AI planning biological attack

Reading time: 5 minutes  

 Newsletter | News | Tools 

Today we will discuss-

• 😲U.S. restricts export of more AI chips to China

• 👨‍⚖️EU plans stricter rules for Gen AI models

• 💣AI chatbots could help in bioweapon attacks

• 😱9 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

Biden administration announces new restrictions on AI chip exports to China 

♨️News - The Biden administration has introduced new restrictions to block the export of advanced artificial intelligence chips, including those produced by Nvidia, to China.

The rules, which will go into effect in 30 days, are aimed at restricting China's access to advanced AI chips, which could be used in critical military applications. However, Commerce Department Secretary Gina Raimondo has emphasized that the intention is not to harm China's economy.

🧐What are these new rules though? 

• Previous restrictions blocked the sale of Nvidia's H100 chip, popular among U.S. AI companies like OpenAI, with Chinese firms buying a slightly slower version such as H800 or A800. The new rules will also ban the sale of these chips.

• Moreover, to prevent workarounds as previously seen in Nvidia's H800, A800 chips that were slightly slower but still had powerful computing capabilities, the new rules limit how much computing power a chip can contain within a set size.

• Export licenses for advanced AI chips in more than 40 additional countries will also be required. This, to prevent them from sending these chips to China.

• Furthermore, chips used in consumer products like game consoles or smartphones won't be restricted for export. However, if these chips are exceptionally fast, companies might have to inform the Commerce Department about their orders.

EU explores stricter rules for regulating Gen AI models to control potential misuse

☕News - The European Union is exploring a three-tiered strategy for regulating generative AI models, as per a recent proposal. This is part of their effort to control the rapidly advancing technology from being misused.

The plan involves setting different rules for various basic AI models under the three tiers. Additionally, it requires extra external testing for the most advanced technology, according to the document.

🪜What are these three tiers? According to the proposal, the first category covers all foundation models. The second tier, termed "very capable" systems, is distinguished by the substantial computing power used to train their large language models. The third category, general purpose AI systems at scale, includes widely used AI tools and will be assessed by the total number of users.

Although there will be tier-specific regulations for these models to follow, it is expected that transparency requirements will apply to models in all three tiers. All models, regardless of their tier, will need to undergo red-teaming, a method in which internal or externally hired experts test a model's security by simulating attacks to identify potential vulnerabilities.

AI chatbots could be used to design and produce bioweapons, report says

📰News - Artificial intelligence chatbots could be used to plan bioweapon attacks, according to a new report by the Rand Corporation, a US think tank. The report tested several large language models (LLMs) and found that they could supply guidance that could assist in the planning and execution of a biological attack.

However, the researchers acknowledged that they had to jailbreak or bypass the chatbot's safety features to get this information out.

🤩What were the key points?

• Previous attempts to turn biological agents into weapons failed because scientists did not understand the bacteria well enough. However, artificial intelligence could quickly fill in these knowledge gaps.

• The researchers clarified that this report was just an assessment, and their final report would examine whether the large language models' responses simply reflected information already available online.

• Lastly, the report emphasized the need for rigorous testing and self-regulation of LLMs to improve the safety standards.

Well, bioweapons, which are a major AI-related threat, will be discussed at next month's global AI safety summit in the UK.

👩🏼‍💻What else is happening?

• FBI and White House warn AI usage is 'new Cold War'

• UK will not seek global framework on AI for nations to sign up to summit

• Google defends use of public data for generative AI training

• Microsoft-backed research finds flaws in OpenAI's GPT-4

• AI gave tech giants a $2.4 trillion boost to their market caps in 2023

🤩AI life hacks

• AI slide deck - Creating slide decks for presentations can be a long process. But with the help of AI slide generators, you can create professional-looking presentations with ease. All you have to do is simply enter your topic, presentation goal, audience, and desired length. The AI will then generate a slide deck tailored to your needs.

• Automate to-do lists - Improve your task management skills by employing AI to work for you. Some AI-powered to-do list apps like Todoist can help you balance your work and personal life. By learning your habits, preferences, and priorities, these apps can provide personalized suggestions, categorizations, and reminders that can help you stay organized and productive.

• Efficient data analysis - Use AI-powered data analysis tools to save time and effort. Tools like Tableau and Power BI can automate complex calculations and visualize data trends, transforming raw data into valuable insights. This can free up your time to focus on more strategic tasks, such as interpreting the data and developing recommendations.

👩🏼‍🚒Discover mind-blowing AI tools

1. Podsnacks - A tool that provides curated summaries of podcasts

2. PhotoFairy - An AI-powered tool designed for photo editing and graphic design

3. Receiptor.ai - An automated tool that extracts receipts and invoices from emails, saving time and effort

4. AI Power - Uses advanced AI models like GPT-3.5, GPT-4 to generate customizable content, images, forms, and chatbots

5. CareerPen - A browser extension that creates personalized cover letters for job advertisements using LinkedIn profile details

6. RoomGPT - A tool that allows users to take a picture of their room and generate new versions of it in different themes

7. MealGenie - An AI tool that uses AI to help users discover vegan recipes, and also gives tips on how to make them

8. Dora - A web design platform that enables users to create stunning 3D and animated sites without coding

9. Sloped - A tool that helps users turn their data queries into shareable and easy-to-read dashboards

🕰️Historic Moments

• 2017: NVIDIA launched StyleGAN, revolutionizing image generation by giving control over styles and features. This milestone marked a leap in AI-generated images, enhancing quality and flexibility. Users could dictate style, resolution, and content, propelling AI-created visuals to new heights.

• 2018: Google's AI research team unveiled BERT, a game-changer in natural language processing. BERT's bidirectional approach transformed language understanding and generation, capturing nuances and context within texts, a leap that reshaped the field by enabling AI to comprehend language more comprehensively.

• 2020: Samsung's AI initiative NEON introduced lifelike AI avatars for versatile applications, signifying the potential for immersive virtual interactions across diverse sectors. NEON stood as the pivotal project, underscoring a substantial leap in AI-generated human-like avatars.