towards leveraging AI for enhanced productivity, security, and innovation across industries.

Sliced:

  • 🆕 Meta’s free Code Llama AI programming tool closes the gap with GPT-4

  • 👶A Camera-Wearing Baby Taught an AI to Learn Words

  • 🦾Meta to deploy in-house custom chips this year to power AI drive

  • 🤖 Amazon announces AI shopping assistant called Rufus

  • 🌏 Asean ministers to discuss AI threats, scams over two-day meeting in Singapore

  • 💳 Mastercard jumps into generative AI race with model it says can boost fraud detection by up to 300%

Meta has introduced Code Llama 70B, a significant upgrade in its AI-powered coding tool lineup, offering enhanced capabilities for generating and debugging programming code. This tool, available for both research and commercial use, marks a leap in accuracy and performance, closely approaching the standards set by GPT-4. With its ability to process more queries and its improved accuracy, Code Llama 70B aids developers in creating more efficient and error-free code, reflecting Meta's commitment to advancing AI technology in the coding domain.

👶 A Camera-Wearing Baby Taught an AI to Learn Words

A recent study demonstrated that an AI model could learn word meanings from a mere 61 hours of video captured from a toddler's perspective, challenging traditional views on language acquisition. This approach, using footage from a camera worn by a child named Sam, allowed the AI to associate words with images without needing the vast datasets typical for training such models. The research underscores the potential simplicity of language learning and suggests that with the right type of data, the gap between human and machine learning might narrow significantly.

Meta plans to integrate a new custom chip, designed in-house, into its data centers this year, supporting its AI initiatives. This second-generation chip aims to lessen Meta's reliance on Nvidia's market-leading chips, potentially saving significant costs in AI operations. The move is part of Meta's broader efforts to enhance computing capacity for AI-driven products across its platforms and devices, aiming for cost-effective and efficient AI workload management.

Amazon has launched Rufus, a new AI shopping assistant designed to make product searching and shopping on Amazon's platform more intuitive and efficient. Rufus allows users to ask questions in natural language about products, leveraging Amazon's vast product catalog, customer reviews, and Q&As to provide helpful answers. Initially tested with a limited user group in the U.S., Amazon plans to expand Rufus nationwide, reflecting its broader strategy to integrate generative AI technologies across its business operations.

ASEAN ministers convened in Singapore for a two-day meeting to address the rising challenges of cyber threats, including scams and misinformation fueled by AI. Spearheaded by Singapore’s Communications and Information Minister Josephine Teo, the discussion focused on AI governance and regional strategies against scams, underlining the importance of international cooperation in enhancing digital security and fostering innovation. The meeting also touched on the significance of adopting UN cyberspace standards and enhancing cross-border data sharing to strengthen the digital economy.

Mastercard has developed a proprietary generative AI model, dubbed Decision Intelligence Pro, aimed at significantly enhancing fraud detection capabilities for banks within its network. Leveraging a proprietary recurrent neural network, this advanced AI technology is expected to boost fraud detection rates by up to 300% in certain scenarios. The model is trained on data from around 125 billion transactions annually, enabling it to identify and predict fraudulent transactions by understanding the relationships between merchants. This cutting-edge approach, which operates in real-time and completes assessments in just 50 milliseconds, represents a major leap in combating fraud, potentially saving financial institutions significant amounts in fraud-related costs.

🛠️ AI tools updates

The article discusses the transformative impact of generative AI tools on Infrastructure as Code (IaC), highlighting their role in accelerating workflow, learning, and career advancement. It emphasizes the utility of tools like GitHub Copilot for code generation, which are becoming indispensable for tasks ranging from shell scripting to writing complex playbooks, thanks to their efficiency and the addition of meaningful comments to the generated code. The article also covers the capabilities of generative AI in code interpretation, enabling the analysis and interpretation of code across different programming languages, and stresses the importance of prompt engineering for achieving accurate responses from large language models. Additionally, it addresses the potential for creating custom GPT models tailored to specific needs, enhancing productivity and decision-making in tech environments.

💵 Venture Capital updates

Reken, an AI and cybersecurity firm, successfully raised $10 million in an oversubscribed seed funding round to combat generative AI threats, including deepfake scams and autonomous fraud. This capital will fuel research and development efforts aimed at pioneering AI solutions to counter these evolving cybercrimes. Supported by a consortium of investors and backed by notable figures connected to Google, Reken's mission is highlighted as addressing the "greatest security challenge of our time," emphasizing the need for new approaches in cybersecurity to tackle the sophisticated threats posed by generative AI.

Antler, a global VC firm, has allocated $5.1M to 37 startups across Southeast Asia, marking its largest pre-seed investment round. These startups span various sectors such as AI, B2B SaaS, fintech, and healthcare, aiming to address regional challenges. This investment underscores Antler's confidence in Southeast Asia's tech landscape and its commitment to supporting early-stage startups, particularly those focusing on AI and Industry 4.0. The initiative highlights a trend towards AI solutions tailored to specific industries and the rise of Industry 4.0 technologies, promising transformative changes across multiple sectors.

🫡 Meme of the day

⭐️ Generative AI image of the day