Today’s top AI Highlights:

1. Elon Musk announces Grok 2 will be available this fall

2. Resemble AI’s new model detects audio deepfakes with >94% accuracy

3. MultiOn’s best-in-class autonomous web information retrieval API

4. Industry’s first Robots-as-a-Service (RaaS) agreement

& so much more!

Read time: 3 mins

Latest Developments 🌍

Advanced Deepfake Audio Detection is Here 🎭

Fake audio is getting so good, it’s hard to tell what’s real anymore. Resemble AI has introduced DETECT-2B, their latest technology for detecting deepfake audio. It boasts a 94% accuracy rate and delivers results in just 200 milliseconds, a near real-time detection capability. DETECT-2B can identify even the most subtle manipulations in audio to combat the growing threat of incredibly realistic AI-generated fake speech. It supports over 30 languages, making it a powerful tool for a global audience.

Key Highlights:

1. Accuracy - DETECT-2B leverages an ensemble of cutting-edge sub-models, each trained on a massive dataset of real and synthetic audio. This approach allows the system to analyze audio from multiple perspectives, picking up on even the smallest inconsistencies that might indicate manipulation.

2. Catches the Subtle Cues - Beyond simply recognizing known deepfake patterns, it can understand the natural flow and nuances of human speech, making it highly sensitive to the subtle temporal inconsistencies often present in synthetically generated voices.

3. How DETECT-2B Works - The model analyzes input audio in short time slices (frames), predicting a “fakeness score” for each frame. These scores are then aggregated and compared to a threshold to determine a final “real” or “fake” classification for the entire audio clip.

4. Ready to Deploy - Resemble AI offers DETECT-2B through both a versatile API for easy integration into existing systems and a user-friendly web dashboard for more direct interaction.

Extract Data From Any Website with a Single API Call 🌐

MultiOn has released a new tool for developers working with web data: the Retrieve API. This API lets you extract structured information from any website using simple natural language commands, eliminating the need for complex parsing or scraping scripts. It integrates seamlessly with MultiOn’s existing Agent API, so you can build truly autonomous web agents that can navigate pages and scrape information in just 3 lines of code.

Key Highlights:

1. Real-Time Data Extraction - Forget static scrapers and outdated data. Retrieve API crawls and parses web pages on demand, ensuring you always work with the most current information available.

2. Handles Dynamic Content - The API effectively processes JavaScript-rendered content, capturing data from dynamically loaded elements and providing a complete view of the website.

3. Parallel Processing for Efficiency - Leverage the power of multi-agent architecture. Retrieve data from multiple pages or websites simultaneously using parallel requests, significantly speeding up your workflows.

See What Happens When Factories Go Humanoid 🦾

Humanoid robots are no longer a futuristic vision but are actively being integrated into real-world manufacturing and warehouse environments through strategic partnerships.

One prime example is the recent multi-year agreement between GXO Logistics, a leading logistics provider, and Agility Robotics, creators of the humanoid robot Digit. This deal marks the industry’s first Robots-as-a-Service (RaaS) agreement for humanoid robots, with GXO deploying Digit in their SPANX facility to work alongside employees. RaaS model allows companies to access cutting-edge robotics without significant upfront investment.

Another significant partnership is between Figure, a company developing autonomous humanoid robots, and BMW Manufacturing which happened in January this year. This agreement is on deploying Figure’s robots, Figure 01, in BMW’s automotive production facility in South Carolina.

Watch this latest update from Figure where Figure 01 is autonomously picking and placing car parts with precision, even self-correcting when needed. This manipulation is fully driven by advanced neural networks trained on simulations, allowing the robot to perceive its environment, plan its actions, and execute tasks with a high degree of accuracy.

Quick Bites 🤌

1. We love how Elon Musk drops big news this casually and this time it was a reply to a tweet on X. He announced that Grok-2 will be available in August this year, and Grok-3 is already in training and will be out by this year end.

2. Meta is updating its photo labeling system. Meta has been labeling AI-generated content with a “Made with AI” tag. However, some users complained that the tag was applied to images that were real but may have used AI editing tools. To clarify this confusion, the tag is being replaced with “AI info”. (Source)

3. Robinhood Markets has acquired Pluto Capital, an AI-powered investment research platform known for crafting custom investment strategies based on individual financial goals and risk tolerance. We can soon expect powerful AI features woven into the Robinhood platform, making AI accessible to everyday investors for their portfolio decisions.

Tools of the Trade ⚒️

1. Paige: AI tool that manages and optimizes your Google Business Profile automatically. It can write posts, upload images, respond to reviews, manage Q&As, and send review request emails, all tailored to your business.

2. Claude Engineer: A command-line tool that helps with software development using Claude-3.5-Sonnet. It lets you chat with the AI, perform file operations, conduct web searches, and manage projects directly from your terminal.

3. Hamming’s Prompt Optimizer: Uses LLM to generate and refine high-quality prompts for your models, saving you 80% of the manual effort involved in prompt engineering. 

4. Awesome LLM Apps: Build awesome LLM apps using RAG for interacting with data sources like GitHub, Gmail, PDFs, and YouTube videos through simple texts. These apps will let you retrieve information, engage in chat, and extract insights directly from content on these platforms.

Hot Takes 🔥

1. Part of the reason why AI triggers accusations of "hype" is that there is a lot of focus on what AI will do next, and especially what AGI will do.
   Yet the capabilities of LLMs today are more than enough for large-scale impacts, but require organizational & technical integration. ~
   Ethan Mollick
   

2. Both GPT-4 and Sonnet 3.5 can answer all the leetcode interview questions.
   As per today's interview standards, you are better off employing an LLM over a human ~
   Bindu Reddy

Meme of the Day 🤡

the average AI Safety enjoyoor be like:

That’s all for today! See you tomorrow with more such AI-filled content.

Real-time AI Updates 🚨

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

PS: We 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!