🗞 OpenAI Addresses Safety Concerns with New Committee as Next-Gen AI Model Looms

🔎 The Latest on the AI Frontier:

• OpenAI Forms Safety Committee Amid Concerns Over AI Risks 🔍

• Ex-OpenAI Board Members Support Altman's Ouster, Cite Toxic Culture 🚨

• Elon Musk's xAI Secures $6B Funding, Aims to Rival ChatGPT 💰

• Musk and Meta AI Chief Clash on X Over Politics and Research 🥊

• Apple to Reveal Practical AI Features and OpenAI Partnership at WWDC 🍏

• Lessons from 19th-Century Telegraph Operators for Human-AI Interaction 🤖

• EU Taskforce Offers Preliminary Insights on ChatGPT's GDPR Compliance 🇪🇺

🔍 OpenAI establishes new safety committee and confirms training of its next large language model amid concerns over long-term AI safety. Read More

• The safety and security committee, led by outside chairman Bret Taylor and other board members, aims to evaluate and improve OpenAI's processes and safeguards within the next 90 days.

• OpenAI CTO Mira Murati previously stated that a major update to the underlying model, succeeding GPT-4, will be unveiled later this year, with Microsoft CTO Kevin Scott suggesting it will be substantially larger than its predecessor.

• The moves come after the resignations of key personnel, including co-founder Ilya Sutskever and Jan Leike, who led the company's long-term safety work, citing concerns over the lack of support for their efforts.

🚨 Ex-OpenAI board members stand by CEO Sam Altman's dismissal, citing concerns about a "toxic culture of lying" and psychological abuse. Read More

• Helen Toner and Tasha McCauley, former OpenAI board members, believe that third-party legislation is crucial to protect the company from its own excesses.

• Following Altman's return, the departure of senior safety-focused talent has raised doubts about the future of OpenAI's self-governance experiment.

• OpenAI is currently under scrutiny for allegedly using Scarlett Johansson's likeness without her consent in an attempt to give ChatGPT her voice

💰 Elon Musk's xAI raises $6 billion in funding to compete against ChatGPT and other AI companies, with a reported valuation of $18 billion. Read More

• The funding will be used to bring xAI's first products to market, build advanced infrastructure, and accelerate research and development of future technologies.

• xAI has launched Grok, an AI chatbot currently available only to X Premium subscribers on the platform formerly known as Twitter.

• Musk reportedly told investors that xAI plans to build its own AI supercomputer by late 2025, which would require 100,000 of Nvidia's current H100 chips, each costing $30,000 to $40,000.

• Tesla shareholders will vote this week on whether to restore Musk's $56 billion pay package ahead of the company's annual meeting on June 13th.

🥊 Elon Musk and Meta AI chief Yann LeCun engage in a heated argument on X, with LeCun criticizing Musk's "vengeful politics" and "conspiracy theories." Read More

• LeCun expressed his dislike for Musk's politics and hype, despite appreciating his cars, rockets, solar panels, and satellite network.

• Musk questioned LeCun's scientific contributions over the past 5 years, to which LeCun responded that he had published over 80 technical papers since January 2022.

• The argument ended with Musk implying that LeCun was "just following orders" and LeCun asserting that Musk doesn't seem to understand how research works.

🍏 Apple to unveil its generative AI strategy at WWDC 2024, focusing on practical features and a potential partnership with OpenAI. Read More

• Apple's AI approach will prioritize practical improvements, such as a "smart recap" feature to summarize missed texts, notifications, web pages, and other media.

• The Voice Memo app may receive AI-generated transcripts, while on-device Spotlight search, Safari internet searches, and writing suggestions for emails and texts could also see AI-powered enhancements.

• Apart from AI features, iOS 18 may introduce the ability to change app icon colors and allow users to place app icons freely on the home screen, moving away from the current top-to-bottom, left-to-right arrangement.

🤖 What Mark Zuckerberg can learn from 19th-century telegraph operators about the potential and pitfalls of human-AI interaction. Read More

• The spread of the telegraph in the mid-19th century enabled real-time conversations at a distance, raising concerns about catfishing and the ability to form genuine connections sight unseen.

• As large language models improve, some suggest that conversational AI could help combat the "loneliness epidemic," but the history of the telegraph suggests that disembodied communication brings new challenges in establishing authenticity and trust.

• Creating chatbots in the image of human conversation may lead to unhealthy dependence and self-deception, so we need to reevaluate these mediated interactions rather than seeing them as direct replacements for human contact.

🇪🇺 The EU's ChatGPT taskforce offers initial insights into the AI chatbot's GDPR compliance, but key legal issues remain undecided. Read More

• The taskforce suggests that OpenAI needs a valid legal basis for all stages of personal data processing, and that "adequate safeguards" could help tilt the balance in favor of the company when relying on legitimate interests.

• On fairness and transparency, the report emphasizes that OpenAI cannot transfer privacy risks to users and must properly inform them about ChatGPT's probabilistic output and potential biases or inaccuracies.

• The taskforce's existence may be influencing GDPR enforcements on OpenAI by delaying decisions, as some data protection authorities appear inclined to wait for the working group's final report before taking action.

Have any feedback? Send us an email