• The AI Collective Word
  • Posts
  • LLMs for Threat Intelligence, Impact of AI on Internet Safety, and AI Chatbots Carry Cyber Risks

LLMs for Threat Intelligence, Impact of AI on Internet Safety, and AI Chatbots Carry Cyber Risks

AI Cybersecurity, Regulations and Privacy Week in Review

Author

Anne Borsohalmi
September 04, 2023

Hello and welcome to today’s newsletter! 

The latest news covers:

  • LLMs for Threat Intelligence

  • Impact of AI on Internet Safety

  • AI Chatbots Carry Cyber Risks

  • MLSecOps Top 10

I hope you enjoy this week’s newsletter!

Please subscribe to The AI Collective Word today to receive your free newsletter directly in your inbox and join our ever-growing network.

Share it with a friend or colleague if you find it helpful.

RISK AND SECURITY MANAGEMENT

Created with Canva

Google will show how to use large language models (LLMs) for threat intelligence at Black Hat USA 2023. The session will explore how LLMs like Google PaLM and OpenAI's ChatGPT can help security teams analyze data and detect threats. AI was also a big topic at RSA Conference 2023, where many vendors launched generative AI products and features.

Tom Siegel, the co-founder and CEO of Trust Lab, is a former Google executive who led the trust and safety team. He started Trust Lab in 2019 to develop software that can detect harmful content on the internet, such as misinformation, hate speech, and AI-generated abuse. In an interview with TechTarget Editorial, he shared his views on the challenges and opportunities of AI and the internet and the role of tech companies and regulators.

Generative AI is a hot trend in IT security, with VMware and Cisco launching new products that use it. These products can help security analysts deal with threats faster and more effectively. However, generative AI is not perfect and still needs human oversight to verify its results.

The NCSC warned that AI chatbots could be hacked to do harmful tasks. They said that LLMs, are not fully secure and could be manipulated by attackers. They advised organizations to be careful when using LLMs and not to trust them with sensitive transactions.

The MLSecOps Top 10 - The Institute for Ethical AI and Machine Learning

The MLSecOps Top 10 is a project that identifies and demonstrates the top 10 security risks in machine learning systems. It is inspired by the OWASP Top 10 Report, but focuses on machine learning security. The project also provides best practices and open-source resources to mitigate these risks. The goal is to promote responsible and secure machine learning development and deployment.

REGULATIONS

Created with Canva

AI has both benefits and risks, depending on how it is used. Frontier models are the most advanced and expensive ones, and they may become more general and human-like in the future. This essay will discuss how to govern these models to align them with human intent.

AI is changing the world in unprecedented ways, attracting huge investments and raising regulatory concerns. From consulting to biotech, many industries are adopting AI to innovate and improve. However, AI also poses risks such as fraud, discrimination, and collusion. Sam Altman, a leading AI entrepreneur, testified before Congress that regulation is needed to mitigate these risks and foster collaboration between the public and private sectors.

Top tech CEOs will join a private AI event hosted by Senator Schumer on September 13. The event is part of a series of forums to discuss the challenges and opportunities of AI regulation. The participants will include representatives from civil society, national security and various industries.

The US Copyright Office is seeking public comments on how to deal with AI and copyright issues. It wants to know how AI models can use copyrighted data, whether AI-generated material can be protected, and who is liable for AI's actions. It also asks about AI's impact on publicity rights and unfair competition laws.

Google DeepMind co-founder Mustafa Suleyman urges the U.S. to impose ethical standards for AI use on buyers of Nvidia's chips. He says companies should follow the voluntary commitments made by leading AI firms to the White House in July. Suleyman is also the CEO of Inflection AI, a Microsoft-backed startup that launched an AI chatbot named Pi in May.

PLATFORM ENGINEERING

ChatGPT Enterprise is a new product that provides secure and fast access to GPT-4, a powerful AI assistant for work. It can help you with various tasks, such as writing, coding, analyzing data, and more. You can customize it for your organization and protect your data with encryption and compliance tools. ChatGPT Enterprise is used by many leading companies and is available for large-scale deployment.

Asana's report shows that AI is becoming a vital tool for work but needs proper guidance. Most executives trust AI to help them achieve their goals and are willing to pay more for AI-powered tools. Employees use AI for various tasks, from data analysis to brainstorming, and want more access to AI at work.

This article shows how to deploy machine learning models with ease using Docker, Kubernetes, GitHub Actions, and web frameworks like FastAPI, Streamlit, and Gradio. These tools create an ecosystem that supports rapid, efficient, and scalable machine learning applications. The article gives essential commands for these tools, useful for data scientists and developers alike.

This article shows how LoRA, a parameter-efficient method, was used to fine-tune an LLM for Dagster-specific technical support. They also share best practices for building a clean production ML pipeline for fine-tuned LLMs. Finally, they explain how they operate, update the model, and monitor the quality of its responses.

Duet AI is a generative AI assistant that works with many Google Cloud products. It can help you with coding, data analysis, and operations using natural language. Duet AI also provides context, citations, and suggestions for your work.

ML is a powerful technology that can improve various industries. How ML teams are organized is very important for their success. Different ML team models suit different company needs and stages.

Machine learning is a powerful way to innovate, customize, automate and compete in any industry. However, it requires a dedicated team of experts who can research, implement and improve ML solutions for specific business needs. A machine learning team can be formed by a strategic AI sponsor who has the vision, budget and authority to support the ML development process and the resulting intellectual property.

The Responsible Machine Learning Principles - The Institute for Ethical AI and Machine Learning

Responsible ML development is a way of creating AI systems that learn from data in a ethical and trustworthy manner. The 8 principles of responsible ML development guide technologists to design, develop and maintain such systems. They cover aspects such as fairness, privacy, security, accountability and transparency.

ML and NLP

Spark is a data processing platform for ML workloads that can run on Kubernetes. Running Spark on Kubernetes simplifies deployment, increases portability, and enables resource management. To deploy Spark on Kubernetes, users need to use container images and pods.

ETHICS

The ACM Code of Ethics and Professional Conduct is a document that expresses the values and responsibilities of computing professionals. It aims to guide their ethical behavior, support the public good, and prevent harm. The Code consists of four sections: fundamental principles, professional responsibilities, leadership obligations, and compliance requirements.

USE CASES

ServiceNow is a cloud computing vendor that offers generative AI capabilities for its customers. These include case summarization and text-to-code, which are powered by the vendor's own language models that are trained on enterprise data. ServiceNow also announced a partnership with Nvidia and Accenture to help enterprises develop their own generative AI solutions.

ChatGPT is the most popular AI technology of 2023, and many organizations want to use it for their purposes. PayPal's chief product officer, John Kim, talks about how his company is using AI to improve security and customer experience. He also reveals some of the upcoming AI-based products that PayPal will launch soon.

RESOURCES

This is a summary of a talk on AI trends, applications, opportunities, and challenges, given by Andrew Ng at Stanford GSB on July 26, 2023. He discussed the latest developments in supervised learning and generative AI and the adoption of AI across various domains and industries. He also shared his insights on how to build successful AI startups and the social impact of AI on society. Video - 36 min.

Windows on Snapdragon is a new way of developing AI applications that leverage the power and efficiency of Qualcomm's Snapdragon chip. In this course, you will learn how to use Windows on Snapdragon as a platform for AI and how to take advantage of the Qualcomm AI stack and the dedicated Neural Processing Engine. You will also get familiar with the tools and frameworks for building Arm64 applications on Windows and Ubuntu. By the end of this course, you will have a solid understanding of the fundamentals of Snapdragon computing and how to use it for AI development.

For: software developers and test engineers
Duration: 2 hours, self-paced

Thank you for reading! Please send me feedback and share the newsletter with others.

The AI Collective Word is sent every Tuesday and Thursday. Think your friend or colleague should know about us? Forward this newsletter to them. They can also sign up here.