UK, US, EU and China sign declaration of AI’s ‘catastrophic’ danger

Today's issue is brought to you by our sponsor - Koala. Accelerate Your Content Creation with this (better than ChatGPT) AI Tool!

📰 Quick News 📰

🏢 Vice President Kamala Harris recently announced the establishment of the United States AI Safety Institute. This institution will create guidelines, benchmark tests, and best practices for potentially hazardous AI systems, as well as provide technical guidance on AI-related issues.

🧑‍🏭 LinkedIn's growth spiked above the one-billion-user benchmark, introducing an artificial intelligence-enhanced chatbot to empower job seekers. It aids in evaluating job postings, framing applications, and recognizing any experience gaps, so finding your perfect job becomes a less stressful and more rewarding journey.

🎨 Stability AI's updated Stable Diffusion platform now features advanced 3D modeling capabilities and image fine-tuning tools like Sky Replacer and Stability Fine-Tuning. This can influence multiple sectors, from graphic design to gaming, by providing sophisticated content creation tools.

📖 The Collins Dictionary has announced "AI" as its 'word of the year' for 2023. The usage of the term AI has reportedly increased four-fold this year, reflecting its growing impact that's as significant as technologies like email and streaming.

🧠 Researchers have developed a physical neural network using silver nanowires that function as a 'brain'. Unlike energy-intensive computer-based AIs, this groundbreaking technology demonstrates real-time learning much like the human brain, with significantly lower energy consumption. It effectively processes and transmits information and remembers patterns showing its potential in neuromorphic computing applications.

⚕️DeepMind's upgraded AlphaFold can now generate predictions for structures of nearly all molecules in the Protein Data Bank. Using this, Isomorphic Labs is exploring therapeutic drug designs, characterizing key molecular structures for disease treatment and surpassing conventional methods like docking simulations in structural characterization.

⚙️ Intel has confirmed the unveiling of its Core Ultra CPUs and 5th-generation Xeon Scalable processors at an "AI Everywhere" event on December 14. Noteworthy is the significant AI-related capabilities of these chips which were built with different manufacturing technologies and targeted primarily at laptops.

📚 Nerd section 📚

🗃️ The Illusion of Data Ownership - on the internet, owning your data is a myth. We might possess our data, but we don't own it. This means we can't fully control its use or decide who profits from it. Decentralized technologies like Web5 could usher in a new era of true data ownership.

📃 From Good to Great: How Pre-processing documents supercharges AI’s output. Document pre-processing is a strategy that greatly augments the performance of Language Learning Models (LLMs). Through a technique called Retrieval-Augmented Generation (RAG), these models can efficiently process large documents and accurately locate and retrieve relevant information.

🌐 Web section 🌐 

So here (but not in the email version) we're adding AI-related stories that may be slightly less attention-grabbing, but still could move the needle in specific sectors and fields.

🤖 Quora's chatbot platform, Poe, is transforming the 'creator' space by introducing a remuneration program for bot creators. This innovative approach not only injects a fresh perspective into the conventional social media creation scene but also gives bot creators a meaningful financial reward for their creativity and tech prowess.

🧪 AI can diagnose type 2 diabetes in 10 seconds from your voice - by analyzing voice nuances, the AI can diagnose type 2 diabetes with notable accuracy, offering a quicker and more accessible method than traditional blood tests.

📚 GPT-4 is vulnerable to jailbreaks in rare languages. Researchers identified that the AI model GPT-4 could be manipulated using less common languages like Zulu or Scottish Gaelic. This opens doors for breeching in AI's language security systems.

💸  MassRobotics is launching an accelerator - offering a 13-week program to early-stage robotics startups, particularly those in hardware, components, and applications. Selected companies will receive $100,000 in non-dilutive funding provided by MassTech.

✍️  Grant Assistant, a start-up, plans to use generative AI for grant proposal writing, potentially saving organizations hundreds of man-hours and thousands of dollars. The AI-powered tool is not meant to replace writers but to provide invaluable support, making the process less intimidating and more efficient.

🦮 Engineers at Binghamton University have successfully developed a robotic guide dog. The innovation responds to leash tugs and assists visually impaired people, only 2% of whom have access to real seeing-eye dogs, as per their research.

💰 AIs can guess where Reddit users live and how much they earn - in an experiment involving large language models (LLMs) like GPT-4, a person's age, location, gender and income were identified with up to 85% accuracy purely by analyzing Reddit posts. Out of 1500 profiles selected randomly, researchers could confidently identify personal attributes of 520 users based on their posts or profile information.

We love to hear back from you!