Good morning! This is your daily ☕️ Techpresso.
On this day in 1860, Herman Hollerith - the inventor of the tabulating machine - was born.
In today's Techpresso:
🤖 OpenAI to put AI into humanoid robots
💥 GitHub besieged by millions of malicious repositories in ongoing attack
🎵 Adobe's new AI tool can generate audio using text prompts
😳 Nvidia just released a new code generator that can run on most modern CPUs
🦙 Metas Llama 3 set to release in July and could be twice the size
⚖️ Three more publishers sue OpenAI
🎁 + 9 other news you might like
🔮 + 3 handpicked research papers and tools
🤖 OpenAI to put AI into humanoid robotsLINK
OpenAI is collaborating with robotics startup Figure to integrate its AI technology into humanoid robots, marking the AI's debut in the physical world.
The partnership aims to develop humanoid robots for commercial use, with significant funding from high-profile investors including Jeff Bezos, Microsoft, Nvidia, and Amazon.
The initiative will leverage OpenAI's advanced AI models, such as GPT and DALL-E, to enhance the capabilities of Figure's robots, aiming to address human labor shortages.
💥 GitHub besieged by millions of malicious repositories in ongoing attackLINK
Hackers have automated the creation of malicious GitHub repositories by cloning popular repositories, infecting them with malware, and forking them thousands of times, resulting in hundreds of thousands of malicious repositories designed to steal information.
The malware, hidden behind seven layers of obfuscation, includes a modified version of BlackCap-Grabber, which steals authentication cookies and login credentials from various apps.
While GitHub uses artificial intelligence to block most cloned malicious packages, 1% evade detection, leading to thousands of malicious repositories remaining on the platform.
🎵 Adobe's new AI tool can generate audio using text promptsLINK
Adobe is developing an experimental generative AI, Project Music GenAI, that creates music based on text prompts and offers editing tools for customization.
The AI tool allows users to manipulate existing music files and generate new compositions with varying structures and tempos.
Competing with similar technologies from Meta and Google, Adobe's solution aims to be more user-friendly with intuitive editing capabilities.
😳 Nvidia just released a new code generator that can run on most modern CPUsLINK
Nvidia, ServiceNow, and Hugging Face have released StarCoder2, a series of open-access large language models for code generation, emphasizing efficiency, transparency, and cost-effectiveness. StarCoder2, trained on 619 programming languages, comes in three sizes: 3 billion, 7 billion, and 15 billion parameters, with the smallest model matching the performance of its predecessor's largest. The platform highlights advancements in AI ethics and efficiency, utilizing a new code dataset for enhanced understanding of diverse programming languages and ensuring adherence to ethical AI practices by allowing developers to opt out of data usage.
🦙 Metas Llama 3 set to release in July and could be twice the sizeLINK
Meta's next open-source language model, Llama 3, is set for a July release and aims to match GPT-4 in performance, possibly doubling in size compared to its predecessor, Llama 2.
Llama 3 is designed to provide clearer contexts for complex topics and will be better at understanding nuances in language, particularly for words that could have multiple meanings based on context.
Despite aiming for GPT-4 quality, Llama 3's development faces challenges, including the recent departure of key personnel in AI safety at Meta, and it remains undecided whether it will incorporate multimodal capabilities.
⚖️ Three more publishers sue OpenAILINK
The Intercept, Raw Story, and AlterNet have filed lawsuits against OpenAI and Microsoft in the Southern District of New York, alleging copyright infringement through the training of AI models without proper attribution.
The litigation claims that ChatGPT reproduces journalism works verbatim or nearly verbatim without providing necessary copyright information, suggesting that if trained properly, it could have included these details in its outputs.
The suits argue that OpenAI and Microsoft knowingly risked copyright infringement for profit, evidenced by their provision of legal cover to customers and the existence of an opt-out system for web content crawling.
Other news you might like
Tim Cook says Apple will "break new ground" in generative AI this year.LINK
Elon Musk provides update on Tesla Roadster timeline.LINK
Coinbase users’ wallets shown wiped to $0 after crash from ‘large surge in traffic’ — as Bitcoin soared.LINK
AI takes priority at TikTok owner ByteDance, as Sora disrupts the future of video creation.LINK
Crypto firm Gemini to pay over $1 billion back to customers in settlement.LINK
Meta’s ‘consent or pay’ data grab in Europe faces new complaints.LINK
Snowflake says Frank Slootman is retiring as CEO, stock plunges more than 20%.LINK
Nvidia GeForce RTX 5090 could be up to 70% faster than the 4090, but its best chips might be reserved for AI.LINK
Google hit with $2.3B lawsuit by 32 media groups over digital ad practices.LINK
Latest research and tools
Talc: a fast and flexible allocator designed for no_std and WebAssembly environments, improving allocation efficiency in WebAssembly apps and multi-core systems.LINK
Fediverse: a federated universe that functions similarly to email, with a few major hubs dominating while some entities host their own servers, making complete decentralization difficult to achieve.LINK
Transparent Image Layer Diffusion Using Latent Transparency: the study introduces a method to create clear image layers by leveraging hidden transparency features.LINK
Want to get the latest news differently? Find us on:
See you tomorrow for a new dose of ☕️ Techpresso!
-min.png){kind=small}
.png){kind=small}
