To help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.
- OpenAI is planning to restructure its business operations in order to remove non-profit control of the company, according to Reuters. The AI company, known for its ChatGPT AI chatbot, is working on a plan to transform its core business into a for-profit benefit corporation, which would no longer be controlled by its non-profit board. This move is expected to make OpenAI more attractive to investors and could end in a $150 billion valuation of the company after restructuring. As part of this change, CEO Sam Altman would reportedly receive equity in the for-profit entity for the first time, though CNBC reports that Altman, in an all-hands meeting this past week, said to employees that it was “just not true” and that he would receive a “giant equity stake.” The non-profit arm of OpenAI would continue to exist and maintain a minority stake in the for-profit company. The unusual corporate structure of OpenAI was intended to safeguard the company’s mission of “safe artificial general intelligence that is broadly beneficial,” though the recent changes are seen as an effort to attract further investment into the company. Indeed, Axios reported this past week that the investors behind a recent funding round of $6.5 billion included a stipulation that they could ask for their money back should OpenAI fail to make the governance changes within the next two years. This restructuring comes amid a tumultuous week for OpenAI, marked by the high-profile departures of key company leaders, including Chief Technology Officer Mira Murati and research chief Bob McGrew, which Altman has denied are related to the proposed corporate governance changes.
- Meta has unveiled its latest AI model, Llama 3.2, new celebrity voices for its Facebook, Instagram, and Whatsapp AI assistants, as well as a new line of AI-powered smart glasses at its annual Connect conference this past week, according to The Associated Press. The company showcased several AI updates, including live translation capabilities and the ability for influencers to create AI versions of themselves for fan interactions. The company, in a blog post, said that Llama 3.2 now has the capability to take information from documents and images and perform a variety of reasoning tasks. Meta also introduced a prototype of wearable technology called Orion, described by CEO Mark Zuckerberg as “the most advanced glasses the world has ever seen,” featuring holographic augmented reality. Users of the company’s Meta AI assistant on Facebook, Instagram and Whatsapp will also now be able to converse with the AI system, which will use a variety of different voices, including those of actors Judi Dench, Kristen Bell and John Cena, after the company struck a deal for their voices. The company also announced that Meta AI-generated content will begin appearing in Instagram and Facebook feeds based upon a user’s interests or current trends.
- The Japan Times reports on a new study that used an AI model to discover over 300 new geoglyphs in Peru’s famous Nazca desert. The study, conducted by researchers at Yamagata University and in collaboration with IBM Research, has doubled the number of known geoglyphs (large incisions made on the desert floor that depict animals, plans, and other entities when seen from far above) in the area, known popularly as the “Nazca Lines.” The AI-powered approach allowed scientists to analyze vast amounts of geospatial data more efficiently than traditional visual identification methods, leading to the discovery of 303 new etchings in just six months. These newly found geoglyphs include large linear designs depicting wild animals and smaller motifs of abstract humanoids and domesticated camelids. The research team, led by archaeologist Masato Sakai, suggests that the larger geoglyphs were likely created for community-level rituals along pilgrimage routes, while the smaller ones may have been used by individuals or small groups to share information about rituals and livestock. This breakthrough demonstrates the potential of AI in accelerating archaeological discoveries, even in well-studied sites like the Nazca Lines, which date back approximately 2,000 years and were only first discovered in 1927.
- Cloudflare, a company providing major “internet backbone” services such as web hosting, cybersecurity and DNS resolving, announced new tools to allow customers to detect and block AI data-scraping bots, according to WIRED. These new AI auditing tools, collectively called Bot Management and offered free-of-charge to all Cloudflare customers, include a real-time bot monitoring dashboard and an expanded bot-blocking service that enables users to selectively block or allow specific AI agents to access their site. Cloudflare CEO Matthew Prince, in an interview with WIRED, emphasized that these tools aim to give website owners more control over how AI bots use their content, regardless of their technical expertise or budget. The company plans to launch a marketplace for negotiating scraping terms with AI companies, potentially involving compensation. This initiative comes in response to growing concerns about AI companies ignoring or circumventing traditional bot-blocking methods like robots.txt files, as well as the use of scraped copyrighted content to train generative AI models. Prince views the new tolls as a necessary step to create a more sustainable ecosystem for content creators in the age of AI. The new tools have been well-received from publishers, including Nick Thompson, CEO of The Atlantic, though Prince said that AI companies have had more mixed reactions.
- NPR’s Morning Edition reports on recent studies of AI-powered analysis tools for reviewing police body-worn camera footage. The study, conducted by researchers at Stanford University, used an AI model to analyze video and audio of over 600 traffic stops conducted by the Oakland Police Department to evaluate the communication of the police officer with the driver. Jennifer Eberhardt, the lead researcher, said to NPR that the researchers “could look at the first 27 seconds of the stop, the first roughly 45 words that the officer spoke, and we could use this model to predict whether that driver was going to be handcuffed, searched or arrested by the end of the stop.” The team also said that the data could be used to analyze racial disparities in law enforcement. AI-powered body camera footage tools are also being used to evaluate officers’ performance. Another study from a team at the University of South Carolina showed that a police department’s use of Truleo, an AI tool that analyzes automated transcriptions of body camera videos to assess how officers perform, was associated with higher professionalism in officers’ interactions with the public. Truleo has not been without its opponents. The system was met with backlash after it was implemented by the Seattle Police Department without informing the police union, though the company has said that Truleo works best in departments where officers are aware of the AI and can watch how it assesses them.
- Microsoft has announced upgrades to “Recall,” an AI-powered feature for Windows PCs, according to Bloomberg. The feature creates a record of everything a user does on a PC by screenshotting everything seen or done on the PC’s display, uses AI models to recognize the text on the screen, and then stores the information in an SQLite database for search and retrieval. After cybersecurity researchers pointed out serious security flaws in the feature (as reported on by this roundup in June), Microsoft has implemented several new security measures, including allowing users to filter out specific apps or websites, enabling sensitive content filtering by default, requiring biometric authentication to access the data and storing the collected information in an isolated environment. David Weston, Microsoft’s vice president for enterprise and operating system security, stated that the company has designed these safeguards to protect against even the most sophisticated hackers. The revised version of Recall will be available in beta next month, but it will remain opt-in and only operate on certain AI-enabled PCs running Windows 11 dubbed as “Copilot+ PCs,” and will not be automatically installed on enterprise PCs without an opt-in from companies.
- James Cameron, director of the blockbuster motion pictures Titanic, The Terminator and Avatar, is joining the board of Stability AI, a company best known for its Stable Diffusion text-to-image generation AI model, according to The Hollywood Reporter. Cameron, known for pushing technological boundaries in filmmaking, said in a statement that he views the convergence of generative AI and CGI (computer-generated imagery) as the next wave of innovation in visual storytelling. Stability AI’s CEO, Prem Akkaraju, a former CEO of the visual effects studio WETA Digital, expressed excitement about Cameron’s involvement, saying in a statement that “Stability AI’s mission is to transform visual media for the next century by giving creators a full stack AI pipeline to bring their ideas to life. We have an unmatched advantage to achieve this goal with a technological and creative visionary like James at the highest levels of our company.” The announcement comes amid a complex relationship between Hollywood and AI, with some studios embracing the technology while others express concerns about copyright issues and possible threats to creative jobs.
[View source.]
