• AI KATANA
  • Posts
  • Securing AI The Next Platform Opportunity in Cybersecurity

Securing AI The Next Platform Opportunity in Cybersecurity

Also: How AI facilitated ISRO achieve Chandrayaan 3 success?

Author

AI KATANA
August 27, 2023

Welcome!

In today's AI digest: As AI becomes foundational to enterprises, securing these systems is paramount. Analogous to the evolution of cloud security, there's a growing need for real-time, AI-specific security interventions. U.S. Commerce Secretary Gina Raimondo's China visit underscores the pivotal role of AI and semiconductors amidst heightened geopolitical tensions. The success of ISRO's Chandrayaan 3 mission can be attributed in part to robust AI integrations, from planning to lunar navigation. On the cultural front, the phenomenon of "Generative Inbreeding" highlights the risks AI poses to human creativity when trained on AI-generated content. Tool-wise, Meta's new AI model, Code Llama, promises to revolutionize coding, while U.S. startup KoBold Metals, backed by major investors, is using AI to pinpoint mineral deposits vital for EV batteries. Don't miss today's meme, an AI-generated image, and IHOP's AI-driven ordering experience.

Sliced

  • 🔒 Securing AI The Next Platform Opportunity in Cybersecurity

  • 🇺🇸 U.S. commerce chief begins China trip with focus on chips, AI

  • 🇮🇳 How AI facilitated ISRO achieve Chandrayaan 3 success?

  • 🌍 ‘Generative inbreeding’ and its risk to human culture

Securing AI systems has emerged as a top priority for many enterprise-level CISOs and CIOs due to the potential value foundation models offer. Drawing parallels to the cloud platform shift, AI represents a transformative wave in software development, where AI properties become central. However, AI introduces unpredictable behaviors since results are often stochastic, not deterministic. As AI applications evolve, potential security risks become apparent. These include the risk of non-deterministic outputs, potential vulnerabilities in AI software, and the increased likelihood of data misuse or leakage. Addressing these challenges, immediate necessities are to ensure visibility, governance, and auditability for foundation model-based solutions. Startups are in a unique position to innovate and offer security solutions at the crossroads of AI and cybersecurity. The evolution in securing AI is reminiscent of the trajectory cloud security took, indicating opportunities for both posture-related and real-time, runtime security interventions.

U.S. Commerce Secretary Gina Raimondo has embarked on a visit to China, aiming to foster dialogue concerning pressing high-tech sectors, notably semiconductors and artificial intelligence, amid escalating tensions between the two nations. Raimondo's tour, encompassing Beijing and Shanghai, underscores the Biden administration's approach of combining competition with diplomacy to prevent potential conflicts. Recent U.S. measures, such as export bans on advanced semiconductor technology to China and investment restrictions in sectors like AI, have heightened the stakes. In response, China has imposed restrictions on its firms purchasing from U.S. chipmaker Micron Technology and strengthened its export controls. Amid this backdrop of intensified rivalry, both countries highlight the necessity of dialogue to avert inadvertent clashes, with a potential summit being discussed for later in the year.

AI was pivotal to ISRO's success in the Chandrayaan 3 mission, playing a multifaceted role from planning to execution. ISRO's Chairman S Somnath highlighted the incorporation of AI in a state-of-the-art sensor array, which combined with velocimetry and altimeters, monitored the lander's descent. Additionally, sophisticated cameras were used for hazard avoidance and tracking, with AI algorithms consolidating data from both sensors and cameras to create detailed, high-resolution images of the lunar terrain, essential for precise landings. Beyond this, AI enabled autonomous lunar navigation, facilitated robotic exploration with the Pragyan rover, ensured predictive maintenance of the spacecraft, expedited vast data analysis, forecasted space weather, optimized mission planning, enhanced data transmission, and continuously monitored lunar environmental changes. The integration of these AI-powered systems and tools significantly augmented the mission's capability and success.

"Generative Inbreeding" is a concept that draws a parallel between the genetic inbreeding seen in biological populations and the potential degradation of AI systems as they increasingly train on AI-generated content. Just as biological inbreeding can lead to offspring with health issues and deformities due to increased expression of recessive genes, training AI on AI-generated content can lead to distorted representations of human culture and the potential breakdown of AI performance over time. Such a recursive process can cause AI systems to produce increasingly distorted cultural artifacts, potentially overshadowing authentic human creativity and influence. While measures like AI classifiers and watermarking are proposed to mitigate this issue, they come with their own set of challenges.

🛠️ AI tools updates

Code Llama, a new large language model (LLM) focused on coding. This advanced model, built on the Llama 2 foundation, facilitates coding by generating, discussing, and even debugging code based on textual prompts. Available in three different sizes with varying parameters, Code Llama caters to a range of requirements, from real-time code completion to more comprehensive coding support. Further specialized versions, namely Code Llama – Python and Code Llama – Instruct, provide enhanced support for Python coding and instruction-based tasks respectively. These tools aim to streamline developers' tasks, enabling them to focus on innovative and high-level aspects of their roles. Open-source and available to the community, Code Llama aspires to boost the development of tools that can change the landscape of the coding world and empower developers across various sectors.

💵 Venture Capital updates

KoBold Metals, a U.S. startup established in 2018, is leveraging an AI platform developed in collaboration with Stanford University to identify critical mineral deposits with increased accuracy. The system uses both private and public geological data from around the world to locate minerals, such as nickel and cobalt, essential for EV batteries, with a precision 10 times better than traditional methods. With the growing demand for these minerals due to the rise of electric vehicles, and with China currently dominating their production, efficient exploration methods like KoBold's can help diversify the supply chain. The startup recently secured $195 million in funding from investors including Bill Gates, Jeff Bezos, BHP, Mitsubishi Corp., and Equinor. KoBold's technology optimizes the mining process by reducing the need for extensive test drilling and increasing the chances of mineral discovery. The company is now partnering with mining giant Rio Tinto and is set to expand its operations using the recently acquired funds.

🫡 Meme of the day

⭐️ Generative AI image of the day