Google's marquee model makes an odd comparison to its primary competition.| www.supervised.news
Existing models like GPT-4 naturally have limitations around the amount of data they can ingest and spit out. New research is looking at breaking past those limitations.| www.supervised.news
Mistral's latest model might spark an actual race to the bottom on cost. Plus, Intel's CEO Pat Gelsinger hates CUDA—does Intel actually have an answer to it?| www.supervised.news
Plus: Hugging Face seemingly gets closer to being a creator hub in AI.| www.supervised.news
Embedding models are a key part of training and deploying large language models. OpenAI wants to win on price as competition on quality grows.| www.supervised.news
The most consequential tech story in decades could spark an even bigger race among startups to chart the future of AI.| www.supervised.news
Plus: Another odd AI leak over the weekend.| www.supervised.news
I've made the extremely difficult decision to end my full-time job with the newsletter, which will shift to more of a hobby. Plus, is the thinning out in the modern data stack finally here?| Supervised
OpenAI has decided it's time to try to handle one of AI's existential crises. Maybe this time it'll work?| Supervised
Plus: Notion, one of AI's brightest stars, rolls out a fuller-fledged AI companion, and several additional executives leave OpenAI.| Supervised
Enterprises are taking a more methodical approach when figuring out how to put AI tools into production—and considering much longer timelines.| Supervised
The data engineers are more important than ever these days.| Supervised
We're obsessed with generating thousands of tokens a second for a reason—and it isn't just to wow end users with text showing up on a screen really fast.| Supervised
The SaaS-ification of AI continues, with OpenAI and Google ready to go full 2010s to get people on board. Plus, durable execution returns as a topic du jour.| Supervised
AI can be far from achieving its potential, but it can also be really useful right now.| Supervised
A few things have changed since a Google researcher sounded the alarm on Google's risk to open source AI in a leaked memo last year.| Supervised
Venture firms are trying to get creative to get in front of the next hottest AI startups. But they have to look in new areas to find them.| Supervised
Plus: the future of an AI data marketplace continues to potentially lie within Snowflake and Databricks—if they can convince customers of the value.| Supervised
Snowflake is converging with its rival, Databricks, faster than ever before. And its new CEO is gearing up for a partial "pivot" into AI.| Supervised
Informatica, of all companies, is creeping back into conversations despite being a legacy vendor. Plus: what happens when RAG isn't enough?| Supervised
What was a research organization is now following dreams of being the next Apple. Plus: Why is everyone suddenly talking about WebGPU again?| Supervised
Another once specialized database technology from the previous big data era might find its way into AI products thanks to retrieval augmented generation.| www.supervised.news
Plus: OpenAI and Google are doing some stuff next week.| www.supervised.news
Plus, a Llama 3 price war is brewing, and what Snowflake's new model means for pre-training.| www.supervised.news
OpenAI now has a batch processing API. But this time around, it’s dealing with more than just a handful of startups—including Snowflake and Databricks.| www.supervised.news
Plus: the death of pre-training seems greatly exaggerated.| www.supervised.news
Long context windows provide an alternative to RAG, but many companies are still seeing an appetite.| www.supervised.news
Let's examine where a structured data startup sits in an unstructured data world. Plus: a new open source framework for AI development.| www.supervised.news
Startups have a new way to tackle idle GPUs: flooding them with tokens. Plus: Apple's potential license for Google's Gemini.| www.supervised.news
AI endpoints are morphing into a race to the bottom, and OpenAI may have trouble winning on cost.| www.supervised.news
Rather than putting out the best model embeddings, it's trying to beat everyone on cost. That should sound familiar.| www.supervised.news
The tools built for managing the data under AI are just as important as the models themselves.| www.supervised.news
I have to do this, apparently, like everyone else. So here are some shots in the dark as to what might happen for a field that's shown it's impossible to predict.| www.supervised.news
If RAG is the solution to hallucinations, how do you determine who gets to see what? Plus: another technique to improve performance emerges in the wild.| www.supervised.news
