• AI Breakfast
  • Posts
  • Tesla's Optimus Robot & CoPilot Release Date

Tesla's Optimus Robot & CoPilot Release Date

Plus, talking to animals with AI

Good morning. It’s Monday, September 25th.

Did you know: Today is the one year anniversary of AI Breakfast? Thanks for following along!

In today’s email:

  • AI & Social Media

  • AI Regulation and Policy

  • AI Enterprise Solutions

  • Robotics

  • Sports and Entertainment

  • Consumer News

  • 5 New AI Tools

  • Latest AI Research Papers

You read. We listen. Let us know what you think of this edition by replying to this email, or DM us on Twitter.

Today’s edition is brought to you by:

Tired of explaining the same thing over and over again to your colleagues?

It’s time to delegate that work to AI. guidde is a GPT-powered tool that helps you explain the most complex tasks in seconds with AI generated documentation.

  • Turn boring documentation into stunning visual guides

  • Save valuable time by creating video documentation 11x faster

  • Share or embed your guide anywhere for your team to see

Simply click capture on our browser extension and the app will automatically generate step-by-step video guides complete with visuals, voiceover and call to actions.

The best part? The extension is 100% free.

Thank you for supporting our sponsors!

Today’s trending AI news stories

AI & Social Media

Meta's AI chatbot plan includes a 'sassy robot' targeting younger users and aiming to compete with platforms like TikTok. These generative AI chatbots, known internally as Gen AI Personas, are designed to increase engagement and may have productivity-related skills. Meta’s shift toward generative AI follows the success of large language models like ChatGPT and their integration into the metaverse. Despite some challenges in adding personalities to chatbots, Meta believes these AI agents can enhance user engagement and potentially improve advertising revenue.

AI Regulation and Policy

Coinbase CEO Calls for Deregulation of Artificial Intelligence development in the United States, arguing that regulation would slow down progress and stifle innovation. Armstrong drew parallels between the AI industry and the crypto sector, emphasizing the importance of decentralization and open-source principles. He expressed concern that regulation could have unintended consequences and hinder competition and innovation. Armstrong highlighted AI’s significance for various applications, including national security, while acknowledging the need to address its potential negative impacts.

White House could force cloud companies to disclose AI customers The law could require companies like Microsoft, Google, and Amazon, to disclose customer information when purchases of computing resources exceed a specific threshold, potentially aiding in the early identification of AI threats. However, the approach faces challenges, as rapidly evolving computing power may render reporting thresholds obsolete, and monitoring could strain the relationships between cloud firms and customers. Advocates argue it could prevent misuse of AI, while critics raise concerns about surveillance implications. The policy would apply primarily to large language models, potentially overlooking other AI technologies with lower computing requirements. British leaders are also exploring AI regulation options.

AI Enterprise Solutions

Microsoft 365 Copilot Release Date Set for November 1st, aimed at business and enterprise users. This AI can extract information from various Microsoft products and interpret text, images, and video, It also includes a prompt-building and sharing application called CoPilot Lab. Microsoft 365 Copilot is priced at $30 per user per month for enterprise accounts. It will unify the AI assistant across Microsoft 365 applications, such as Word, Excel, Powerpoint, Outlook, and Teams, enabling users to streamline their work processes.

Amazon to invest up to $4 billion in AI startup Anthropic, marking its move to compete with tech giants like Microsoft, Meta, Google, and Nvidia in the burgeoning AI sector. Initially, Amazon will invest $1.25 billion for a minority stake in Anthropic, with an option to increase it to $4 billion. Anthropic aims to build a “frontier model” called “Claude-Next,” 10 times more capable than the current Claude model, which you can try here.


Tesla’s Optimus robot can now sort objects autonomously using end-to-end trained neural network. The robot can self-calibrate, precisely locate its limbs, and sort colored blocks into trays. Even when a human interferes by moving blocks, Optimus adapts to changes and continues sorting accurately. Tesla encourages collaboration to further develop Optimus and highlights its AI and robotics initiatives on its website. This progress follows CEO Elon Musk’s announcement that Optimus production was affected by actuator supply shortages, prompting Tesla to design and produce its own actuators.

Sports and Entertainment

The NFL and Amazon are using AI to invent new football stats using Amazon’s Next Gen Stats, utilizing AI and data collection tools to analyze player performance and game patterns. Amazon Web Services (AWS) developed AI algorithms that dig into player behavior, offering granular insights into elements like defender aggression and response times. The partnership focuses on machine learning models for identifying blockers and pass rushers, quantifying “pressure,” and detecting individual matchups, aiding player selection and game strategy. AI tracking helps experts gain valuable player statistics. Insights manifest in interactive graphics, enhancing football analysis. This multidisciplinary project combines the NFL, Zebra Technologies, Wilson Sporting Goods, and AWS.

Consumer News

The Google Pixel 8’s latest leak shows off big AI camera updates such as the “Magic Editor” which allows users to manipulate photos, combining multiple shots and altering elements for desired effects. The AI-driven camera app also offers DSLR-style manual controls, including shutter speed and ISO adjustments. The Pixel 8 Pro boasts a 48MP ultrawide camera and telephoto lens, while both models feature a 50MP wide camera. This leak showcases Google’s commitment to integrating AI for advanced photography experiences on its flagship smartphones.

AI is policing the package theft beat for UPS as 'porch piracy' surge continues across U.S. UPS’s AI system, called DeliveryDefense, assigns a “delivery confidence score” to addresses using historical data and machine learning algorithms. This score helps identify high-risk delivery locations, allowing recipients to choose in-store collection or UPS pick-up. UPS is also launching a web-based version of DeliveryDefense for small- and medium-sized businesses in October. The rise in package theft has promoted various logistics companies, including Amazon, DHL, and FedEx, to adopt technology-driven solutions to protect deliveries.

Artificial Intelligence Could Finally Let Us Talk with Animals and decipher intricate sound patterns from creatures such as crows and sperm whales. These breakthroughs draw from extensive datasets and “self-supervised” AI models, offering practical benefits in conservation, animal welfare, and pet understanding. Still, challenges include maintaining AI precision, tackling multimodal animal communication, and navigating ethical dilemmas tied to potential misuse. AI in commercial industries for precision fishing can detect target species or their predators by listening, while poachers might misuse this tech to locate and mimic calls for endangered animals. For species like humpback whales, synthesizing their songs could lead to unknown societal effects.

5 new AI-powered tools from around the web

CloseVector is a versatile vector database tailored for easy integration and scalability. It operates on users’ machines for optimal performance, serving machine learning applications. Ideal for those seeking efficient article and PDF interactions or recommendation systems with limited candidates.

FireCut, the AI-powered tool, revolutionizes video editing. With features like silence cutting, automated camera switching, zoom cuts, and effortless chapter generation, it transforms hours of footage into polished content with speed and precision.

VOMO empowers users to convert spoken words into organized notes, effortlessly generating slide decks, tables, meeting minutes, and more with GPT-4. Boost productivity, ignite creativity, and bid farewell to excessive typing.

Dataaxy connects job seekers and employers in the data and artificial intelligence field. It offers precision matchmaking, tailored job alerts, and a thriving community. Job seekers can stand out with candidate profiles, while employers find their ideal talent.

Animant empowers users to create immersive AR experiences without 3D design expertise. Its AI enables natural language commands for 3D model creation and animation. Users can even scan physical objects and map their surroundings in 3D. The Storyline feature simplifies animation composition.

arXiv is a free online library where scientists share their research papers before they are published. Here are the top AI papers for today.

DualToken_ViT is an innovative vision transformer model addressing efficiency challenges. It combines convolutional neural networks (CNNs) and self-attention-based vision transformers (ViTs) to efficiently process local and global information. Position-aware global tokens enrich global data and introduce position information, enhancing performance in tasks like image classification, object detection, and semantic segmentation. DualToken-ViT outperforms existing models, achieving accuracies of 75.4% and 79.4% on ImageNet-1K with low computational costs. It introduces a new approach to efficiently harness the power of ViTs for computer vision applications.

The research introduces MetaMath, a specialized language model designed to enhance mathematical problem-solving abilities. MetaMathQA, a novel dataset created through various question bootstrapping techniques, enriches the diversity of mathematical questions. By augmenting questions through rewriting, backward reasoning, and answer augmentation, MetaMathQA enhances LLaMA-2 models, resulting in MetaMath models that outperform existing open-source LLMs. The top model, MetaMath-7B achieves substantial accuracy improvements on two standard mathematical reasoning benchmarks, GSM8K and MATH. Both the MetaMathQA dataset and MetaMath models are released for public use, representing a significant advancement in mathematical reasoning with LLMs.

In this study, researchers propose an adaptive masking approach for efficient pruning of a multilingual automatic speech recognition (ASR) model. Traditional neural network pruning methods involve multiple rounds of pruning and re-training for each language, which can be computationally expensive. The proposed approach, called Dynamic ASR Pathways, dynamically adapts sub-networks during training, avoiding premature decisions about a fixed sub-network structure. The method is shown to outperform existing pruning methods when targeting sparse monolingual models and reduces the need for language-specific pruning in multilingual models. The research demonstrates its effectiveness across several languages and offers a promising avenue for more efficient multilingual ASR models.

CodePlan is a pioneering framework developed by Microsoft Research, India, addressing repository-level coding challenges through LLMs and planning. It focuses on automating complex coding tasks that span entire code repositories, including package migration and temporal code edits. Unlike existing tools that handle localized coding issues, CodePlan formulates repository-level coding as a planning problem. It integrates incremental dependency analysis, change impact assessment, and an adaptive planning algorithm to orchestrate multi-step edits across interconnected code. Evaluations on C# and Python repositories reveal CodePlan’s superiority over baseline methods, marking a significant advance in automating complex coding tasks.

In the study, Boolformer, a transformer-based architecture exhibits remarkable performance across varying conditions. In the noiseless context, it consistently generates precise and concise formulas for previously unseen Boolean functions, extending its capabilities to complex, untrained scenarios. In the noisy regime, the model showcases resilience, effectively managing incomplete datasets, bit flips, and noisy variables. These findings underscore Boolformer’s versatility, underscoring its suitability for real-world applications, such as binary classification tasks and the inference of gene regulatory networks. This model’s adaptability and robustness make it a promising solution for a wide range of practical use cases.

Thank you for reading today’s edition.

Your feedback is valuable.

Respond to this email and tell us how you think we could add more value to this newsletter.