ChatGPT Enterprise usage high as Cisco eyes $350 million deal

Key Takeaways

• Alexis Martinez-Arizala was arrested for using AI deepfake videos to mislead individuals, including a Seminole County deputy and a shopper.
• Experts warn that AI deepfake technology poses significant challenges for public safety and law enforcement due to its realism.
• Japanese company CyberAgent achieved a 93% monthly active usage rate for ChatGPT Enterprise and Codex, enhancing efficiency and decision-making with strong security.
• Cisco Systems is reportedly in talks to acquire Astrix Security, an Israeli AI cybersecurity startup specializing in monitoring AI agents, for $250 million to $350 million.
• The Gemini app now offers interactive simulations and models directly within chats, allowing users to adjust variables and explore complex topics.
• The primary bottleneck in AI development is not models, but specialized hardware infrastructure like GPUs and data centers, leading to increased spending and long lead times.
• Startup Onix launched "Personal Intelligence," offering paid access to AI versions of human experts, trained by the experts themselves and storing data on user devices for privacy.
• AI's impact on journalism jobs and the broader writing industry is a growing concern, as highlighted by recent layoffs.
• There is a call for mandatory AI Basics courses in colleges and high schools to promote AI literacy and address skepticism.

AI deepfake video used to trick shopper

A woman named Melanie Valentine was tricked by an AI-generated deepfake video at a Home Depot in West Palm Beach. A man showed her a realistic video of her truck being stolen, trying to get her to follow him outside. This incident is part of a pattern where a content creator, Alexis Martinez-Arizala, has targeted customers with fake videos. Martinez-Arizala has since been arrested and faces charges after a similar prank targeted a Seminole County deputy. Experts explain that AI can create convincing fake videos and audio with just a photo and voice sample, raising concerns about public safety and privacy.

Man arrested for AI deepfake prank on deputy

Alexis Martinez-Arizala, 22, was arrested in Puerto Rico for allegedly using an AI deepfake video to mislead a Seminole County deputy. On March 24, Martinez-Arizala showed the deputy a fake video of a car break-in, claiming it was happening to the deputy's vehicle. Investigators confirmed the video was AI-generated and the incident did not occur. Experts warn that deepfake technology poses challenges for law enforcement, as it can create realistic content that requires split-second decisions. Martinez-Arizala also posted the video online, and faces a $7,000 bond upon extradition.

CyberAgent boosts speed with ChatGPT Enterprise and Codex

Japanese company CyberAgent is using ChatGPT Enterprise and Codex to improve efficiency and decision-making across its businesses. The company views AI as a foundational technology and has invested in AI research since 2016. By adopting ChatGPT Enterprise, employees can confidently use AI for tasks like research and drafting, supported by strong security. Codex helps teams work faster in design discussions and documentation. This adoption has led to 93% monthly active usage of ChatGPT Enterprise within the company.

CyberAgent achieves high AI adoption with secure tools

Japanese internet company CyberAgent has achieved a 93% adoption rate for ChatGPT Enterprise by focusing on security and fostering a supportive culture. The company sees AI as a core technology, investing in its development through an AI Lab and an AI Operations Office. ChatGPT Enterprise provides enterprise-grade security, allowing employees to use AI confidently while adhering to internal guidelines on confidential data. Codex is also being used to enhance decision-making and speed up tasks like code review and documentation. This organic growth is supported by training and internal initiatives.

Cisco may buy AI security startup Astrix for $250M

Cisco Systems is reportedly in talks to acquire Astrix Security, an Israeli cybersecurity startup, for $250 million to $350 million. Astrix specializes in software that monitors and secures AI agents, which are increasingly used by companies. This potential acquisition would strengthen Cisco's cybersecurity offerings in the growing field of AI security. The deal would represent a significant premium over Astrix's previous valuation. Neither company has officially commented on the potential acquisition.

Gemini app now creates interactive simulations

The Gemini app now allows users to generate interactive simulations and models directly within chats, transforming complex topics into explorable visualizations. Previously, responses were mostly text with static diagrams. Users can now adjust variables like gravity or velocity to see how they affect systems, such as a moon's orbit around Earth. This feature is rolling out globally to all Gemini app users, enhancing understanding through hands-on exploration. It is not yet available for Education and Workspace accounts.

AI's impact on journalism jobs debated

Austin Harvey, a writer, reflects on how AI may have contributed to his recent layoff, impacting readership at his former outlet. He shares his journey into journalism, finding passion in writing about mental health and later working for ATI. Harvey details his experience writing over 1,200 articles and hosting podcasts, aiming to make himself invaluable. The article touches on the broader impact of AI on the writing industry and the search for stability in a career.

AI hardware, not models, is the main bottleneck

The current focus on AI models overlooks the critical infrastructure needed to train and operate them, such as specialized hardware and data centers. Major tech companies are significantly increasing their spending on this infrastructure, indicating a tight market for compute power. Building new data centers and acquiring advanced chips involves long lead times and complex supply chains. Access to this hardware, including GPUs, is becoming a major constraint, influencing research agendas and the pace of AI innovation. This scarcity affects costs and competitiveness, as organizations must plan around available capacity rather than their ideal timelines.

Startup Onix offers AI versions of experts for a fee

A new company called Onix is launching technology called Personal Intelligence, allowing users to pay for access to AI versions of human experts. These AI bots store information on the user's device for privacy and are trained by the experts themselves. While designed to minimize hallucinations and stay on topic, early testing showed some instances of the AI deviating from the subject. Onix is starting with 17 vetted experts, primarily in health and wellness, and emphasizes that these AI interactions are not a substitute for professional medical treatment. The company aims to augment users' ability to explore topics thoughtfully.

Share your AI use in Open Thread

This week's Racket Open Thread invites readers to discuss how they use, or choose not to use, artificial intelligence. The prompt acknowledges the widespread marketing of AI since ChatGPT's launch in late 2022. Readers are encouraged to share their experiences, whether they see AI as a tool for improving life, a path to freedom from labor, or a concerning technology that erodes creativity. As always, participants are free to discuss any topic they wish.

Axios C-Suite: Mandate AI education for all

Axios suggests that all colleges and ideally high schools should mandate an AI Basics course starting this fall. This course should cover different AI models, prompting, agents, and basic safety and ethics. The author expresses concern over the skepticism and fear surrounding AI among students, emphasizing that while caution is important, ignorance or refusal to use AI will hinder future job prospects. The message is directed at CEOs and their teams to step up and promote AI literacy.

Sources

• ‘It was very real:’ Shopper targeted with A.I. deepfake video
• Man arrested in Puerto Rico for using AI deepfake to mislead Seminole County deputy
• CyberAgent moves faster with ChatGPT Enterprise and Codex
• CyberAgent's AI Leap
• Cisco in Talks to Buy AI Security Startup Astrix for at Least $250 Million
• The Gemini app can now generate interactive simulations and models.
• Did AI kill my job, or open up a next chapter?
• AI’s Biggest Bottleneck Isn’t the Model
• This Startup Wants You to Pay Up to Talk With AI Versions of Human Experts
• Tell Us How You Use—Or Don't Use—AI in This Week's Open Thread
• Axios C-Suite: AI education for all