Google DeepMind Finds Way to Put AI into Robots

Plus: Researchers find a way to turn AI chatbots evil, Together AI's LLaMA-2-7B-32K.

Hello Engineering Leaders and AI Enthusiasts!

Welcome to the 74th edition of The AI Edge newsletter. This edition brings you Google DeepMind’s novel AI model that can control robots.

And a huge shoutout to our incredible readers. Your support is invaluable to us!😊

In today’s edition:

💡 Google DeepMind finds way to put AI into robots
😈 ChatGPT to Bard– Researchers find a way to turn AI chatbots evil
🚀 Together AI extends Llama-2 to 32k context
🧠 Knowledge Nugget: Exploring Large Language Models: A Dive Into Your Top Questions by

ScaleDown

Let’s go!

---

Google DeepMind finds way to put AI into robots

Google DeepMind has introduced Robotic Transformer 2 (RT-2), a first-of-its-kind vision-language-action (VLA) model that learns from both web and robotics data. It then translates this knowledge into generalized instructions for robotic control. This helps robots more easily understand and perform actions– in both familiar and new situations

The approach results in very performant robotic policies and, more importantly, leads to a significantly better generalization performance and emergent capabilities due to web-scale vision-language pretraining. Thus, internet-scale text, image, and video data can now be used to help robots develop better common sense.

Why does this matter?

With this AI model, robots can learn more like we do– transfer learned concepts to new situations. It shows the promise of building general-purpose physical robots that can reason, problem-solve, and interpret information for performing a diverse range of tasks in the real world.

Source

---

ChatGPT to Bard– Researchers find a way to turn AI chatbots evil

LLMs today undergo extensive fine-tuning to ensure they do not produce harmful content in their responses. However, new research has introduced an approach that automatically produces adversarial suffixes to prompt the models, which results in affirmative responses for objectionable queries.

Unlike traditional jailbreaks, these are built in an entirely automated fashion, allowing one to create virtually unlimited number of such attacks. Although built to target open-source LLMs,  the strings easily transfer to many closed-source, publicly-available chatbots too, like ChatGPT, Bard, and Claude.

Why does this matter?

It questions the process of adversarial training of AI models, in which during training or finetuning of a model, it is attacked with such methods and then iteratively trained to respond correctly. Adversarial training is still proven to be effective, but does it eventually lead to models susceptible to such attacks?

Source

---

Together AI extends Llama-2 to 32k context

Together AI has released LLaMA-2-7B-32K, a 32K context model built using Meta’s Position Interpolation and Together AI’s data recipe and system optimizations, including FlashAttention-2. You can fine-tune the model for targeted, long-context tasks– such as multi-document understanding, summarization, and QA. Here’s the model in Playground completing a book:

Upon evaluation, the model achieves comparable quality than the original LLaMA-2-7B base model.

Why does this matter?

Open-source LLMs have been catching up with closed-source models. But this move by Together opens up opportunities to extend the context length of open models to the regime of 32K-128K (or even surpass it) and match the SoTA closed-source models.

Source

---

Knowledge Nugget: Exploring Large Language Models: A Dive Into Your Top Questions

•  Overfitting in LLMs: What is it, and how does LLM training deal with it?

• Tailoring LLMs: Can we modify pre-trained models according to our needs?

• Deploying LLMs: Given their size, how do you navigate the complexities of deploying LLMs?

And so on…Many intriguing questions surround the realm of LLMs. This article by

ScaleDown

addresses numerous queries, peeling back the layers of LLMs. It serves to enlighten both curious beginners and seasoned AI enthusiasts alike.

Why does this matter?

Covering a broad spectrum of LLMs' nuances, the blog offers a deeper understanding of powerful language models, their potential applications, and practical insights into implementing AI technology.

Source

---

What Else Is Happening❗

🌍ChatGPT for Android now available in 12 more countries! Check the link for details (Link)

🗣️Forget subtitles. YouTube now dubs videos with AI-generated voices with tool Aloud (Link)

💼Paris-based Capgemini will invest 2Bn euro in AI and double AI team over 3 years (Link)

💡Intel CEO: The company plans to build AI into every product it builds (Link)

🔬AI identifies protein snippets from Neanderthals that have anti-biotic powers (Link)

🎓GPT-4 passes first Harvard semester in humanities and social sciences experiment (Link)

---

🛠️ Trending Tools

• Likewise Learn: AI app to test, refine, and generate engaging Twitter/Mastodon content. Get detailed performance predictions.

• Signnme: Create stunning email signatures reflecting your brand identity. No design skills are required. 10+ templates are available.

• CodeSense AI: Cutting-edge VS Code extension to enhance code documentation and comprehension through AI.

• Interacly AI: Transforms digital interactions with simplicity and efficiency. Streamlined experiences and customized chatbots for businesses.

• Sttabot: Official WordPress plugin. Add your Sttabot AI bots to your WordPress website in a single click.

• Doc Dog: Document AI platform for extracting data from invoices and receipts with high accuracy. Use its API and webhook.

• Unthread: Helps companies scale up support on Slack. Tracks, tags, prioritizes conversations, and generates responses with AI.

• Vocol AI: One-stop voice collaboration platform powered by advanced speech and NLP technologies. Turns voice and data into insights.

---

🧐Monday Musings: Bringing the Llama-2 experience on your local devices!

We have been talking much about the free, open-source Llama-2. But ever wondered if you could have your own Llama-2 adventure right on your M1/M2 Mac, Windows, Linux, or even your phone? You totally can, and the coolest part is you don’t even need an internet connection.

Here’s an example using a locally-running Llama 2 to whip up a website about why llamas are cool:

Check out this blog to uncover open-source tools you can use to run Llama 2 on your own devices.

(Source)

---

That's all for now!

If you are new to ‘The AI Edge’ newsletter. Subscribe to receive the ‘Ultimate AI tools and ChatGPT Prompt guide’ specifically designed for Engineering Leaders and AI Enthusiasts.

Thanks for reading, and see you tomorrow. 😊

---