Mark Zuckerberg announced in April that the company would significantly increase its capital expenditures in the coming year to maintain its position in the highly competitive field of artificial intelligence. He followed through with this commitment last week by establishing a $15 billion “AI superintelligence” team, which is rumored to include nine-figure salaries and a 49% investment in Scale AI. Among the notable hires was Scale’s 28-year-old founder, Alexandr Wang, a former roommate of OpenAI’s Sam Altman.

Prior to Meta’s investment, Scale already had most of the top players in AI as clients, and some were apprehensive about the implications of this deal. As Bloomberg summarized, “Scale AI’s Wang Brings to Meta Knowledge of What Everyone Else is Doing.” Google, Scale’s largest client, expressed concerns, reportedly indicating that their partnership would end due to this development, according to Reuters.

A Silicon Valley analyst characterized the announcement as a strategy of “a wartime CEO.” Superintelligence refers to an AI that can outperform humans across all tasks, while the current state of AI has yet to reach artificial general intelligence (AGI), which denotes a human-level performance across a broad spectrum of activities. Meta’s efforts to leapfrog the current advancements and focus on superintelligence appear to be an attempt to reclaim its leadership position in AI following significant strides made by competitors like OpenAI and Google, as well as the lackluster return on Meta’s substantial investment in the Metaverse.

While the scale of this move seems to mark a decisive entry into the AI race, Meta is still catching up, as its latest AI models fall short compared to its competitors. In April, the company released a model tailored to gaming benchmarks, indicating a level of urgency. With its massive investment and the formation of a high-profile team, Meta is effectively buying its position in the AI landscape.

The Week in AI News

• Thousands of UK university students caught cheating using AI
• Disney and Universal sue AI image creator Midjourney for copyright infringement
• Can ChatGPT assist in managing your finances?
• Researchers develop AI tool that restores age-damaged artworks within hours
• Keir Starmer addresses technology’s potential for a ‘better future’ amidst AI fears

Wikipedia’s Response to AI Integration

Last week, Wikipedia halted a test of an AI-driven article summarizer following backlash from its editors, who contribute to the encyclopedia. The AI-generated summaries were to be displayed at the top of pages, preceding introductions crafted by humans. The trial lasted two weeks and was available to around 10% of users opting into the mobile version of Wikipedia.

The feedback received was overwhelmingly negative, according to the nonprofit Wikimedia, with one editor describing the initiative as “a truly ghastly idea.” Many expressed concern over potential inaccuracies and the impact on Wikipedia’s credibility. A consensus among editors leaned towards keeping AI out of the platform, fearing that such mistakes would undermine its reliability.

“We’re grappling with the balance between AI and Wikipedia. While some minimal AI integrations already exist, the question remains: ‘What distinguishes human-generated content from robotic output?’” said Pacita Rudder, executive director of Wikimedia New York City. Discussions on this topic are ongoing within the Wikipedia community.

In contrast, Reddit has opted for a more commercial approach, having begun charging for access to its application programming interface (API) earlier this year. Reddit’s extensive archive of posts benefits AI companies that require large datasets for training models like ChatGPT. This move has led to user dissatisfaction, resulting in widespread protests among moderators and users alike. Despite user opposition, Reddit executives chose to prioritize financial gains over community concerns.

Shortly after Wikipedia canceled its AI test, the Wikimedia Foundation organized an in-person “edit-a-thon” at the United Nations, focusing on the history of the organization. The activity, part of the UN’s Open Source Week, encouraged collaboration among both new and seasoned editors, showcasing the proactive engagement within the Wikipedia community.

Rudder emphasized the importance of building community connections: “It’s community, right? There’s a misconception that a lot of Wikipedia editors are alone in their basements, typing away, but that’s not the case.” She also highlighted how this event marked a vibrant gathering in contrast to the typical solitary editing experience.

Wikipedia’s vast text corpus plays a fundamental role in training the majority of large language models. While it has a revenue-generating model through Wikimedia Enterprise, it remains a non-profit. Unlike Reddit’s aggressive monetization strategy, Wikipedia emphasizes human contributions and community feedback, as evidenced by its decision to roll back certain AI features.