AI Roundup : AI worms, Elon Musk Sues OpenAI, Mistrals new model, Llama 3,Apples Car, Klarna's AI-Powered Support Bot,Figure AI humanoid and more....
Dhamodharan Sankaran
Technology Leader | Conversational AI | Digital Transformation | Customer Centricity | Agile | Results-Oriented | Engineer | Architect | Shaping High Performing teams | Collaborator
AI Worms
- AI worms are a new type of cyber threat that can spread through generative AI-powered emails, compromising security measures of AI systems.
- Researchers have created an AI worm named Morris II, which overwhelms AI-powered email clients by continually forwarding messages.
- These AI worms can automatically spread, steal data, and deploy malware by using adversarial self-replicating prompts.
- Generative AI worms have not been observed in the wild yet, but they pose a significant security risk that developers and tech companies should be aware of.
- Research findings have been shared with major companies like Google and OpenAI to enhance system resilience against these attacks.
Mistral AI releases new model
- Mistral AI has launched a new flagship large language model called Mistral Large to rival models like GPT-4.
- Mistral Large is multilingual, supporting English, French, Spanish, German, and Italian, with top-tier reasoning capabilities.
- Mistral AI has introduced a beta chat assistant named Le Chat that users can try for free.
- Mistral AI has partnered with Microsoft to offer its models on the Azure platform, becoming the second company after OpenAI to do so.
- The pricing for Mistral Large is $8 per million input tokens and $24 per million output tokens, making it 20% cheaper than GPT-4 Turbo.
Elon Musk Sues OpenAI and Altman for Breaching Firm’s Founding Mission
- Elon Musk has filed a lawsuit against OpenAI and its CEO, Sam Altman, alleging a breach of the firm's founding mission of developing AI for the benefit of humanity.
- Musk claims that OpenAI has shifted its focus towards profit, particularly through a partnership with Microsoft, deviating from its original nonprofit roots.
- The lawsuit seeks to compel OpenAI to adhere to its founding agreement and return to its original mission of benefiting humanity rather than prioritizing profits for Microsoft.
Apple Shifts Focus from Autonomous?Cars to Generative AI
- Apple shifted focus from its autonomous car project to prioritize generative AI initiatives, redirecting around 2,000 employees previously working on the self-driving vehicle to work on AI projects.
- The decision, led by COO Jeff Williams and Kevin Lynch, aims to boost Apple's generative AI efforts under the guidance of executive John Giannandrea, aligning with peers like Microsoft and Google who have made strides in this area.
- This strategic move reflects Apple's adaptation to the evolving tech landscape and increasing prominence of artificial intelligence technologies across diverse industries.
Meta plans launch of new AI language model Llama 3 in July
- Meta Platforms is set to launch its new AI language model, Llama 3, in July, aiming to provide better responses to contentious questions with enhanced nuance and context awareness.
- Meta plans to appoint an internal overseer for tone and safety training to improve the accuracy and appropriateness of Llama 3's responses, following Google's criticism over historical inaccuracies in its Gemini chatbot.
领英推荐
Klarna's AI-Powered Support Bot
Klarna has introduced an AI assistant that handles two-thirds of customer service chats, effectively doing the work of 700 human agents. The bot is proficient in dealing with a variety of customer queries
Figure AI and OpenAI Collaboration
- Figure secured $675 million at a $2.6 billion valuation from notable investors like Microsoft, NVIDIA, Jeff Bezos, and others.
- Figure and OpenAI entered a collaboration agreement to work on developing advanced AI models for humanoid robots, intending to expedite Figure's commercial progress by improving the robots' language processing and reasoning capabilities.
The Groq LPU? Inference Engine
- Groq's LPU Inference Engine is known for being fast and efficient in handling complex tasks, especially for large language models.
- It boasts features like exceptional sequential performance, single-core architecture, and high accuracy even at lower precision levels, outperforming GPUs in addressing compute density and memory bandwidth challenges in LLMs.
- Groq's LPU system can process over 300 tokens per second per user in models like Llama 2 70B, making it a standout choice for AI developers and positioning Groq as a key player in the AI chip industry.
EMO: Emote Portrait Alive
Alibaba recently announced EMO, a model capable of generating realistic talking and singing videos from any image or audio.?
Google Genie's 2D Game AI
- Google DeepMind introduced Genie, an AI model capable of generating complete 2D platformer games from a single image or text prompt, leveraging its training on a vast dataset of 2D platformers.
- With a deep understanding of game mechanics like player control and movement, Genie, trained on over 200,000 hours of data, excels in creating playable virtual environments, offering potential applications in areas like robotic navigation training.
- The AI model's proficiency in game development signifies a notable innovation in generative AI technology, presenting opportunities for diverse interactive experiences and AI agent training in virtual settings.
Microsoft's Introduces 1-bit LLMs
- Microsoft introduced BitNet b1.58, a Large Language Model (LLM) using 1.58 bits per parameter, significantly reducing computational demands while maintaining performance.
- By utilizing ternary values (-1, 0, 1) instead of traditional 16-bit models, BitNet cuts GPU memory use and energy consumption by up to 3.5 and 71 times, respectively, without compromising accuracy.
- BitNet b1.58 achieves comparable or better outcomes to FP16 models like LLaMA 3B in perplexity and language tasks from a 3 billion parameter size, with a 1-bit precision that speeds up computations, lowers energy costs, and scales efficiently to larger models, boosting performance and reducing latency.
Weekend takeaways
Founder & CEO, Group 8 Security Solutions Inc. DBA Machine Learning Intelligence
7 个月Grateful for your post!
Strategic Account Executive at SingleStore
7 个月Really insightful and concise! And subscribed!