Episode #36 - AI Weekly: by Aruna

Welcome back to "AI Weekly" by Aruna - Episode 34 of my AI Newsletter!

I'm Aruna Pattam, your guide through the intricate world of artificial intelligence.

Now, let's delve into the standout developments in AI and Generative AI from the past week, drawing invaluable lessons along the way.

#1 Meta’s Orion AR Glasses: A Glimpse into the Future Beyond Smartphones

Meta has unveiled the?Orion AR glasses, a prototype that could redefine our digital interactions by potentially replacing smartphones. The glasses offer a futuristic experience, overlaying digital elements like 3D content onto the real world. With AI integration, they can recognize objects, allow gesture control, and perform tasks like video calls or suggesting recipes based on what’s in your fridge.

However, Orion is still in its prototype phase, with significant hurdles ahead. The glasses rely on a separate processing device and are currently too bulky for everyday use. Yet, Meta promises to refine and miniaturise the product for eventual mass consumption.

While these AR glasses are not yet ready for the public, they offer a sneak peek into a future where screens are no longer needed, and reality becomes our digital interface.

?? Read more here:?

#2: Law Students Take On ChatGPT: Can AI Hold Up in Legal Exams?

University of Wollongong law students recently faced off against ChatGPT in a criminal law exam, highlighting AI’s potential—and limitations—in legal education. Dr. Armin Alimardani’s study explored whether AI could outperform human students by evaluating standard and prompt-engineered responses from ChatGPT. While most AI-generated papers struggled, scoring in the lower percentile, prompt-engineered responses performed significantly better, scoring as high as 78%.

This raises an important question: Can AI replace human lawyers? Not quite.

The AI models lacked the critical thinking and depth needed to rival human reasoning, especially in handling legal complexities. However, this experiment revealed that AI can assist in legal processes, offering support in research and routine tasks.

Perhaps most surprising was that none of the law tutors detected that any papers were AI-generated, indicating how sophisticated the technology has become. As Dr. Alimardani noted, educators need to teach students how to effectively collaborate with AI tools like ChatGPT, as proficiency in prompt engineering could become a crucial skill in the future legal job market.

While AI can’t replace human judgment, it can enhance productivity in legal practice. Universities must now focus on teaching critical thinking and AI collaboration to prepare future lawyers for this evolving landscape.

Click here for more details!

#3: Telstra’s AI Integration: A Blueprint for the Future of Connectivity

As AI becomes more integrated into everyday life, Telstra is gearing up to play a major role in this transformation. CEO Vicki Brady outlined the company’s ambitious AI roadmap, focusing on building infrastructure to handle rising data demand from AI processes. Telstra’s ongoing investments in 14,000km of intercity fiber and partnerships with Microsoft highlight its commitment to AI.

Incorporating AI into operations, Telstra has already saved millions through preventive actions and energy management. However, as automation rises, job losses loom, forcing a re-evaluation of the workforce.

AI promises growth, but challenges lie ahead.

Click the link to know more:

#4: MIT’s Clio: The Future of Task-Aware Robots

MIT engineers have developed Clio, a groundbreaking AI that enables robots to prioritize task-relevant parts of a scene, much like humans instinctively do. Clio can analyze complex environments and focus only on the elements needed for a specific task, such as retrieving a dog toy or sorting through clutter.

Tested in real-world settings like cluttered offices and apartments, Clio enhances robots’ efficiency, especially for tasks like search and rescue, domestic assistance, and industrial operations.

With AI growing rapidly, Clio may soon redefine robotic interaction in everyday life.

#5: Microsoft Unveils New Copilot Features, Including Voice Interaction and Daily News Summaries

Microsoft is introducing new features for its Copilot AI assistant, including “Copilot Voice” for smoother conversations and “Copilot Daily,” which provides a personalized daily news summary. The company is partnering with major news outlets like Reuters and USA TODAY for curated content. Copilot Labs, part of a $20/month Pro subscription, will allow users to test experimental features like real-time screen interpretation and deeper, more complex responses. Microsoft also announced new Copilot+ PCs, with additional AI-driven capabilities, set to roll out later this year.

Click here?to read the full story!

#6: SoftBank to Invest $500 Million in OpenAI

SoftBank’s Vision Fund is set to invest $500 million in OpenAI during its latest funding round, according to reports. This marks SoftBank’s first major investment in the AI company, which is currently raising $6.5 billion in convertible notes, valuing OpenAI at $150 billion. The valuation is contingent on potential changes to OpenAI’s corporate structure, including removing a profit cap for investors. Apple was reportedly interested but dropped out of the funding round.

Read the link for more details.

?#7: Researchers Introduce Self-Correction Mechanism for Safer AI Models

MIT and Peking University researchers have unveiled a groundbreaking self-correction mechanism for large language models (LLMs) to enhance their safety and autonomy. Traditionally reliant on external feedback, LLMs can now self-critique and refine their responses using in-context alignment (ICA). This innovation significantly improves error reduction, alignment, and resilience against sophisticated jailbreak attacks. The self-correction framework also addresses social biases, enhancing LLM performance in real-world applications while minimising human intervention.

This approach could revolutionise future AI model development across industries.

#8: Liquid AI Unveils State-of-the-Art Non-Transformer Models

MIT spinoff Liquid AI has introduced Liquid Foundation Models (LFMs), a series of non-transformer AI models that outperform popular transformer-based models like Meta’s Llama and Microsoft’s Phi. Leveraging principles from dynamical systems and signal processing, LFMs deliver high performance across tasks such as text, audio, and video processing while using significantly less memory.

These models offer exceptional efficiency and flexibility, positioning Liquid AI as a leader in foundation models. A full launch is scheduled for October 23, 2024.

#9: OpenAI’s Transformation: From Research Lab to Tech Giant

OpenAI, initially launched as a nonprofit with an altruistic mission, is transitioning toward a more conventional tech company structure. Amid reports of restructuring and a potential public benefit corporation shift, senior leaders, including CTO Mira Murati, have left, raising questions about the company’s future direction. OpenAI’s increased focus on rapid product launches, like the GPT-4o model, signals a move toward profit-driven goals, placing safety and research on a backburner. CEO Sam Altman remains at the helm, steering the company’s evolving identity.

#10: IBM and NASA Collaborate to Tackle Climate Crisis with AI

IBM and NASA have introduced a cutting-edge AI foundation model to address climate and weather challenges. Developed in collaboration with Oak Ridge National Laboratory, the model can forecast severe weather and long-term climate trends, providing localized predictions and enhancing global climate simulations. Trained on 40 years of Earth data, the model is available on Hugging Face, allowing researchers to fine-tune its applications. This innovation marks a significant step forward in using AI to mitigate climate risks.

That wraps up our newsletter for this week.

Feel free to reach out anytime.

Have a great day, and I look forward to our next one in a week!

Thanks for your support.