🤖 New Humanoid robot for your home!

Also: China’s Qwen2 model beats GPT-4o

Author

Evolving AI
September 02, 2024

In partnership with

Welcome, AI enthusiasts

1X Technologies has introduced NEO Beta, a prototype humanoid robot designed for home use. Alibaba has released its new AI model, Qwen2-VL, which can analyze videos over 20 minutes long — outperforming GPT-4o. Meanwhile, Magic has announced LTM-2-mini, an AI model with a massive 100-million-token context window, capable of processing the equivalent of 10 million lines of code or 750 novels. Let’s dive in!

In today’s insights:

  • New Humanoid robot for your home!

  • China’s Qwen2 model beats GPT-4o

  • Magic's new AI model has a 100M token context window

Read time: 3 minutes

🗞️ LATEST DEVELOPMENTS

Evolving AI: It may look like a guy in a suit, but 1x Technologies have introcuced their new humanoid robot, NEO Beta.

Key Points:

  • Handles delicate tasks like placing glassware or lifting heavy objects with a human-like touch.

  • Unlike competitors, NEO operates quietly, fitting seamlessly into home environments.

  • Learns from interactions, improving its ability to assist humans in daily tasks over time.

Details:

1X Technologies has introduced its new humanoid robot, NEO Beta, designed for home use. This new robot is designed to help people with household chores. The company is working with full-time operators to increase the robot’s proficiency in various tasks. They are currently preparing for mass production of NEO Beta at their Moss factory in Norway. NEO Beta was developed specifically for use in home environments such as the kitchen. It will be available for trial use in selected homes later this year. CEO Bernt Børnich emphasized that the safety of the robot is a top priority and stated that they placed great emphasis on safety measures during the development process to ensure NEO Beta can operate safely in homes.

Why It Matters:

NEO Beta is seen as an important step towards the use of humanoid robots for household chores and is expected to generate great interest in the consumer market. What do you think? When the time comes, would you love to have a humanoid robot doing your chores at home?

Below are some more examples of NEO Beta doing chores.

The 32,481% Boom: First Disruption to $martphones in 15 Years

Mode saw 32,481% revenue growth from 2019 to 2022, ranking them the #1 overall software company, on this year's Deloitte 500 fastest-growing companies list. Mode is on a mission to disrupt the entire industry with their “EarnPhone”, a budget smartphone that’s helped consumers earn and save $325M for activities like listening to music, playing games, and… even charging their devices?!

Mode has over $60M in revenue - this is your chance to invest in a $1T+ market opportunity!

This is a paid advertisement for Mode Mobile Reg A offering. Please read the offering statement at https://invest.modemobile.com/.

Source: Alibaba Cloud

Evolving AI: Alibaba has launched Qwen2-VL, a new vision-language model that boosts visual understanding, document analysis, and video comprehension.

Key Points:

  • Qwen2-VL can analyze videos over 20 minutes, handle multilingual text, and interpret visual content in real time.

  • The model comes in three sizes, with two open-source versions available for commercial use.

  • Key features include handwriting recognition, function calling, and integration with other third-party tools.

Details:

Alibaba's Qwen2-VL is a next-gen vision-language AI model designed to understand and analyze visual data, including videos over 20 minutes long, in multiple languages like English, Chinese, Arabic, and more. This model excels at real-time video analysis, offering capabilities such as live tech support, summarization, and interactive feedback. It can recognize handwriting in different languages, identify objects in images, and maintain continuous conversations around video content. Available in three sizes, with two open-source versions, Qwen2-VL integrates unique features like Naive Dynamic Resolution and Multimodal Rotary Position Embedding for enhanced processing.

Why It Matters:

The model does have its limitations, the company noted, as it is unable to extract audio from video files, given that it’s designed only for visual reasoning. Its training is also only up to date as of June 2023 and it cannot guarantee complete accuracy for complex instructions or scenarios. However, Alibaba said that the model’s performance and visual capabilities showcased top-tier benchmarks across most metrics, even surpassing closed-sourced models such as OpenAI’s flagship GPT-4o and Anthropic PBC’s Claude 3.5-Sonnet.

Source: Alibaba

Source: Magic

Evolving AI: Magic AI Lab introduces HashHop, a new tool to robustly evaluate AI models' ability to handle ultra-long contexts effectively.

Key Points:

  • Magic’s new AI model, LTM-2-mini, can handle 100 million tokens—equal to 10 million lines of code or 750 novels.

  • To evaluate its model, Magic developed "HashHop," a benchmark that uses random hashes to challenge the AI’s memory and reasoning without giving away any hints.

  • Magic’s model is 1,000 times more computationally efficient than previous models like GPT-3.5, using far less memory and hardware.

Details:

Magic has unveiled LTM-2-mini, their first AI model capable of handling an enormous 100 million token context window. This means it can manage the equivalent of 10 million lines of code, which opens new possibilities for software development. To truly test this model's capability, Magic created the "HashHop" benchmark, which uses random data to assess how well the model can recall and link information without relying on any contextual clues. Their technology also introduces a highly efficient algorithm that vastly reduces computing power and memory needs, unlike traditional models such as GPT-3.5. With the support of Google Cloud and new supercomputers, Magic aims to train even larger models in the future.

Why It Matters:

Yes, a 100 million token context window — that is not a typo. A 100 million token context window enables very large bodies of knowledge, including the entire code base of a good-sized enterprise, to be processed at once. Magic's innovation could transform software development by enabling AI to handle large-scale tasks more efficiently.

🎯SNAPSHOTS

Direct links to relevant AI articles.

🤖 OpenAI says ChatGPT now has 200M users.

🤝 Amazon hires the founders of AI robotics startup Covariant.

📈 NVIDIA results shown AI boom continues despite recent bubble fears.

❤️ AI could help identify patients most at risk of developing conditions that could lead to fatal heart problems.

🗣️ Amazon plans to release a paid version of its Alexa voice assistant with advanced AI features this October.

🚗 Tesla plans Robotaxi reveal at California Warner Bros. Studio.

📈Trending AI Tools

  • 🛠️ Sana - Your ultimate work companion. It's the powerhouse assistant you need for everything from document analysis to task automation (link)

  • 📝 ProDream - An academic writing assistant with outlining, editing, citation formatting, proofreading, and plagiarism checking (link)

  • 📋 Jamie - A tool to automate meeting note-taking by providing summaries, transcripts, and action items from audio recordings (link)

  • 🩺 Docus - A chat assistant for personalized health consultations and second opinions (link)

  • 💬 Empaithy - An app for emotional support and self-reflection through micro-journaling (link)

Reply

or to participate.