New today: You can now subscribe to Poe within the Android app! Subscribing provides access to GPT-4 (including 32k), Claude 2, and all user-created bots that use them as a base. Subscriptions work across platforms, so if you subscribe on Android you can use your access anywhere.
LLMS
-
LLMs Generate Synthetic Datasets Solving Data Labeling
By
–
Harnessing LLMs to Generate Synthetic Datasets In the context of machine learning, there's often a lack of labeled data or very little of it. This data collection and labeling process typically ends up taking a lot of effort and time. With the advent of LLMs, the paradigm
-
Understanding LLMs: Learning Deep AI Fundamentals First
By
–
AI Scientist Luis Serrano likens it to learning to drive a car before familiarizing yourself with its intricate parts, suggesting a deep AI enthusiast must intimately understand their Large Language Models (what a great analogy!)
-
Understanding LLMs’ Core Infrastructure: Embeddings, Transformers, Attention
By
–
Understanding AI and LLMs' basic infrastructure, like embeddings, transformers, and attention, aids in effectively using these models.
-
Study Questions LLM Trustworthiness for Critical Tasks
By
–
Large language models present powerful capabilities, but a new study suggests they are not trustworthy enough for critical tasks. “Under the hood, these are machine learning models with vulnerabilities," says HAI faculty affiliate @sanmikoyejo
. -
DEF CON Hackers Launch Attacks on Generative AI Models
By
–
Legions of DEF CON hackers will attack generative AI models https://
bit.ly/3Luys3h #AI #MachineLearning #DeepLearning #LLMs #DataScience -
Multi-Agent Systems Defined by Competing Prompting Agents
By
–
Deux agents ou plus qui se promptent en concurrence jusqu'au résultat hors du commun = système multi agents
-
Language Agent Tree Search Unifies Reasoning Acting Planning
By
–
Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models Zhou et al.: https://
arxiv.org/abs/2310.04406 #ArtificialIntelligence #DeepLearning #MachineLearning -
Transformers Library Bug Breaks Training Arguments in Latest Version
By
–
FYI I think the latest Transformers from main might have broken this, since I get
“` File "transformers/src/transformers/training_args.py", line 1513, in __post_init__ if self.half_precision_backend != "apex" and len(self.sharded_ddp) == 0:
TypeError: object of type 'bool' -
Fine-Tuning Linear Adapters for Embedding Models
By
–
Fine-Tuning a Linear Adapter for Any Embedding Model https://
bit.ly/44YcZqx #AI #MachineLearning #DeepLearning #LLMs #DataScience