TLDR AI 2023-10-02

Gretel's novel tabular large language model reliably generates synthetic tabular data with zero-shot prompting. You can use it to create a new correlated dataset with the correct datatypes, based on an SQL schema or query; or to expand existing datasets by automatically adding columns and records, or filling in gaps and null values.

The LLM was trained on half a trillion tokens of public data and generates highly realistic tables while ensuring relationships between records remain consistent. It’s even aware of regional differences (company names, emails, etc.) Request early access here

AI is great at doing tasks, but not so much doing full jobs. Sam Altman argues that AI will enable humans to focus on different work that AI can’t do.

Building an online store requires lots of high-quality product and marketing images. This is an early demo of a background replacement tool built with Stable Diffusion XL by the Shopify AI team that makes it easy to use your existing product images to make something new.

Meta has released Llama 2 Long, a new AI model that outperforms GPT-3.5 Turbo and Claude 2 for long user prompts.

This study presents Transformer-VQ, a new design of transformer that processes attention faster thanks to unique vector-based keys and caching.

This study uses visual and language-based tools to improve how systems recognize real vs. fake faces.

One of the coolest and simplest vision papers in recent weeks. Vision Transformers use "useless" pixel values as places to store global information. This makes attention maps uninterpretable. However, if you add a simple [reg] token to the vocabulary, the model will use it and won't store information in pixel values.

Researchers have developed a method to create videos that closely match a sound they're paired with, both in overall theme and moment-by-moment details.

SapientML is an AutoML technology that can learn from a corpus of existing datasets and their human-written pipelines and efficiently generate a high-quality pipeline for a predictive task on a new dataset.

With all the new models coming out, it can be somewhat challenging to find a way to pull and run them. It's also hard to remix and remake them easily. This neat little tool, Ollama, makes it easy to test the latest models.

ChatGPT and other AI technologies, while critiqued as "unscientific," can be viewed as an "aspiring science" akin to how alchemy preceded chemistry.

Dataiku has unveiled the LLM Mesh, addressing the critical need for an effective, scalable, and secure platform for integrating LLMs in the enterprise. With the LLM Mesh sitting between LLM service providers and end-user applications, companies can choose the most cost-effective models for their needs to ensure the safety of their data and responses.

More startups are popping up to help us read and manipulate our mental states and help us relax, learn, and reduce pain. Neuralink, Mendi, and FocusCalm are just some of these startups. These companies will gather data from users if they are given access to their brains. Who owns this data?

Never deal with ChatGPT’s character limit again, as BulkCorrector segments your text into manageable pieces for bulk corrections.

A subreddit about AI-driven music, speech, audio production, and all other emerging AI audio technologies.

Getty Images, in collaboration with Nvidia, has launched "Generative AI by Getty Images," a tool that allows users to create images using Getty's extensive photo library.