SMRTR AI

AI stories

Get hand-picked daily summaries of the best, most informative AI articles from around the web. Browse recent summaries, then jump to the original source when you want the full article.

Latest story• Sep 25, 2024

Microsoft’s Nuclear Deal Could Spark New Market for AI Power

Microsoft has secured a 20-year deal to restart the Three Mile Island nuclear plant in Pennsylvania to power its AI operations, becoming the first major tech company to directly invest in nuclear power for AI. This highlights the enormous energy requirements for advancing AI technologies. The 835-megawatt reactor will provide reliable, high-capacity power as global AI spending is expected to reach $300 billion by...

Read summary Original source

Latest stories

‘Robot lawyer’ company faces $193,000 fine as part of FTC’s AI crackdown

DoNotPay, a self-proclaimed AI "robot lawyer" company, has settled with the FTC for $193,000 over false claims about its AI's capabilities. The settlement requires DoNotPay to warn past subscribers and provide evidence for future...

Read summary Original

The Most Capable Open Source AI Model Yet Could Supercharge AI Agents

Ai2 has released Molmo, an open-source AI model with visual and conversational abilities that could enable developers to create more advanced AI agents. The model comes in various sizes, including a 70-billion-parameter version...

Read summary Original

Meta’s going to put AI-generated images in your Facebook and Instagram feeds

Facebook and Instagram are testing a feature that adds AI-generated images to users' feeds based on their interests and current trends. The "Imagined for You" content may include images incorporating users' faces, with options to...

Read summary Original

Meta AI gets celebrity voices and lip-synced translations

Meta AI's voice assistant now works across multiple platforms, offers celebrity voice options, analyzes images, and is testing an AI translation tool for Instagram Reels that dubs and lip-syncs creators' speech in different...

Read summary Original

Introducing Llama 3.2 models from Meta in Amazon Bedrock: A new generation of multimodal vision and lightweight models

Llama 3.2, Meta's latest AI models in Amazon Bedrock, pack multimodal vision, 128K context, and 1B-90B parameters, delivering image reasoning, text generation, and translation prowess while offering lightweight options for edge...

Read summary Original

SMRTR AI

Get the next batch of curated stories in your inbox.

This archive is built from SMRTR newsletter stories. Subscribe for hand-picked stories without the extra noise.

More recent stories

How Meta’s AI Advancements May Impact Social Commerce

Meta showcased major AI enhancements at its Connect conference, potentially transforming online shopping through voice assistants and image recognition. Over 400 million people use Meta AI monthly, with 185 million weekly users....

Read summary Original

How sales teams can use AI today to optimise conversions

AI is turbocharging B2B sales, supercharging everything from personalization to forecasting, helping sales teams close more deals and focus on relationship-building while streamlining operations and staying ahead of the...

Read summary Original

Figma to React with CodeParrot AI: Build Dashboard in minutes

CodeParrot, a VSCode extension, converts Figma designs to React code and other frameworks, streamlining the development process. It offers speed, accuracy, and customization options, allowing developers to generate...

Read summary Original

OpenAI rolls out Advanced Voice Mode with more voices and a new look

OpenAI is expanding access to Advanced Voice Mode (AVM) for ChatGPT's paying customers. The feature now has a new blue sphere design and five additional nature-inspired voices, bringing the total to nine. AVM includes...

Read summary Original

88% of US parents see AI as essential to their children's education - survey

Parents increasingly recognize AI's importance for their children's futures, with 88% believing AI knowledge will be crucial. However, 81% are unsure if AI is part of current curricula. The survey reveals a need for improved AI...

Read summary Original

New AI model breaks barriers in cross-modality machine vision learning

A new cross-modality machine vision AI model has been developed by Chinese researchers, overcoming limitations of traditional single-domain models. The WRIM-Net creates global region interactions to extract detailed associations...

Read summary Original

The Raspberry Pi of AI: Korean startup wants to bring ultra-cheap, high performance NPUs to billions of devices using its own secret quantization sauce, with $1/TOPS target on the horizon

DeepX, a South Korean AI company, presented its V1 and M1 chips at the 2024 Embedded Vision Summit. The V1 SoC features a 5-TOPS NPU and quad-RISC-V CPUs for under $10, while the M1 accelerator offers 25-TOPS performance for...

Read summary Original

James Cameron is joining Stability AI’s board of directors

James Cameron, renowned director and early tech adopter, has joined Stability AI's board of directors. This move signals growing interest in AI within the film industry, with Cameron highlighting the potential of generative AI in...

Read summary Original

Language agents help large language models 'think' better and cheaper

Zero-Shot AgentInstruct, a new method from Washington University, uses large language models to create task-specific instructions, significantly improving smaller models' performance in reasoning tasks, especially math and logic,...

Read summary Original

The Ownership Issue: Three Big Questions With New AI Models

Artificial intelligence is raising significant concerns about data ownership and intellectual property rights. Key issues include whether AI outputs based on human creations constitute fair use, whether companies should build...

Read summary Original

Google Reports 2 Million Developers Building With Its GenAI Solutions

Companies are leveraging Google's Gemini AI for diverse applications, with millions of developers on board and users saving nearly two hours weekly, while major brands report boosted engagement and performance across customer...

Read summary Original

How to build your own custom AI voice agent - for free

Supernormal has launched Voice Agents, a platform allowing users to create custom AI voice assistants in under two minutes. These agents can handle calls independently and integrate with popular meeting software. Users can choose...

Read summary Original

Spotify’s AI playlist builder is now available in the US

Spotify's AI Playlist feature, now available in more English-speaking areas, lets Premium users create custom playlists from text prompts, generating 30 songs that can be refined for better results.

Read summary Original

Better Instructions, Better Results: A Look at Prompt Optimization

LLMs can now optimize prompts for themselves and others, boosting task performance significantly. The OPRO method uses one LLM to generate better instructions for another, improving accuracy on math and reasoning tasks like GSM8K...

Read summary Original

Source-Disentangled Neural Audio Codec (SD-Codec): A Novel AI Approach that Combines Audio Coding and Source Separation

SD-Codec, a new neural audio codec, splits audio sources into separate codebooks, boosting interpretability and control while maintaining the efficiency gains of tokenized compression for music, speech, and sound effects.

Read summary Original

Practical Recipe for an AI-based Chatbot in the Browser

Build a browser-based AI chatbot using Transformers.js, with a simple interface and web worker for smooth performance, using a Hugging Face model—all explained in a new tutorial that also suggests ways to enhance your creation.

Read summary Original

BART Model for Text Summarization

BART is a powerful pre-trained model for natural language processing tasks, combining bidirectional and auto-regressive transformers. It excels in comprehension and generation tasks, achieving state-of-the-art results in...

Read summary Original

Microsoft claims its new tool can correct AI hallucinations, but experts advise caution

Microsoft unveils Correction, a service to automatically revise AI-generated text containing factual errors. The tool uses small and large language models to flag and rewrite potential hallucinations by comparing text to trusted...

Read summary Original

Back to all archives Load more stories