AI, QUANTUM COMPUTING & ROBOTICS WEEKLY REVIEW

Navigating the Next Wave: Quantum Computing, AI, and Robotics

OpenAI's Q*, also known as 'Q Star', is creating waves in the Artificial General Intelligence (AGI) sphere with its human-like cognitive abilities. This model, integrating aspects of Q-learning and the A* search algorithm, signifies a significant shift towards more adaptable AI systems.

Recent Controversy and CEO's Return

Sam Altman, after a brief removal as CEO of OpenAI, has returned to his position. Amidst this turmoil, reports have surfaced about Q*, hinting at its potential implications for humanity and its role in Altman's temporary departure. In his recent comments, Altman acknowledged the leak regarding Q* but primarily focused on AI safety and the broader goals of OpenAI.

Q*'s Mathematical Prowess

Q* is reportedly capable of solving mathematical problems, a feat challenging for generative AI models. Though it currently performs at a grade-school level, its success in this area has sparked optimism about its future capabilities.

Enhancing Reasoning in AI: The Q Strategy

Q* aims to elevate reasoning within LLMs, a move that could revolutionize tasks requiring logical progression and problem-solving. This development brings to light critical discussions about AI autonomy and ethics.

Introducing SDXL Turbo: A Real-Time Text-to-Image Generation Model

SDXL Turbo Performance

SDXL Turbo, a new real-time text-to-image generation model, achieves state-of-the-art performance with its innovative distillation technology, Adversarial Diffusion Distillation (ADD). This technology allows single-step image generation with high fidelity, representing a leap forward in image synthesis efficiency and quality.

Performance and Speed

In comparative tests, SDXL Turbo outperformed other multi-step models like LCM-XL and SDXL in terms of prompt adherence and image quality, all while requiring substantially less computational power. It can generate a 512x512 image in just 207ms on an A100, demonstrating remarkable speed improvements.

Global AI Security Agreement

Collaborative International Effort

Eighteen nations, including the U.S., Britain, Germany, and others, have officially endorsed the new "Guidelines for Secure AI System Development." These guidelines, developed collaboratively by major cyber security agencies and contributions from tech giants like Google and OpenAI, aim to create a unified approach to mitigating AI-related risks.

Key Points of the Guidelines

1. Secure-by-design and secure-by-default: Emphasizes proactive protection against cyber intrusions for AI-based products, pushing developers to prioritize security in model architecture and training datasets.

2. Supply Chain Security: Advises careful consideration of code component sources and the implementation of security measures for externally acquired components.

3. Unique AI Code Risks: Highlights specific cyber threats like prompt injection attacks and data poisoning, necessitating unique security considerations.

4. Collaborative and Continuous Approach: Encourages sharing information within the AI community to evolve and advance systems, covering all stages of the code lifecycle.

Global Participation and Notable Absence

The agreement has been signed by 18 countries, but notably, China, a leading AI developer ***************, has not yet endorsed the document.

QUANTUM COMPUTING SECTION

1. AWS's New Quantum Chip Amazon Web Services (AWS) recently unveiled a new quantum computer chip that significantly enhances error correction capabilities. This chip, created entirely in-house, can suppress bit flip errors by 100x using a passive error correction approach. By integrating both passive and active error correction techniques, this chip could theoretically achieve quantum error correction six times more efficiently than standard methods. The use of superconducting quantum circuit technology allows for scalability in the number of physical qubits, which is crucial for suppressing the total logical error rate. This advancement is an essential step toward practical and reliable quantum computing, potentially benefiting industries like pharmaceuticals, materials science, and financial services.

2. Multiverse Computing's Quantum Digital Twin Project Multiverse Computing has pioneered a project using a digital twin and quantum optimization to increase the efficiency of green hydrogen production. This breakthrough could significantly improve the economics of hydrogen production and reduce greenhouse gas emissions. In collaboration with IDEA Ingeniería and AMETIC, the digital twin numerically simulates a green hydrogen production plant and employs quantum algorithms to optimize the electrolysis process. The quantum solution led to a 5% increase in hydrogen production and associated revenue compared to classical solvers, demonstrating the potential for quantum algorithms to make renewable energy more cost-effective and sustainable.

3. Fundamental Limits of Quantum Computers Research from the Vienna University of Technology has revealed a fundamental limitation for quantum computers: no clock can achieve both perfect resolution and precision simultaneously due to finite energy or entropy generation. This discovery implies inherent limits for quantum computers, as the achievable resolution and precision in timekeeping restrict the speed and reliability of quantum computations. Quantum state changes in a quantum computer correspond to rotations in higher dimensions, which require precise timekeeping for accuracy. The study highlights a trade-off between time resolution and precision, suggesting that either rapid functioning or high precision is achievable, but not both simultaneously. This finding sets a natural boundary for the future development of quantum computers.

ROBOTICS

1. Wombo's New AI Avatar App - Wombo Me Wombo, known for its AI-generated art app Dream by Wombo, has launched a new AI-powered app called Wombo Me. This app allows users to turn a single selfie into multiple lifelike avatars. It stands out in the crowded AI avatar app market by requiring only one selfie and generating images almost instantly. Wombo Me is designed for entertainment, allowing users to experiment with different personas, hairstyles, and makeup trends. The app also includes a Tinder-like swiping feature for selecting preferred avatars and plans to use user feedback to enhance future image generation.

2. Amazon's Titan Image Generator Amazon introduced the Titan Image Generator at its AWS re:Invent 2023 conference. Available on Bedrock, Amazon’s AI development platform, this generative AI model can create new images from text descriptions or modify existing images, such as changing backgrounds. Titan Image Generator was trained on diverse datasets across various domains and includes built-in mitigations for toxicity and bias. Its capability for customization and image generation positions it as a significant player in AI-powered media creation.

3. Kognitos' Funding for Automating Business Processes Kognitos, an AI startup, raised $20 million to help businesses automate back-office processes. Leveraging AI, Kognitos aims to increase efficiency by automating mundane tasks. Its platform allows users to automate tasks in plain English, making it accessible to users with mid-level technical abilities. Kognitos is competing with established RPA platforms and differentiates itself with its more intuitive and sophisticated automation technology. This funding will be used for product development and expanding partnerships.

4. Robotics Advancements Recent developments in robotics have shown significant progress in various applications:

• Acoustic Wave Sensor Robots: A network of mobile robots has been developed to inspect large pipe structures using acoustic wave sensors. This technology has demonstrated the capability to detect multiple defects in steel pipes.
• Teaching Robots to Recognize Uncertainty: Engineers have devised a method to make robots smarter by programming them to recognize when they lack information. This technique quantifies the fuzziness of human language to instruct robots when to seek further directions.
• Crowdsourced Feedback for Robot Training: A new method uses crowdsourced feedback from nonexpert human users to guide AI agents through reinforcement learning. This technique allows for more effective and diverse training of robots.

The latest advancements in AI robotics and quantum computing demonstrate the ongoing integration of AI in consumer applications and the increasing complexity in robotics, with the potential to revolutionize technology and everyday life in various ways.

UPDATE: HOW THE WORLD IS USING AI AND THE DEVELOPING DISPARITIES

Here is an extremely compelling and technical article about ChatGPT, Character AI, and Google Bard being the top gainers in AI traffic:

After an explosive start to 2023, the AI hype train kept charging ahead, reaching peak monthly visits of around 4.1 billion in May 2023 according to WriterBuddy's Sujan Sarkar[^1](https://zapier.com/blog/chatgpt-vs-bard/). This was followed by a pullback of 1.2 billion visits across the industry.

The top 50 AI tools combined garnered over 24 billion total visits with a 10.7x growth rate and 236.3 million average monthly visit increase. Overall, average monthly visits reached 2 billion, surging to 3.3 billion in the last six months[^2](https://zapier.com/blog/chatgpt-vs-bard/).

Although Quillbot and Midjourney were top 5 most-visited, they both saw traffic drops in contrast to explosive growth in other AI tools. Midjourney lost 8.66 million visits and Quillbot lost 5 million in the past year[^3](https://zapier.com/blog/chatgpt-vs-bard/).

The average AI session lasted 12 minutes 34 seconds, enough to query ChatGPT and get a cocktail recipe. However, ChatGPT sessions averaged less than 10 minutes while tools like Character AI, Dezgo and Janitor AI averaged 25 minutes[^4](https://zapier.com/blog/chatgpt-vs-bard/).

AI chatbots and writing generators like ChatGPT, Character AI and Google Bard topped categories. Image generators were second highest, including Midjourney, DeepAI and Craiyon[^5](https://zapier.com/blog/chatgpt-vs-bard/).

The US led AI use with 5.5 billion total visitors, about 22.6% of worldwide traffic. India was second with 2.1 billion visits, 8.5% of global use[^6](https://zapier.com/blog/chatgpt-vs-bard/).

69.5% of users were male versus 30.5% female. The study didn't analyze this gap but different gender roles likely play a role[^7](https://zapier.com/blog/chatgpt-vs-bard/).

80.5% of the top 50 AI tools were direct traffic indicating a loyal user base. Only 0.12% were paid search[^8](https://zapier.com/blog/chatgpt-vs-bard/).

Stay Ahead. Stay ZEN

SCIENTISTS, DEVELOPERS, ANALYSTS & MORE

JOIN US ??

https://www.dhirubhai.net/groups/12925135/