1
2023: The Year of AI
Explore the significant AI advancements, impactful partnerships, and legal debates that defined 2023.
New
The latest blog posts to take off on Hacker News. 387 posts.
2
Text-to-CAD
Introducing our machine learning model for generating CAD models from a text prompt.
3
Mistral 7B Fine-Tune Optimized
The OpenPipe platform has migrated to Weights & Biases and CoreWeave. Browse migration details and the OpenPipe technical archive.
4
Waymo outperforms comparable human benchmarks over 7M+ miles
Safety leads everything we do at Waymo. This year alone, Waymo has served over 700,000 ride-hailing trips with public riders and no human driver. We couldn’t have hit that milestone without putting safety front and center, and we are working hard to improve the measurement, transparency, and performance of our fleet.
5
VideoPoet A large language model for zero-shot video generation
A Large Language Model for Zero-Shot Video Generation. VideoPoet demonstrates simple modeling method that can convert any autoregressive language model into a high quality video generator.
6
7
RAG to Riches
Today I will be regaling you with tales of marvelous cherry-picked new developments in AI, from Sam Altman's recent Uno Reverse incident to bear spraying job-hungry engineers.
8
Advancements in machine learning for machine learning
Posted by Phitchaya Mangpo Phothilimthana, Staff Research Scientist, Google DeepMind, and Bryan Perozzi, Senior Staff Research Scientist, Google Re...
9
Do large language models need all those layers?
Finding that 70% of attention heads and 20% of feed-forward networks can be excised with minimal effect on in-context learning suggests that large language models are undertrained.
10
Data exfiltration from Writer.com with indirect prompt injection
Authors: PromptArmor and Kai Greshake
11
12
Weak-to-Strong Generalization
We present a new research direction for superalignment, together with promising initial results: can we leverage the generalization properties of deep learning to control strong models with weak supervisors?
13
FunSearch: Making new discoveries in mathematical sciences using LLMs
We introduce FunSearch, a method for searching for “functions” written in computer code, and find new solutions in mathematics and computer science. FunSearch works by pairing a pre-trained LLM, whose goal is to provide creative solutions in the form of computer code, with an automated “evaluator”, which guards…
14
Google Imagen 2
Today, we’re sharing a significant upgrade to Google Cloud’s image-generation capabilities with Imagen 2 , our most advanced text-to-image technology, which is now generally available for Vertex AI customers on the allowlist (i.e., approved for access).
15
16
Phi-2: The surprising power of small language models
Phi-2 is now accessible on the Azure model catalog. Its compact size and new innovations in model scaling and training data curation make it ideal for exploration around mechanistic interpretability, safety improvements, and fine-tuning experimentation on a variety of tasks.
17
A new old kind of R&D lab
Answer.AI will figure out the fundamental research needed to tame AI, and the development path needed to make it useful in practice
18
GigaGPT: GPT-3 sized models in 565 lines of code
GigaGPT is Cerebras’ implementation of Andrei Karpathy’s nanoGPT – the simplest and most compact code base to train and fine-tune GPT models.
19
Mistral: Our first AI endpoints are available in early access
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
20
Mixtral of experts
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
21
22
Paving the way to efficient architectures: StripedHyena-7B
StripedHyena-7B is an open-source model built beyond the Transformer architecture, matching Llama-2 and Mistral 7B while running faster on long sequences.
23
StableLM Zephyr 3B
StableLM Zephyr 3B is a new chat model representing the latest iteration in our series of lightweight LLMs, preference tuned for instruction following and Q&A-type tasks.
24
Purple Llama: Towards open trust and safety in generative AI
Meta AI announces Purple Llama, a project for open trust and safety in generative AI, with tools for cybersecurity and input/output filtering.
25
Audiobox: Meta's new foundation research model for audio generation
Audiobox is Meta’s new foundation research model for audio generation.
26
Behind the scenes of Sound ID in Merlin – Identify birds using your phone (2021)
What is Sound ID? Today we announced one of our biggest breakthroughs—Sound ID, a new feature in the Merlin Bird ID app—and a major leap forward in sound identification and machine learning
27
Long context prompting for Claude 2.1
Claude 2.1 excels at retrieving information across its 200K context window, with a simple prompt adjustment improving accuracy from 27% to 98%. Claude 2.1’s performance when retrieving an individual sentence across its full 200K token context window.
28
29
Cloud TPU v5p and AI Hypercomputer
The new TPU v5p is a core element of AI Hypercomputer, which is tuned, managed, and orchestrated specifically for gen AI training and serving.
30
Gemini: Google's most capable AI model yet
Gemini is our most capable and general model, built to be multimodal and optimized for three different sizes: Ultra, Pro and Nano.