Apple And Meta Discuss Integrating Llama Gen AI Into iOS 18, Possible Addition Imminent

Welcome to Towards AGI, your premier newsletter dedicated to the world of Artificial Intelligence. Our mission is to guide you through the evolving realm of AI with a specific focus on Generative AI. Each issue is designed to enrich your understanding and spark your curiosity about the advancements and challenges shaping the future of AI.

Whether you're deeply embedded in the AI industry or just beginning to explore its vast potential, "Towards AGI" is crafted to provide you with comprehensive insights and discussions on the most pertinent topics. From groundbreaking research to ethical considerations, our newsletter is here to keep you at the forefront of AI innovation. Join our community of AI professionals, hobbyists, and academics as we pursue the ambitious path towards Artificial General Intelligence. Let’s embark on this journey together, exploring the rich landscape of AI through expert analysis, exclusive content, and engaging discussions.

TheGen.AI News

Apple And Meta Discuss Integrating Llama Gen AI Into iOS 18, Possible Addition Imminent

Apple is currently in talks with Meta to potentially incorporate the latter's generative artificial intelligence technology into Apple devices. There has been ongoing speculation for the last six months about Apple's plans to adopt AI technology in its iPhones and Macs, with possibilities ranging from direct integration to the creation of an AI-focused app store.

Recent reports suggest that Apple's discussions with Meta about generative AI technology are part of a broader strategy hinted at previously by Apple’s executive, Craig Federighi, involving partnerships with various companies to broaden AI offerings for Apple users.

Meta is one of several tech giants rumored to be in talks with Apple for AI integration. Other notable mentions include OpenAI, with expectations of ChatGPT 4o being integrated in the upcoming iOS 18, and Google. Lesser-known entities like Anthropic and Perplexity have also been discussed as potential partners.

The proposed collaboration with Meta could follow a similar pattern to what has been speculated with OpenAI. Rather than paying directly for AI integrations, Apple may allow Meta and other AI providers to offer their services through subscriptions on Apple’s platforms such as iOS, iPadOS, and macOS. This approach could provide Apple users with more options and help Apple avoid additional costs.

However, these negotiations with Meta are reportedly not yet finalized, with potential for the deal to not proceed, as indicated in a recent Wall Street Journal report. Apple’s strategy seems to involve treating AI providers akin to how it manages its hardware suppliers, promoting a competitive environment and choice diversity for consumers.

Apple is planning to start testing its new AI capabilities in the summer of 2024, although specific dates and the scope of the initial testing phase remain unclear.

Overall, Apple is actively seeking partnerships with various AI firms, including Meta, to expand its AI capabilities across its device ecosystem. The company aims to offer consumers more options and maintain cost efficiency while integrating the latest AI technologies into its products. The full details of these negotiations are still developing, and Apple's approach to integrating AI technology continues to evolve.

Wipro Upskills 30,000 Employees in Advanced Generative AI Techniques

In his letter to shareholders in the annual report for FY24, Wipro Chairman Rishad Premji stated that the company has extensively trained over 225,000 employees in Generative Artificial Intelligence (Gen AI) basics, with 30,000 receiving advanced role-specific training. Wipro is intensifying its focus on Gen AI, incorporating it into all client discussions and internal processes. The company’s AI-driven enterprise chatbot, WiNow, has resolved around 65 lakh queries, boosting productivity and reducing operational costs.

Premji emphasized the importance of strong change management to facilitate AI adoption, ensuring employees are prepared to work alongside AI through training and communication. In FY24, Wipro invested heavily in its 'ai360' strategy to integrate AI across all platforms and solutions, and expanded its partnerships with major tech companies like AWS, Google, IBM, Microsoft, and Nvidia, as well as enhancing its in-house innovation labs.

Despite the challenging economic climate affecting technology spending, Premji noted that Wipro remains committed to innovation and has seen promising signs in its consulting business in the latter half of the year. He reaffirmed the company's dedication to leveraging AI for reducing costs, increasing operational efficiency, and driving growth, positioning Wipro at the forefront of industry transformation.

Legal Sector Explores Gen AI for Enhanced Efficiency and Error Reduction

Generative Artificial Intelligence (GenAI) is making significant inroads into the legal sector, as noted by both tech-savvy legal professionals and experts. Law firms and corporate legal teams are beginning to experiment with GenAI tools such as ChatGPT and Microsoft’s Copilot, according to young legal professionals speaking to The Hindu.

This early experimentation signifies a shift towards more structured planning and the creation of policies for GenAI usage, they noted.

Kalindhi Bhatia, a Partner at law firm BTG Advaya, highlighted the benefits of GenAI in streamlining contract management processes, including monitoring, categorization, risk assessment, and compliance. "GenAI platforms help in reducing errors and increasing efficiency," she explained.

Anju Gandhi, a Partner at SNG & Partners, also praised GenAI for improving contract analysis, tracking, and compliance, while reducing the need for manual work. She particularly noted the advantages for young lawyers, such as easier research and drafting.

However, the integration of GenAI in law is not without challenges. Lawyers who are used to traditional practices may be reluctant to adopt GenAI due to concerns about confidentiality and the technology’s unfamiliarity. As GenAI technology is still developing, issues such as bias and a limited grasp of legal intricacies need addressing. "GenAI is not a universal solution and should be used in conjunction with traditional methods. Concerns over costs, evolving technology, and client confidentiality remain, along with resistance from some legal practitioners," Gandhi stated.

According to Bhatia, while GenAI facilitates certain legal tasks, it cannot replace the nuanced human judgment necessary in legal work. Lawyers are finding ways to balance the use of GenAI with conventional methods to serve their clients effectively.

Additionally, the World Development Corporation, a Mumbai-based firm specializing in AI advancements for the legal field, is working on empowering the legal community in India with AI tools. The firm’s Group MD and CEO, Zeeshaan Pathan, mentioned that their AI solution, LawTech, assists lawyers with drafting, research, legal analysis, and precedent checks. "Over 500 technicians are currently enhancing LawTech, our pioneering AI solution for the legal sector," Pathan said.

TheOpen.AI News

Based Hardware Introduces Friend: The Unique Open Source AI Necklace for Conversation Transcription; Priced at $69 for Consumers

Based Hardware has launched Friend, the only fully open-source AI necklace designed to help users remember daily conversations. This device, which now also comes in a consumer version styled more like jewelry, pairs with a mobile app to provide transcriptions, summaries, and follow-up checklists, along with proactive notifications.

Already in the hands of 1,000 developers, Friend allows for the creation of custom plug-ins and enhancements to the app. The newly introduced consumer model features added components like a speaker, onboard memory, and a vibration motor for instant alerts, alongside its microphone and Bluetooth connectivity. An LED indicator signals when the device is recording.

The developer version of Friend is currently available for immediate purchase at $49, while the consumer version is priced at $69 and will be delivered in three months. The associated mobile app is available at no cost.

Nik Shevchenko, CEO and Founder of Based Hardware, emphasized the device’s role in facilitating genuine, hands-free interactions, allowing users to engage more fully in conversations without the distraction of taking notes.

Friend is particularly useful at networking and professional events, recording interactions and helping users recall details like names and discussion topics. The device does not store conversations but instead uses AI to transcribe and summarize them on the user’s phone, ensuring privacy.

Noteworthy features include a six-day battery life, live transcription with high accuracy, and the ability to customize the experience through an app marketplace. The device supports various AI software options for generating transcripts, including WhisperX.

Looking forward, Friend could be developed into tools for monitoring sales meetings, assisting medical professionals, or aiding those with memory impairments. 

The project, which began with a highly successful Kickstarter campaign, has developed the largest open-source community for any wearable to date. Nik Shevchenko, a Thiel Fellowship recipient, along with advisor Monisha Perkash, who previously worked on Meta's Ray-Ban smart glasses, continue to explore future innovations like video capabilities, brain-computer interfaces, and various wearable formats.

Ollama's Open Source AI Server Faces Critical Security Breach

A recent security flaw in Ollama, a widely-used open-source software for operating Large Language Models (LLMs), was found to enable remote code execution. The vulnerability, known as CVE-2024-37032 or Probllama, was identified by Wiz Research on May 5 and quickly patched by the maintainers in the new release, version 0.1.34, available on GitHub since May 6.

Ollama is popular for its ability to perform inference tasks using neural networks from various sources like Meta's Llama family, Microsoft's Phi, and Mistral models. It is frequently accessed via Docker Hub, with hundreds of thousands of pulls each month, and operates either from the command line or through a REST API.

The vulnerability stemmed from inadequate validation mechanisms in Ollama's server-side REST API. According to Sagi Tzadik from Wiz's research team, attackers could exploit this by sending a specially crafted HTTP request to the server. This was particularly concerning in Docker setups where the API server typically runs with root privileges and is publicly exposed by default.

The issue was specifically linked to the API endpoint /api/pull, which allows users to download models from both the Ollama registry and private registries. Researchers discovered that the download process could be manipulated by using a malicious manifest file in the digest field, enabling path traversal attacks. This could lead to file corruption, unauthorized data access, and ultimately, remote system control.

Despite the availability of a patched version, over 1,000 instances of the Ollama server were still vulnerable and exposed online as of June 10. Users are urged to upgrade to the latest version immediately and implement additional security measures. These include using a reverse-proxy for external access since Ollama lacks built-in authentication and isolating the server from direct internet access via firewalls.

Tzadik also highlighted the broader implications of such vulnerabilities in modern software tools designed for deploying LLMs. Even with advancements in programming languages that prioritize safety, traditional security issues like path traversal still pose significant risks.

TheClosed.AI News

How OpenAI's Latest Acquisition, Rockset, Could Transform Digital Marketing?

OpenAI has recently acquired technology from Rockset that promises to revolutionize product creation, real-time data analysis, and recommendation systems, potentially reshaping search marketing soon. Rockset's technology, known as Hybrid Search, combines vector search, text search, and metadata filtering to enhance document retrieval. This method improves the generation process in Retrieval-Augmented Generation (RAG) systems, which are used to produce more accurate and relevant results, similar to those employed in Bing’s AI search and Google’s AI Overviews.

Hybrid Search's significance lies in its ability to index and utilize various data types—including vectors, text, and geospatial data—enabling real-time data application. This versatility is critical for developing both internal and consumer-facing applications, such as contextually relevant product recommendations, customer segmentation, targeted marketing, personalized content aggregation, and location-based services. Rockset has demonstrated the effectiveness of this technology in numerous case studies.

While OpenAI plans to integrate this technology into their retrieval infrastructure, they have only broadly hinted at how this could lead to new APIs and products that businesses and individuals might use. One potential direction could involve OpenAI developing a product based on an API that businesses could use for internal and consumer applications. In this scenario, OpenAI would provide the core infrastructure—similar to its role with ChatGPT and foundation models—encouraging global innovation with OpenAI at the helm.

Apple Declines Meta AI Partnership, Opts for OpenAI and Alphabet Instead: Report

According to a Bloomberg report citing insider sources, Apple Inc. turned down proposals from Meta Platforms Inc. to integrate Meta's Llama chatbot into the iPhone as part of a potential AI partnership. The discussions, which took place briefly in March, did not advance to a formal stage, and Apple currently has no plans to pursue the integration.

Simultaneously, Apple engaged in talks with OpenAI and Alphabet regarding their AI products—ChatGPT and Gemini, respectively. A deal with OpenAI was announced in early June, while a collaboration with Alphabet's Gemini is anticipated in the future.

The report suggests that Apple found Meta's privacy practices insufficiently rigorous, which may have influenced its decision. Apple has also been openly critical of Llama, making its integration a contradictory move. In contrast, ChatGPT is seen as a more robust offering, and Apple's existing relationship with Google for its Safari browser makes a partnership on the Gemini project plausible.

Additionally, Apple is reportedly seeing promising results from its AI startup, Anthropic, which could lead to the launch of a chatbot. The OpenAI deal also includes a provision allowing ChatGPT subscribers to access their services on iOS, generating potential revenue for OpenAI with Apple benefiting from App Store commissions.

During the Worldwide Developer Conference 2024 in California on June 10, Apple unveiled 'Apple Intelligence,' a new suite of AI features. These features will be available to iPhone 15 Pro users and later models, as well as iPad users with M1 chipsets or newer. This upgrade will enhance Siri's capabilities, enabling it to interact with and perform actions within third-party apps based on on-screen content and user commands. For example, users will be able to locate specific images via Siri or have Siri perform calculations and note-taking within apps.

Amazon's New AI Chatbot 'Metis' Aims to Rival OpenAI's ChatGPT, Says Report

Amazon.com Inc. is developing a new AI chatbot named "Metis," aiming to rival OpenAI's ChatGPT, according to a report from Business Insider. This initiative, based on Amazon's internal AI model called Olympus, promises to deliver more current responses compared to other chatbots. Metis is expected to automate intricate tasks by leveraging existing data, with plans to debut the chatbot at a significant Alexa-related event in September, though the launch date may adjust.

The development is spearheaded by Amazon’s AGI team, led by Rohit Prasad, the company's senior vice president and head scientist. Amazon CEO Andy Jassy noted that AI initiatives are widespread across the company, which has long been a pioneer in cloud computing and machine learning. Jassy also mentioned that Amazon's AI ventures are projected to contribute over $1 billion in annual revenue, potentially generating tens of billions in the future.

Amazon's recent AI efforts have not been without their hurdles. The company's earlier AI chatbot, Q, was criticized for its inaccurate outputs, an issue Amazon attributed to a hasty launch and an underperforming AI model. Improvements have been underway to enhance these capabilities.

Additionally, Amazon's generative AI operations recently hit a multi-billion dollar revenue run rate, as disclosed during their first-quarter 2024 earnings call. Jassy emphasized the strong demand for AI solutions, which customers are eager to utilize to revamp their businesses and improve client interactions.

In May, Amazon introduced AI-driven search functionalities to its Fire TV devices, signaling its ongoing commitment to embedding AI throughout its product range. This new feature is set to enhance search on Fire TV, positioning it in competition with other streaming platforms such as Roku and Apple TV.

In our quest to explore the dynamic and rapidly evolving field of Artificial Intelligence, this newsletter is your go-to source for the latest developments, breakthroughs, and discussions on Generative AI. Each edition brings you the most compelling news and insights from the forefront of Generative AI (GenAI), featuring cutting-edge research, transformative technologies, and the pioneering work of industry leaders.

Highlights from GenAI, OpenAI and ClosedAI: Dive into the latest projects and innovations from the leading organisations behind some of the most advanced AI models in open-source, closed-sourced AI.

Stay Informed and Engaged: Whether you're a researcher, developer, entrepreneur, or enthusiast, "Towards AGI" aims to keep you informed and inspired. From technical deep-dives to ethical debates, our newsletter addresses the multifaceted aspects of AI development and its implications on society and industry.

Join us on this exciting journey as we navigate the complex landscape of artificial intelligence, moving steadily towards the realisation of AGI. Stay tuned for exclusive interviews, expert opinions, and much more!

Keep reading