World's Most Responsible, No Controversy LLM🖐️

PLUS: Sam Altman seeks $ 7 TRILLION, Autonomous Drone learns to fly in 18 seconds

Today’s top AI Highlights:

1. The world’s safest AI model that won’t answer almost anything straight

2. Enhancing AI voice assistants’ conversational quality, naturalness and empathy

3. Tiny drone learns to fly in 18 seconds

4. Sam Altman seeks a staggering $5 to $7 trillion to jumpstart global AI chip production

5. DSPy by Stanford: The framework for programming - not prompting - foundation models

& so much more!

Read time: 3 mins

Latest Developments 🌍

The “Outrageously Safe” AI Model 🧯

In light of the growing misuse of AI, ensuring that the output of AI models is extremely safe is crucial. But how far do you want to go down the safety path? Meet Goody-2, proclaimed as the world’s most responsible AI model that distinguishes itself by its unconventional method of addressing user requests. The model perceives literally all kinds of requests as potentially unethical and offensive and conveniently avoids answering.

Developed by Brain, “a very serious” studio, GOODY-2 is a satire on the exaggerated level of caution. It has an extreme approach to avoiding controversy and bias, to the extent that it might refuse to answer even the most benign questions like “What 2+2 is?” if there's a remote possibility of them being misconstrued. This over-the-top safety feature humorously highlights the challenges of striking the perfect balance between having a safe AI interaction and a frustrating experience.

---

AI Learning Hour ⏰

Build GPTs that leads the OpenAI Leaderboard 📈

Learn the strategies for integrating custom actions in GPTs and get tips for optimizing GPT performance and efficiency in various development scenarios. Reserve your spot now for a FREE live hands-on session 👉 Register Now!

---

Conversations with AI More Human-like 🗣️

Despite the widespread adoption of AI voice assistants in daily life, users frequently encounter frustrations due to the assistants’ inability to engage in truly natural and empathetic conversations. This disconnect stems from the assistants' challenges in grasping the full spectrum of human emotions, context, and the subtle nuances that characterize our interactions. Recognizing these limitations, a collaborative effort led by LAION has open sourced BUD-E (Buddy for Understanding and Digital Empathy) to bridge the gap between the mechanical responses of current AI voice assistants and the fluidity of human conversation.

Key Highlights:

1. BUD-E is designed to respond to user requests in real time, utilizing natural voices that convey empathy and emotional intelligence. This capability is complemented by the system's understanding of long-term conversational context and its ability to handle complex interactions involving multiple speakers, interruptions, and thinking pauses, all while operating fully locally on consumer hardware.

2. BUD-E has achieved response latencies of 300 to 500 milliseconds with the Phi 2 model, indicating a significant step towards real-time interaction. The team is optimistic about further reducing these times to below 300 milliseconds using larger models like LLama 2 30B, on consumer-grade hardware such as the 4090 GPU. This level of performance is key for seamless, natural dialogues, minimizing the awkward pauses that disrupt the flow of interaction.

3. BUD-E is designed with a holistic approach to improving conversational AI, addressing not just the technical aspects of latency and voice recognition but also the nuanced challenges of emotional intelligence, context awareness, and multi-speaker conversation handling. Features such as long-term context memory, empathy and emotional intelligence, and the ability to handle complex conversational dynamics with interruptions and affirmations, are at the core of BUD-E’s development.

18 Seconds to Takeoff 🪽

Researchers at New York University's Agile Robotics and Perception Lab, together with the TII, have achieved a remarkable feat in drone autonomy by developing a system that trains drones to fly in just 18 seconds! Utilizing a consumer MacBook Pro, this system not only makes drone training remarkably efficient but also showcases the potential for rapid deployment in research and commercial applications. What stands out is the system's ability to train drones to maintain stable flight and navigate specific trajectories quickly, leveraging advanced simulation and reinforcement learning techniques on readily accessible hardware.

Key Highlights:

1. The collaboration has led to a drone training process that is significantly faster than existing methods, achieving stable and controllable flight in a mere 18 seconds of simulation on a MacBook Pro. This process is even more time-efficient than the time it takes to compile and flash the firmware onto the drone itself, indicating a substantial leap in the speed of drone readiness from simulation to real-world application.

2. The system employs a strategic curriculum approach in its reinforcement learning model, starting with a forgiving reward function and gradually increasing penalties. This method ensures that drones are not only trained quickly but are also robust and reliable in various conditions, avoiding the common pitfall of overfitting to a simulated environment.

3. This training approach is designed to be universally applicable across different types of drones, from small, open-source models to larger, more complex constructions. Furthermore, the project is open-source and plans for future integration with the PX4 open-source drone autopilot. Researchers are focused on refining the system to adapt to different environmental conditions and vehicle configurations for diverse flying robots beyond just quadrotors.

Altman's Trillion-Dollar Bid for Global AI Chip Production 💰

Sam Altman, CEO of OpenAI, is reportedly pursuing an investment between $5 to $7 trillion to expand global AI chip production significantly to boost the manufacture of high-powered computer chips crucial for catering to the escalating demand for advanced AI technologies. The plan addresses semiconductor production and its substantial energy requirements. Altman has discussed funding with potential investors, including officials from the United Arab Emirates, to secure the necessary funding. The projected financial requirement surpasses the combined market values of Apple and Microsoft.

Altman has also engaged with the US government, including Commerce Secretary Gina Raimondo, seeking support for this venture, which aligns with efforts to enhance domestic chip production and lessen dependence on foreign sources like China. Microsoft supports Altman's plans, which include collaboration between OpenAI, chip makers, and energy companies to build factories for increased chip output.

---

Tools of the Trade ⚒️

1. CodeAI: A platform where you can test out any code in HTML, CSS & JS and see its preview on the platform itself. It has an AI-integrated chatbot to which you can ask any query. It has a live code context of your code.

2. DSPy: A framework designed to optimize language model prompts and weights, particularly in complex pipelines involving multiple language model uses. It separates program flow from parameters and introduces optimizers for tuning, aiming for less manual prompting and enhancing task reliability.

3. Onedoc: Simplifies the process for developers to create beautiful PDFs by offering an open-source library for document design, a simple API for document generation, and a cloud platform for hosting and tracking documents​.

4. Jan: An open-source, self-hosted alternative to ChatGPT that runs 100% offline on your computer. It brings AI to your desktop with features like customizable AI assistants, global hotkeys, and in-line AI, ensuring privacy by keeping conversations, preferences, and model usage secure and local on your device​.

😍 Enjoying so far, TWEET NOW to share with your friends!

---

Hot Takes 🔥

1. Prediction: Sam Altman will get his 5 $7 trillion, in big part by the United States government and allies, in order to ensure the West achieves AGI before other regions.
   Whoever achieves AGI first rules the world. It is likely the last invention humans ever create. Have a great Friday. ~ Jeremiah Owyang

2. openai now generates about 100 billion words per day.
   all people on earth generate about 100 trillion words per day. ~ Sam Altman

---

Meme of the Day 🤡

Ouch..

Source

That’s all for today!

See you tomorrow with more such AI-filled content. Don’t forget to subscribe and give your feedback below 👇

Real-time AI Updates 🚨

⚡️ Follow me on Twitter @Saboo_Shubham for lightning-fast AI updates and never miss what’s trending!!

PS: I curate this AI newsletter every day for FREE, your support is what keeps me going. If you find value in what you read, share it with your friends by clicking the share button below!