- The AI-ronman 🚀
- Posts
- Researchers create self-spreading AI malware 🧪
Researchers create self-spreading AI malware 🧪
The AI-ronman 🚀
Welcome to the forefront of AI innovation, where ideas meet impact. Get ready to be inspired by stories of breakthroughs and resilience in the ever-evolving world of artificial intelligence. 🌟
Quick Takes ⚡
ChatGPT feature updates
Inflection-2.5: Nears GPT-4 performance
Google engineer accused of stealing AI trade secrets faces indictment
Microsoft employee warns FTC about Microsoft's AI image generator
ArtPrompt bypasses safety measures to unlock malicious queries
Microsoft’s AI PCs are here!
Transform webpages into desktop apps with Chrome Canary's latest feature
Researchers create self-spreading AI malware
Meta AI introduces Searchformer for improving planning efficiency
Deep Dive 🔍
ChatGPT feature updates:
You can now enhance the security of your ChatGPT and API accounts by enabling multi-factor authentication. This feature can be activated through the 'Settings' menu in your account, providing an additional layer of protection.

A new feature dubbed "Upgraded response" is under experimentation for free users, displaying a response crafted by GPT-4 and nudging towards an upgrade to ChatGPT Plus.

ChatGPT now offers a "Read Aloud" feature, providing responses in five distinct voices across web and mobile platforms in 37 languages, enhancing multimodal AI interactions and user accessibility.

Inflection-2.5: Nears GPT-4 performance🏆
Inflection's latest update, Inflection-2.5, supercharges the Pi personal AI assistant with capabilities rivaling GPT-4, using only 40% of the training power. With a 94% performance match, enhanced coding and math skills, and a booming 6M user base, and an average chat length of 33 minutes. Inflection is redefining what personal AI can do, complete with real-time web searches.

Google engineer accused of stealing AI trade secrets faces indictment 🔏
Linwei Ding, also known by Leon Ding, a Google engineer, has been formally accused by a federal grand jury of misappropriating proprietary AI technology information, including Google's TPU chip designs and GPU specifications. Allegations suggest Ding was secretly aiding China-based entities, leveraging the stolen data to advance in the AI sector. The indictment details Ding's discreet transfer of critical data to a personal cloud account, circumventing detection mechanisms. Conviction on the trade secret theft charges could lead Ding to face severe penalties, including lengthy incarceration and hefty fines. Watch here.
Microsoft employee warns FTC about Microsoft's AI image generator 😶
Shane Jones, an AI engineer at Microsoft, voices grave concerns about Copilot Designer's capability to produce troubling imagery, citing violence and sexual content, and claims the company has dismissed his warnings and discouraged public discourse. Microsoft asserts its openness to employee input without addressing Copilot Designer's cited issues, spotlighting the ethical dilemmas in AI advancement.
ArtPrompt bypasses safety measures to unlock malicious queries 💣️
A method named ArtPrompt, devised by researchers, can circumvent the protective measures of large language models such as GPT-3.5 and GPT-4. This approach employs ASCII art in prompts to elicit responses on otherwise restricted subjects, by transforming sensitive terms into ASCII art, thereby bypassing built-in safety features. This reveals a critical susceptibility in AI safeguards, underscoring the possibility of exploitation.

Microsoft’s AI PCs are here! 💻️
Microsoft is setting a new standard with the introduction of its Surface Pro 10 and Surface Laptop 6, the first in its lineup of "AI PCs." These devices feature the powerful Intel Core Ultra/Qualcomm Snapdragon X Elite processors and are endowed with advanced AI capabilities. Users can enjoy natural language searches across various mediums, benefit from a smart, searchable activity log for personalized tips, and use text for image editing. Moreover, there is talk of an 'AI Explorer' feature being integrated into Windows 11 soon.
Transform webpages into desktop apps with Chrome Canary's latest feature
The next Chrome Canary update will introduce a capability that lets users convert webpages into standalone desktop apps, through a new "install page as app" option found in the webpage settings menu. For early access to this functionality, download Canary and activate the options at chrome://flags/#web-app-universal-install and chrome://flags/#shortcuts-not-apps.

Researchers create self-spreading AI malware 🦠
A groundbreaking discovery by researchers has led to the identification of "Morris II," an AI-driven malware. This "AI worm" zeroes in on widely-used Generative AI platforms and email aids such as Gemini Pro and ChatGPT 4.0, leveraging "prompt-injection" flaws for proliferation and data theft. Morris II ingeniously embeds self-duplicating prompts into standard text or graphic files to breach systems, potentially exposing private data like credit card and social security numbers. Notifications have been sent to Google and OpenAI to enhance protective measures.

Meta AI introduces Searchformer for improving planning efficiency
Meta AI's Searchformer merges Transformer models' versatility with traditional planning's structured approach, enhancing complex decision-making. Trained on simulated data and expert insights, it optimizes search strategies to outperform conventional methods by 27% in puzzle-solving efficiency.
GlazeGPT just announced their industry-leading innovation - text-to-MongoDB queries, a revolutionary feature that translates natural language into precise MongoDB queries.🔥This latest update empowers users to effortlessly interact with complex MongoDB datasets and generates complex queries in seconds while understanding your business glossary.
Level up your skills with this 14-day free trial - don't miss out! ⚡️
As we close this chapter, remember that every end is just a new beginning. Keep the spirit of inquiry alive, and we'll continue to bring you the best of AI. Stay tuned, stay inspired! 🌌
Ciao for now!
(Author: Poonam 👧 )
Karan 😎 🚀
CEO, GlazeGPT.com