AI Frontier Weekly | 4th June 2024

OpenAI exposes criminals:

Over the past three months, OpenAI researchers have found and banned five covert Influence Operations (IOs) that were using ChatGPT to spread disinformation, in an attempt to manipulate public opinion.

Key Points:

• The IOs originated from Russia, China, Israel, and Iran, and several operators were already known by authorities: In March, the US treasury sanctioned 2 Russians behind one of the campaigns.
• The malicious actors were using ChatGPT to create and post propaganda content (translating it into different languages), conducting open-source research, and debugging simple code.
• None of the IOs gained traction or reached large audiences, and were discovered quickly because human error exposed the content they were posting was generated by AI.

Why you should care: These particular campaigns didn’t make much impact, but it shows how malicious actors are exploiting AI tools, like ChatGPT, to scale up propaganda campaigns and change public opinion, putting pressure on companies, like OpenAI, to integrate stricter guardrails and protocols that restrict the use of their tool. Source.

Google’s Magic Leap Forward:

Google has formed a strategic partnership with Magic Leap—a leader in Augmented Reality (AR) devices—a collaboration designed to explore new possibilities in AR.

Key Points:

• The partnership will combine Magic Leap’s AR experience with Google’s tech platforms to “foster the future of the XR [Extended Reality] ecosystem with innovative product offerings.”
• Magic Leap will benefit from the collaboration by getting its AR technology out to a wider audience, and Google will get to redefine its place in the AR world.
• This partnership aligns with Google’s Project Astra, which they demoed at the I/O conference, which allows people to interact with an AI chatbot, via a pair of smart glasses.

Why you should care: This isn’t Google’s first foray into AR: They were forced to discontinue Google Glass—voice and motion-controlled smart glasses—following privacy concerns, high costs, and poor functionality, so this partnership could produce some interesting possibilities. Source.

OpenAI lands its biggest customer:

PricewaterhouseCoopers (PwC)—one of the ‘big four’ accounting firms—has signed an agreement with OpenAI to give ChatGPT Enterprise (the business version of ChatGPT) to its employees and clients.

Key Points:

• PwC will give ChatGPT Enterprise licenses to its 100,000+ employees across the US and UK—making it OpenAI’s biggest customer to date—as part of its $1B commitment to AI.
• It will also use its “first-hand experience of AI transformation” to “help clients leverage ChatGPT Enterprise for better and faster ways of working,” having already identified over 3,000 use cases.
• The company will develop custom GPTs to complement its audit, tax, and consulting services, and help its workforce review tax returns, generate proposal responses, and generate reports.

Why you should care: The agreement between OpenAI and PwC marks the first time OpenAI has agreed to a resale model, as it tries new revenue-making avenues, in the face of huge computing bills. Source.

NVIDIA's 2026 AI breakthrough:

During a speech at the Computex trade show in Taipei, NVIDIA CEO, Jensen Huang, revealed its Blackwell AI chip successor (announced just a few months ago): 'Vera Rubin', which is expected to launch in 2026.

Key Points:

• Although details are light, Rubin will have new graphics processors (GPUs) and central processors (CPUs) designed to power AI applications with high-bandwidth memory.
• Alongside the introduction of Rubin, Huang also announced that NVIDIA will upgrade its AI accelerators every year, revealing that they’re launching the Blackwell Ultra chip launch in 2025.
• Huang said NVIDIA’s chips will produce 98% cost savings and use 97% less energy than “traditional computing methods” that can’t process the data needed to power AI systems.

Why you should care: Announcing a new next-generation chip and accelerating its release schedule from every two years to every year, it's clear that NVIDIA is making a bold move to secure its dominance over the AI chip market, which it currently has an 80% share of. Source.

iOS 18 Brings an AI to Apple's Siri:

With a major AI boost coming in iOS 18, Apple is preparing to radically alter Siri, making speech interactions with iPhone apps even more natural and intuitive.

• Improving Siri's Voice Control Capabilities: In the near future, users will be able to command a variety of Apple app functionalities just by speaking to Siri.
• Siri's ability to multitask will be enhanced in future versions. For example, she will be able to summarize a meeting and send it off all at once.
• Apple is putting a lot of effort into developing AI capabilities that can be executed directly on the iPhone. This will improve efficiency and privacy by utilizing the device's strong on-device CPU.

If Apple's forthcoming Siri upgrade delivers on its promises of improved usability and interactivity, it might revolutionize voice-controlled assistants. This change has the potential to make using the iPhone in daily life easier and more natural. Source.

AI News

• OpenAI former safety leader Jan Leike?joins?Anthropic.
• The first AI “bilingual” brain-reading device decodes Spanish and English words.
• Google’s weird AI answers hint at a fundamental problem.
• OpenAI inks deal with accounting powerhouse PwC to resell ChatGPT to businesses.
• The Atlantic, Vox Media ink licensing, product deals with OpenAI.
• OpenAI sets up a safety?committee as it starts training its new model.
• According to a report, Apple plans to unveil new “Project Greymatter” AI tools at WWDC?2024.
• Elon Musk is feuding with “AI Godfather” Yann LeCun.
• How cyberscammers use AI to manipulate Google search results.
• Financial institutions face rising threat from sophisticated AI fraud.
• Inside China’s first AI hospital.
• Is the record stock market riding on questionable AI assumptions?
• How AI can boost cancer, depression, and perioperative care.
• The good and bad news about AI and jobs.
• EU watchdog says banks must take full responsibility when using AI.

Responsible AI is not just about liability — it's about ensuring what you are building is enabling human flourishing. Thomas Kurian (CEO at Google Cloud, 2023)