News/AI Models
OpenAI is Hoping Publisher Partnerships and GenAI Will Give SearchGPT an Edge in Competitive Market
OpenAI's new SearchGPT prototype aims to chip away at Google's dominance in web search by leveraging generative AI and partnerships with publishers to provide users with more efficient and conversational search results. Key details of SearchGPT: OpenAI has revealed limited information about its new search engine prototype, which is designed to help users find information more quickly and easily: SearchGPT uses generative AI to gather links and provide answers to user queries in a conversational manner, potentially integrating with the popular ChatGPT chatbot in the future. The search engine will draw upon information from the broader web as well as...
read Jul 26, 2024Meta’s Open-Source AI Sparks Debate Over Safety, Innovation, and Accountability
A debate over open-source versus closed AI models is emerging, as Meta releases an open-source model while OpenAI keeps its code private. This development raises important questions about the implications of these different approaches for AI safety, competition, and innovation. Meta's open-source approach sparks controversy: Meta CEO Mark Zuckerberg has called for open-source AI development and released an open-source model, Llama 3.1, which the company claims can compete with closed models like OpenAI's ChatGPT. Anthony Aguirre, executive director of the Future of Life Institute, suggests that open-source models are incompatible with safety regulation, as they lack the necessary guardrails to...
read Jul 26, 2024IBM Says Quantum Computing May Solve AI’s Energy and Cost Crisis
The exponential growth in AI training costs is raising sustainability and equity concerns, but quantum computing may offer a solution. The AI training cost problem: The rapid increase in AI usage is driving up energy consumption and costs, which could surpass the US GDP by 2030 and limit AI's potential if left unchecked. ChatGPT queries consume at least 10 times more energy than Google searches, contributing to the unsustainable growth in AI's energy footprint. By 2030, the cost to train a single AI model once could exceed worldwide IT spending, creating a hard ceiling for the technology's development and accessibility....
read Jul 26, 2024How to Get on the SearchGPT Waitlist
SearchGPT prototype launched by OpenAI, aiming to revolutionize AI-powered search: OpenAI has surprise-revealed SearchGPT, a new AI-powered search engine running on GPT-4 models that has real-time access to internet-based information, potentially challenging Google's search dominance. Key details about the SearchGPT prototype: SearchGPT is currently a prototype being tested by OpenAI with a limited waitlist for early access: SearchGPT provides a large textbox for user queries and attempts to summarize and organize the information it finds, with links available in a sidebar. The results can be followed up with relevant queries using a chatbox, and summaries are expected to have in-line...
read Jul 26, 2024Microsoft Launches Serverless Fine-Tuning for Its Phi-3 Model
Microsoft introduces serverless fine-tuning for its Phi-3 small language model, allowing developers to adapt the AI for specific use cases without managing their own servers, intensifying competition in the enterprise AI market. Key features of Phi-3 small language model: Microsoft's Phi-3, a 3 billion parameter model, offers affordable performance on coding, common sense reasoning, and general knowledge, now as part of a family of 6 models with varying parameters and context lengths: Phi-3 models range from 4,000 to 128,000 tokens in a single input, with costs from $0.0003 to $0.0005 USD per 1,000 input tokens. Designed with safety features for...
read Jul 25, 2024Strategies for Ensuring AI Safety and Chatbot Flaws
The large language models (LLMs) powering chatbots like ChatGPT are capable of impressive feats, but still routinely produce errors and can be made to behave in undesirable or harmful ways. Making these AI systems safe and robust to misuse is a critical challenge as they become more widely deployed. Vulnerabilities rooted in fundamental properties of LLMs: Many of the problems with LLMs stem from how they work - by predicting the most likely next word based on statistical patterns in their training data: Performance can fluctuate wildly depending on how common the output, task or input text is on the...
read Jul 25, 2024NVIDIA’s NeMo Retriever Boosts Accuracy and Throughput for Enterprise LLMs
NVIDIA introduces NeMo Retriever NIM microservices to enhance the accuracy and throughput of large language models (LLMs) for enterprises by leveraging their proprietary data. Key components of NeMo Retriever: The microservices consist of embedding and reranking models that work together to efficiently retrieve the most relevant data for generating accurate responses: Embedding models transform diverse data into numerical vectors, capturing meaning and nuance, and are computationally efficient. Reranking models score the retrieved data based on its relevance to the query, improving accuracy but being more computationally complex. NeMo Retriever combines both model types to ensure the most helpful and accurate...
read Jul 25, 2024Google’s Gemini AI Chatbot Gets Major Upgrade, Enhancing Speed and Quality
Google's Gemini AI chatbot is receiving a significant upgrade with the Gemini 1.5 Flash AI model, which promises to enhance its performance and capabilities. Gemini 1.5 Flash brings speed and quality improvements: The upgrade to Gemini 1.5 Flash is expected to deliver faster response times and improved overall quality across various aspects of the AI's performance: Users can expect noticeable improvements in Gemini's reasoning abilities and image understanding, according to Amar Subramanya, VP of engineering for Gemini. The AI's context window, which determines the amount of text it can process at once, is being quadrupled to 32K tokens, enabling Gemini...
read Jul 25, 2024DeepMind’s AlphaProof AI Model Can Solve Math Olympiad Problems
The significance of DeepMind's progress in tackling complex math problems lies in the potential for AI to enhance mathematical research and education, despite current limitations. Key developments in DeepMind's math-focused AI: DeepMind introduced AlphaProof, a model specializing in math reasoning, and an updated version of AlphaGeometry, which focuses on geometry problems. The AI programs successfully solved four out of six problems from the International Mathematical Olympiad, demonstrating their proficiency in topics like algebra and geometry. Challenges and limitations: Math reasoning remains a difficult area for AI models, which have historically struggled with this domain due to their reliance on text-based...
read Jul 25, 2024Unstructured Data Poses Important Considerations for Privacy, Governance and Ownership in the AI Era
The rapid advancements in AI's ability to analyze unstructured data are raising important questions about data privacy and ownership. Key Takeaways: As AI systems become increasingly capable of extracting insights from vast amounts of unstructured data, it's crucial to consider the privacy implications: While unstructured data may seem less sensitive than structured databases containing personal identifiers, AI can still pull together inferences, timelines, and narratives that could be highly intrusive. The era of AI moving from structured data sets to a more general technology approaching "universal knowledge" is both thrilling and potentially terrifying from a privacy perspective. Advancements in AI...
read Jul 25, 2024Nvidia Bets Big on Custom AI Models with AI Foundry Launch
Nvidia's AI Foundry service aims to help businesses create and deploy custom large language models, signaling the company's push to capture a larger share of the booming enterprise AI market. Customization drives accuracy: Nvidia claims that customizing open-source models like Meta's Llama 3.1 for specific business use cases can significantly improve model performance: The AI Foundry service provides access to pre-trained models, high-performance computing resources through Nvidia's DGX Cloud, and the NeMo toolkit for model customization and evaluation. Nvidia reports seeing almost a ten-point increase in accuracy by simply customizing models for enterprise clients. NIM: Nvidia's unique approach to AI...
read Jul 24, 2024Mistral Large 2 Just Dropped, and It’s the Latest Open-Source Powerhouse AI Model
Mistral AI shocks the AI world with the release of Mistral Large 2, a powerful open-source model that takes on Meta's Llama 3.1 in the rapidly evolving AI landscape. Key details of Mistral Large 2 release: Mistral's new flagship model boasts 123 billion parameters and is licensed as open-source for non-commercial research use, while commercial applications require a separate license: The model offers advanced multilingual capabilities, supporting dozens of languages including English, French, Spanish, German, Italian, Portuguese, Arabic, Hindi, Russian, Chinese, Japanese, and Korean. Mistral Large 2 delivers strong performance in reasoning, code generation, and mathematics, making it ideal for...
read Jul 24, 2024Why Some Experts Believe Synthetic Data Will Degrade Future Models
The proliferation of AI-generated junk web pages poses a significant challenge to the future development and performance of AI models, as training on increasingly synthetic data can lead to degraded output quality and potential model collapse. Key takeaways from the research: A study published in Nature demonstrates that the quality of an AI model's output gradually deteriorates when trained on data generated by other AI models: The effect worsens as subsequent models produce output that is then used as training data for future models, likened to taking photos of photos and eventually being left with a dark square or "model...
read Jul 24, 2024Stability AI Just Announced “Video 4D” with with Dynamic 3D Object Viewing
Stability AI has introduced Stable Video 4D, a new generative AI model that adds a new dimension to video generation by accepting video input and generating multiple novel-view videos from 8 different perspectives. A next-gen AI video model: Stable Video 4D builds upon Stability AI's existing Stable Video Diffusion model, taking the concept further by enabling users to view dynamically moving 3D objects from arbitrary camera angles: The model combines the strengths of Stability AI's previously-released Stable Video Diffusion and Stable Video 3D models, fine-tuned with a carefully curated dynamic 3D object dataset. Stable Video 4D is a first-of-its-kind network...
read Jul 24, 2024Winning Patient Trust: Balancing Privacy and Progress in Medical Machine Learning Research
Winning patient trust is essential for advancing machine learning applications in medicine, which rely on access to large and diverse health datasets. Data sharing challenges: While openly available medical datasets have been highly beneficial for research, in many cases health data cannot be shared due to privacy concerns or participant preferences: Federated learning schemes allow models to be trained on local datasets without sharing the data directly, but additional privacy-preserving tools are still needed to prevent data reconstruction from model updates. Synthetic data generated by AI algorithms can help protect patient privacy by replacing or augmenting real datasets, but potential...
read Jul 24, 2024MIT Researchers Are Automating Neural Network Interpretability to Improve Transparency in AI
Researchers at MIT's CSAIL developed an AI system called MAIA that automates the interpretation of neural networks, enabling a deeper understanding of how these complex models work and uncovering potential biases. Key capabilities of MAIA: The multimodal system is designed to investigate the inner workings of artificial vision models: MAIA can generate hypotheses about the roles of individual neurons, design experiments to test these hypotheses, and iteratively refine its understanding of the model's components. By combining a pre-trained vision-language model with interpretability tools, MAIA can flexibly respond to user queries and autonomously investigate various aspects of AI systems. Automating neuron-level...
read Jul 24, 2024China’s Open-Source AI Strategy Challenges Western Dominance, Fosters Global Collaboration
For Alibaba and several Chinese AI startups, open-source AI presents an opportunity for faster commercialization and global recognition. Chinese companies are embracing open-source AI models as a strategy to compete with Western tech giants, find alternative paths to innovation, and gain international visibility. Key drivers behind China's open-source AI push: Chinese companies are betting on open-source AI for several reasons, including accelerating product development, finding niche markets, and attracting global developers: Alibaba's decision to open-source its Qwen model is aimed at growing its cloud business by encouraging developers to build applications using Alibaba Cloud and its open-source tools. For Chinese...
read Jul 24, 2024AI Startups Tackle Looming Data Shortage with Innovative Solutions
The AI industry is facing a looming data shortage as companies have already exhausted much of the available training data, but startups are exploring innovative solutions to address this challenge. Synthetic data emerges as a potential solution: Gretel, a startup valued at $350 million, is creating AI-generated synthetic data that closely mimics real information without the privacy concerns: Synthetic data has been used by companies working with sensitive information, such as patient data, to protect privacy while still providing valuable training data for AI models. Gretel's CEO, Ali Golshan, sees an opportunity to supply data-starved AI companies with fake data...
read Jul 24, 2024Meta’s Llama 3.1 Launch Ignites AI Arms Race, Challenging OpenAI’s Dominance
Llama 3.1 launch signals Meta's bold challenge to OpenAI's dominance in the AI market: Meta's release of its most advanced AI model to date, Llama 3.1, and its integration into the company's AI assistant, Meta AI, across multiple platforms, marks a significant escalation in the AI competition, directly targeting OpenAI's market-leading position. Open-source strategy disrupts the AI landscape: Meta's decision to make the powerful Llama 3.1 405B model openly available contrasts with the proprietary approaches of its competitors, particularly OpenAI, potentially reshaping the future of AI accessibility and development: The open-source release comes at a critical juncture for Meta, following...
read Jul 23, 2024Meta Just Released the Biggest and Best Open-Source Model Ever
Meta's release of the largest-ever open-source AI model, Llama 3.1, outperforms rival models like GPT-4o and Claude 3.5 Sonnet on several benchmarks, marking a significant milestone for the AI industry: The 405-billion parameter model was developed at an estimated cost of hundreds of millions of dollars, utilizing over 16,000 of Nvidia's H100 GPUs. Meta is working with more than two dozen major tech companies to help developers deploy their own versions of Llama 3.1, which is claimed to be more cost-effective than proprietary models. The release includes model weights for companies to train and tune Llama 3.1 on custom data,...
read Jul 23, 2024China’s AI Chatbots are Becoming More Accessible Across the World
China's AI chatbots are becoming more accessible to users worldwide, enabling experimentation and comparison with Western models. Key pathways to access: Chinese AI companies are providing various avenues for users outside China to interact with their large language models (LLMs) and chatbots: Some platforms, such as Doubao, ChatGLM, and DeepSeek, allow direct registration using non-Chinese phone numbers, granting access to a range of AI features like text and image generation. Open-source AI community Hugging Face hosts demos of several Chinese LLMs, including Alibaba's Qwen and Tencent's Hunyuan-DiT, which can be accessed without a Chinese number. Alibaba's ModelScope platform, a Chinese...
read Jul 22, 2024Sakana’s Latest AI Models Are Recreating Historic Japanese Ukiyo-e Art
The AI startup Sakana AI has released two new image generation models focused on recreating Japan's historic ukiyo-e art style, aiming to spread appreciation for Japanese culture and find applications in education and literature. Bringing ukiyo-e art to the modern era with AI: Sakana AI's new models, Evo-Ukiyoe and Evo-Nishikie, generate images closely resembling the Japanese ukiyo-e art style that flourished between the 17th and 19th centuries: Evo-Ukiyoe is a text-to-image model that generates ukiyo-e style images based on text prompts, even allowing for the inclusion of modern elements like hamburgers or laptops, though results may sometimes deviate from the...
read Jul 22, 2024Why Trolling LLMs to Elicit Outrageous Responses Has Legitimate Scientific Value
The hidden value of trolling large language models: Internet trolls fiddling with prompts to elicit outrageous or nonsensical responses from LLMs are actually engaged in a legitimate scientific pursuit that reveals the models' limitations and challenges the deceptive practices of LLM vendors: Contrary to vendors' stated objectives of making models helpful and accurate, they pour significant resources into responding to every viral troll-generated LLM transcript, suggesting their true priorities may differ from their public stance. Commercial LLM applications rely on the models appearing human-like as a proxy for their reliability, since customers need to understand how and when the models...
read Jul 22, 2024Cohere Secures $500M, Valuing AI Startup at $5.5B
Cohere, a Toronto-based enterprise AI startup, has raised $500 million in a Series D funding round, valuing the company at $5.5 billion amid growing skepticism surrounding the return on investment of generative AI tools for businesses. Significant investment despite industry concerns: The substantial funding round, led by PSP Investments and joined by new and returning investors, demonstrates continued investor enthusiasm in the enterprise AI sector: New investors include Cisco Systems, Fujitsu, AMD Ventures, Magnetar, and Export Development Canada, while returning investors consist of Oracle, Salesforce Ventures, and Nvidia. The investment comes on the heels of other notable fundraising efforts in...
read