🐛Hungry worm targets AI

Reading time: 5 minutes  

 Newsletter | News | Tools | Grow My Tool  

Today we will discuss-

• 🤖Anthropic debuts Claude 3

• 🏑China to level AI playing field

• 🖥️New computer worm threatens AI models 

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

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

AI startup Anthropic launches its most powerful models yet

🤹🏻‍♂️News - On Monday, Anthropic unveiled Claude 3, a set of artificial intelligence models that the company says are its fastest and most powerful yet. The series comprises three models: Opus, Sonnet, and Haiku. 

✨Key points to note -

• The star of the lineup is Opus, which Anthropic claims is better than any other openly available AI system on the market, even outperforming OpenAI's GPT-4 and Google's Gemini Ultra on industry benchmark tests, such as undergraduate level knowledge, graduate level reasoning and basic mathematics.

• Sonnet, positioned as the mid-range model, offers businesses a cost-effective solution for routine data analysis and knowledge tasks, delivering high performance without the premium price tag of the flagship model.

• Meanwhile, Haiku is tailored to be fast and economical, making it suitable for applications such as consumer-facing chatbots, where quick responses and cost efficiency are key considerations.

• Sonnet and Opus are now available in 159 countries, with Haiku set to be released soon.

🌞What else is new?

• Anthropic has introduced multimodal support within the Claude 3 series, allowing users to upload various types of unstructured data such as photos, charts, and documents for analysis and responses.

• According to Anthropic co-founder Daniela Amodei, Claude 3 can now summarize up to 150,000 words, which is equivalent to the length of "Harry Potter and the Deathly Hallows," doubling its previous version's capacity.

• Amodei noted that Claude 3 demonstrates improved risk understanding in its responses compared to its predecessor, Claude 2, which tended to be overly cautious, especially with sensitive topics and safety concerns.

• According to the company, Claude 3 also has an improved understanding of prompts.

China fosters fair play for AI startups amid tech dominance, chip shortages

🤓News - China is implementing measures to create fair competition opportunities for its AI startups. This comes as the country's tech giants have hogged AI-training computing resources, which are further limited due to US chip restrictions.

🤔What's their plan? Seventeen city governments, including Shanghai, have pledged to provide "computing vouchers" to help AI startups that are facing rising data center costs amid shortages of crucial chips.

According to official statements, the vouchers will range in value from $140,000 to $280,000. They are intended for use in AI data centers to facilitate the training of the companies' large language models (LLMs). 

🤕Easier said than done? A government official involved in data center construction stated that the vouchers would lower computing costs for AI companies by approximately 40 to 50%.  They noted that although this will "ease financial burden on startups" temporarily, there are still strict criteria for applicants, such as a minimum revenue threshold or involvement in a government-sponsored research project.

Moreover, while the initiative reflects the dedication of Chinese governments to AI technology,  it does not address the actual problem - scarcity of resources.

New AI worm can steal your data and compromise popular chatbots

😳News - Various experts have highlighted the risks of using generative AI in cybersecurity, warning that its adoption may lead to a surge in advanced social engineering attacks. 

Researchers have now reportedly developed a computer worm to target generative AI-powered applications, aiming to raise awareness about this potential threat.

📖For context - Worms are a type of malware that can spread across computer networks on their own, often without needing any input from users.

🕵🏻‍♂️What did the researchers find? According to the researchers, the worm, named Morris II (inspired by the first internet worm launched in 1988), could send spam emails and steal data from users without them having to click on any malicious links, a form of malware known as zero-click malware.

The researchers tested their worm against popular AI models like Gemini Pro, ChatGPT, and LLaVA to demonstrate how it could be used for harmful purposes.

They claimed their worm could infiltrate groups of generative AI systems using self-replicating prompts. By using jailbreaking and machine learning techniques, the researchers suggested the worm could exploit connections between these systems to spread malware or spam emails.

🥸What's the conclusion? In the experiment, the AI worm targeted generative AI email assistants, stealing email data and sending spam. However, the study suggests that attackers can use such worms to manipulate AI models into generating harmful content, enabling them to carry out all kinds of malicious activities.

🙆🏻‍♀️What else is happening?

• India asks tech firms to seek approval before releasing 'unreliable' AI tools

• House Republicans demand Google reveal what role US government had in developing woke Gemini AI

• Nvidia CEO says AI could pass human tests in five years

• AI outshines humans in creative thinking

• 41% of architects now using AI, says RIBA report

🤩Interesting facts

• The "CloudPainter" team, consisting of both a human artist and an AI, collaborated to create a unique masterpiece in the 2018 RobotArt competition. The human artist initiated the painting, while the AI, trained in various art styles, offered creative directions, resulting in a one-of-a-kind collaborative creation.

• OpenAI, in the year 2019, developed a robotic hand called Dactyl that was capable of solving a Rubik's Cube.

• A self-driving Uber car fatally struck a pedestrian in Tempe, Arizona in 2018. The incident showcased the limitations of self-driving technology, triggered public skepticism, and spurred debates on regulations, ethics, and ongoing research for safer autonomous vehicles.

👩🏼‍🚒Discover mind-blowing AI tools

1. Arcwise - An AI-first data analytics platform in your Google Sheet. Instant insights, zero CSV imports, one-click reports (Free Trial)**

2. Frequentli - An app that simplifies the process of creating FAQs for websites (Free)

3. VEED - An online video editing platform that helps users create professional-quality videos quickly and easily ($290/month)

4. FaceFiesta.io - Allows users to create personalized stickers and emojis based on their face (Free)

5. CardZap - An app that allows users to create and share professional business cards (Free)

6. Launch Teddy - An AI assistant designed to provide specialized support for product launches (Exclusively available to ChatGPT Plus users)

7. Ask Abe - A helpful information assistant designed to answer legal questions based on the California Legal Code (Free) 

8. Banner GPT - An AI tool that generates banner images for blog posts (Free)

9. AliceCam - An AI-powered tool that allows users to create professional-looking headshots without a real photoshoot ($11.99 one-time payment) 

💣Mythbuster

Myth - AI can solve any problem

Fact - AI has the potential to solve a wide range of problems, but it isn't a universal solution for every problem. While AI systems are well-suited for tasks such as pattern recognition, and data analysis, there are many problems AI may not be able to solve, especially those that require common sense, the ability to reason, and ethical judgement.

[The tool marked with ** is a sponsored feature.]