Hacker Breaches OpenAI's Internal System, Steals AI Design Details

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 toward 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

Meta Introduces AI-Powered Meta 3D Gen for Instant 3D Asset Creation

Meta has unveiled a new research paper detailing a generative artificial intelligence (AI) model capable of creating 3D assets, named Meta 3D Gen. This AI model can produce high-quality 3D models from text prompts in under a minute. Meta emphasized that the model can generate comprehensive high-resolution textures and material maps by using in-house developed texture generation and text-to-mesh techniques. This method sets Meta apart from companies like Tencent and Stability AI, which focus on converting images to 3D models.

The research paper, currently available on Meta's website, claims that the Meta 3D Gen model can render 3D models with high accuracy to the text prompts while maintaining high quality. Additionally, the model supports physically based rendering (PBR), enabling the generated models to be relit in various real-world applications. However, the research paper has not undergone peer review, so some claims might be overstated, and verification will only be possible once the model is publicly released.

The paper details the development approaches for the AI model, which includes two proprietary systems: Meta 3D AssetGen and Meta 3D TextureGen. AssetGen generates 3D objects from text prompts through a two-stage process. In Stage I, it creates the 3D model based on the prompt, and in Stage II, it refines the model by reconciling view-based textures and completing hidden parts. TextureGen is responsible for finding and applying appropriate textures to the model. Meta claims that the 3D Gen AI model has a 68 percent success rate in a single stage, meaning it correctly generates the desired 3D asset from the text prompt 68 times out of 100. The company also asserts that the model outperforms industry standards in terms of prompt accuracy and visual quality for complex prompts.

Morgan Stanley's AI Debrief: Pioneering Global Data Analysis

When Morgan Stanley announced its new generative AI support tools for financial advisors last week, it emphasized the efficiency gains from its notetaking capabilities. However, analysts following the $54 billion financial firm believe its objectives are far more ambitious.

The stated purpose of AI @ Morgan Stanley Debrief, part of a suite of generative AI tools being developed for financial advisors, is to record, transcribe, and summarize key points from over 1 million conference calls held annually by Morgan Stanley employees.

Powered by OpenAI, the Debrief tool will also draft an email for the financial advisor to review and send, and log a note about the call in Morgan Stanley’s Salesforce system. The true potential lies not in the AI's ability to capture and summarize information, a capability already offered by Apple, Android devices, Zoom, Microsoft, Google, and various independent apps. The real advantage is that all Morgan Stanley employees and contractors will use the same tool, ensuring uniform data formats for comprehensive analysis.

This would mean that Morgan Stanley management could, in theory, analyze every call across the enterprise within minutes of its completion. This analysis would reveal if employees are delivering the intended messages, understand client priorities, and track changing customer sentiments. Aaron Cirksena, founder and CEO of MDRN Capital, highlighted the value of achieving deep global analytics insights from consistent data.

“It’s essentially Big Brother, giving them an eye on every conversation,” Cirksena said. “Did our people actually discuss the topics they were instructed to cover?”, he added.

Jonathan Murray, the chief strategy officer at marketing firm Mod Op, acknowledged the significant analytics potential of Morgan Stanley's AI initiative. “Client interactions typically represent an untapped source of insight for most companies. Morgan Stanley is demonstrating how new large language models can transform these interactions into valuable insights that every organization will want to leverage,” Murray stated.

His colleague at Mod Op, executive-in-residence Monica Richter, noted that the value will depend on how deeply Morgan Stanley executives explore their newly acquired data. “Will they use AI to aggregate client feedback for generating research ideas or making buy/sell decisions? If a client mentions setting up a retirement account, will AI not only note this request but also automatically send relevant articles on IRAs, Roths, etc., tailored to the client’s portfolio needs?” Richter wondered.

Even if everything goes as planned, some experts question whether this level of analysis could pose risks for Morgan Stanley. Rebecca George, managing director of AI consulting firm Slalom, pointed out that any data accessible to the company could be subject to legal discovery in litigation.

“What if someone wanted to check if Morgan Stanley was giving different advice based on gender, ethnicity, or race? This opens them up to potential bias in guidance litigation,” George remarked. “This is where culture and ethics intersect, making all this information discoverable. Are they prepared for this level of exposure?”

Observers also highlighted a notable omission in Morgan Stanley’s announcement: no mention of protecting sensitive client data given the nature of OpenAI systems. “Clients will ask, ‘How are you protecting us?’ Morgan Stanley needs a clear strategy and should communicate it publicly,” said HP Newquist, executive director of AI consulting firm The Relayer Group.

Lenovo Introduces Cutting-Edge AI Solutions and Liquid Cooling Innovations

Lenovo has unveiled its latest comprehensive enterprise AI solutions designed to help businesses of all sizes leverage the power of artificial intelligence (AI). This new suite includes the sixth generation of Lenovo Neptune liquid cooling innovations and AI solutions powered by NVIDIA, aimed at supporting large-scale AI deployment both efficiently and sustainably.

These new AI solutions from Lenovo are created to make AI accessible and practical for various sectors. The offerings include turnkey services, business-ready vertical solutions, and energy-efficient innovations that expedite the practical application of AI. Sumir Bhatia, President of Lenovo ISG, emphasized the transformative potential of these innovations: “AI is more than just a technology; businesses require advanced technologies, tailored services, and energy-efficient infrastructure for effective deployment across industries. Lenovo’s new solutions provide exactly this, enabling companies to leverage AI in real-time while maintaining operational efficiency and sustainability. Our objective is to empower businesses to utilize AI effectively, driving transformation across all sectors.”

Central to Lenovo’s new AI strategy is its collaboration with NVIDIA and the Lenovo AI Center of Excellence. This partnership addresses various AI applications by offering comprehensive services, from pocket to cloud. The new AI Innovator solutions and the Lenovo Neptune liquid cooling system are designed to support the mainstream rollout of AI-ready computing without compromising energy efficiency.

Amit Luthra, MD of Lenovo ISG India, highlighted the strategic importance of these innovations: “With 95% of Indian CIOs recognizing AI’s competitive advantage, our partnership with NVIDIA and the expansion of the Lenovo AI Center of Excellence enable us to provide tailored, scalable AI solutions. Furthermore, these innovations ensure energy efficiency and high performance.”

Generative AI: The Next Frontier

Lenovo underscores its commitment to AI innovation with a significant investment in generative AI. A joint research study with IDC and NVIDIA, commissioned by Lenovo, shows that generative AI is the top tech investment priority for 2024 among IT and business decision-makers globally. This expansion is part of Lenovo’s $1 billion investment over three years to accelerate generative AI deployment, aiming to deliver faster outcomes and maximize performance and efficiency at any stage of the AI journey.

Smarter AI Solutions for Every Business

Lenovo’s third annual global CIO report identifies AI as the most urgent priority for IT, though many organizations struggle with its implementation. To address this, Lenovo offers a five-stage AI Services Center of Excellence (COE) that combines business advisors, data scientists, and infrastructure to maximize outcomes. The Lenovo AI Discover service guides businesses through the initial steps of AI adoption, while the COE’s new services and NVIDIA offerings accelerate progress.

TheOpen.AI News

Open-Source AI Platform Sentient Raises $85M in Major Funding Round

San Francisco-based open-source AI development platform Sentient announced today that it has raised $85 million in seed funding to challenge the dominance of centralized AI model development.

The funding round was co-led by Peter Thiel’s Founders Fund, Pantera Capital, and Framework Ventures. Other investors included Ethereal Ventures, Robot Ventures, Symbolic Capital, Delphi Ventures, Hack VC, Arrington Capital, HashKey Capital, Canonical Crypto, and Foresight Ventures.

Sandeep Nailwal, co-founder of Polygon, a blockchain network solution addressing Ethereum scaling issues, is one of Sentient's core contributors. He works alongside other key contributors such as Professor Pramod Viswanath from Princeton University, co-inventor of Flash OFDM, the technology behind the 4G wireless standard, and Professor Himanshu Tyagi from the Indian Institute of Science.

Another significant contributor is the AI studio startup Sensys, founded by Kenzi Wang, co-founder of Symbolic Capital. Sensys will support companies within the Sentient ecosystem.

“The rapid advancement of AI has the potential to transform every aspect of our lives, but the concentration of power in the hands of a few centralized entities poses significant risks,” Nailwal said in a statement. “By building an open platform for AGI development, we aim to ensure that the benefits of AI are distributed equitably and that its development aligns with the interests of humanity as a whole.”

Artificial general intelligence (AGI) is a theoretical field of AI research focused on developing software capable of human-like intelligence. AGI is considered the “holy grail” of AI, as it would enable the creation of highly capable, broadly task-oriented AI assistants.

Currently, most AGI development and research are controlled by large firms due to their substantial resources. Closed-source models pose challenges because their parameters and variables cannot be examined, verified, or distributed by third parties, remaining under centralized control.

Sentient plans to build an AI platform that allows contributors to collaborate and monetize their innovations, aiming to advance "Open AGI." Blockchain technology will be used to incentivize people to build, copy, and expand AI models, rewarding pioneers for their contributions.

“Currently, anyone can just copy models without paying for them, and Sentient aims to solve this incentive problem which disincentivizes open-source AI,” said Joey Krug, a partner at Founders Fund. “This is a really interesting research area.”

Sentient will enter the testnet phase this quarter and plans to use the funding to hire experts to expand its team, including AI talent and blockchain engineers.

Latham & Watkins' New AI Director Unveils OpenContracts: A Free, Open-Source Document Analysis Tool"

John Scrudato, recently appointed as the Director of Artificial Intelligence and Innovation at Latham & Watkins, has developed OpenContracts, a free and open-source AI tool designed for document analysis and annotation. Scrudato describes it as a generative AI native system that leverages LlamaIndex for both querying and data extraction. It enables users to manage, process, and query contracts, corporate legal documents, and more. LlamaIndex, for those unfamiliar, is a comprehensive interface toolkit that enhances large language model (LLM) applications with external data. It offers a versatile data framework that connects various data sources—unstructured, structured, or semi-structured—to large language models, allowing businesses to develop robust end-user applications effortlessly.

OpenContracts, which is available for free, offers the following capabilities: -

Document Management: Organize document collections (Corpuses).

Layout Parser: Automatically extracts layout features from PDFs.

Automatic Vector Embeddings: Generates embeddings for uploaded PDFs and extracted layout blocks.

Pluggable Microservice Analyzer Architecture: Facilitates document analysis and automatic annotation.

Human Annotation Interface: Allows manual annotation of documents, including multi-page annotations.

LlamaIndex Integration: Uses vector stores powered by pgvector and manual or automatic annotations to enable an LLM to answer questions intelligently. - Data Extraction: Allows users to ask multiple questions across numerous documents using complex LLM-powered querying.

Custom Data Extraction: Supports custom data extraction pipelines for bulk document querying on the frontend.

TheClosed.AI News

Hacker Breaches OpenAI's Internal System, Steals AI Design Details

A hacker infiltrated OpenAI's internal messaging systems last year, stealing details about the company's artificial intelligence technology design, according to a New York Times report on Thursday. The hacker obtained information from online forum discussions where OpenAI employees discussed their latest technologies, the report, which cited two people familiar with the incident, revealed.

However, the hacker did not access the systems where OpenAI, the company behind the popular ChatGPT, stores and develops its AI, the report noted.

OpenAI, which is backed by Microsoft Corp, did not immediately respond to a request for comment from Reuters.

According to the report, OpenAI executives informed employees at an all-hands meeting in April last year and notified the company's board about the breach. However, they decided not to make the incident public as no customer or partner information had been compromised.

The executives did not view the incident as a national security threat, believing the hacker to be an individual without known ties to any foreign government. As a result, OpenAI did not inform federal law enforcement agencies about the breach.

In May, OpenAI announced it had disrupted five covert influence operations that attempted to use its AI models for deceptive activities across the internet, highlighting ongoing safety concerns about potential misuse of the technology.

Meanwhile, the Biden administration is planning new measures to protect U.S. AI technology from countries like China and Russia, with preliminary plans to impose safeguards on advanced AI models, including ChatGPT. In May, 16 companies developing AI pledged at a global meeting to ensure the technology's safe development, as regulators struggle to keep pace with rapid innovation and emerging risks.

Apple Joins OpenAI Board as Observer, Strengthening AI Strategy

Last month, Apple announced a significant strategic partnership with OpenAI, gaining an observer role on OpenAI’s board. This move aims to enhance Apple’s AI capabilities and strengthen its position in the tech industry. Phil Schiller, head of Apple’s App Store and former marketing chief, will take on this role later this year.

As an observer, Phil Schiller will attend OpenAI’s board meetings, gaining insights into decision-making processes without having voting power. This position allows Apple to understand OpenAI’s strategies and developments better, aligning Apple with Microsoft, OpenAI’s major supporter and AI technology provider. This keeps Apple competitive in the rapidly evolving AI landscape.

However, the partnership might bring complications due to the long-standing rivalry between Microsoft and Apple. Microsoft may prefer Schiller to be excluded from discussions about future AI initiatives involving both companies.

Phil Schiller, an Apple Fellow, manages the App Store and oversees product launches. Although he stepped down as marketing head in 2020, he remains influential at Apple and serves as a director at medical technology company Illumina.

Integration of ChatGPT with Apple Devices

During the WWDC 2024 event in June, Apple announced the integration of ChatGPT into iPhones, iPads, and Macs, as part of its AI feature suite. The upcoming iOS 18 will introduce enhanced AI capabilities on Apple devices, with ChatGPT handling complex tasks and on-device AI models managing simpler tasks.

This partnership is not financially driven. Instead, OpenAI gains access to millions of Apple users, while Apple benefits from integrating ChatGPT, potentially offering a paid version through App Store fees. Apple is also negotiating with other AI developers, including Google, Anthropic, Baidu, and Alibaba Group Holding, to explore further chatbot integrations. The new AI features, branded as “Apple Intelligence,” will launch first in American English, with plans for an international rollout.

OpenAI has faced controversies, including the temporary removal of CEO Sam Altman and criticisms regarding AI safety. For Apple, securing an observer role in a high-profile partner like OpenAI underscores the importance of this partnership.

Apple’s observer role on OpenAI’s board and the integration of ChatGPT into its devices represent a major advancement in Apple’s AI strategy. This collaboration enhances AI capabilities on Apple devices, providing users with advanced features and setting the stage for future AI innovations.

OpenAI Elevates Vodafone's Chatbot, Marking a Setback for IBM

TOBi, the familiar virtual assistant, has been upgraded with OpenAI's generative AI (GenAI) technology and is now running on Microsoft Azure. Known as SuperTOBi, this enhanced version can interpret entire sentences and phrases, overcoming the limitations of traditional chatbots that can only respond to simple questions based on a few keywords. It also offers more natural conversation capabilities.

SuperTOBi is being deployed across Europe, already live in Italy and Portugal, with launches in Germany and Turkey expected later this month. In Portugal, where SuperTOBi is used for appointment booking, the first-time resolution rate has improved from 15% to 60%, and Vodafone's online net promoter score (NPS) has risen by 14 points to 64.

Every superhero has an origin story. TOBi's began in 2017, designed to provide instant responses to customer queries, avoiding the wait for a human assistant. Initially, TOBi, powered solely by IBM's Watson AI technology, understood 90% of queries and could interact conversationally and manage customer accounts. However, the rapid evolution of the GenAI market in recent years has seen Vodafone switch to OpenAI for SuperTOBi, highlighting IBM's declining influence.

IBM's Watson gained fame in 2011 by defeating two champions of the US quiz show Jeopardy, bringing terms like 'natural language processing' into tech conversations. However, as detailed in a 2021 New York Times feature titled "What Ever Happened to IBM's Watson?" IBM's ambitious focus on using Watson for cancer treatment revealed the technology's limitations, diverting IBM's attention and resources.

This shift allowed other players, notably OpenAI, to take the lead in the generative AI market. Today, OpenAI, led by CEO Sam Altman, is seen as a key driver of GenAI technology. Although IBM continues to offer Watsonx, a portfolio of GenAI solutions for developers and enterprises, it's a far cry from Watson's Jeopardy triumphs. Now, it appears even SuperTOBi cannot restore Watson's former glory.

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 organizations 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 realization of AGI. Stay tuned for exclusive interviews, expert opinions, and much more!

Keep reading