Nils Durner's Blog Ahas, Breadcrumbs, Coding Epiphanies

Web Resources in LLM Training, and Hallucination

There is some concern over whether companies and content creators are, or should be, in agreement with their data being utilized for training AI models. The Google C4 dataset, a massive collection of web-crawled data, is at the center of this debate. Read more...

Hands on Training an Own OpenAI Model (Fine-Tuning 2.0)

Upfront summary: model training has evolved significantly. Adding new knowledge may actually be viable now. Read more...

Amazon Bedrock is GA

Amazon Bedrock is Generally Available: Announcement. For everyone who has enjoyed preview access, these details caught my attention: “Amazon Titan model usage will be billed starting 00:00 PST on Sept 28, 2023.” “Effective Sept. 18, 2023: Titan Text G1 - Express is the new name for model Titan Text Large.” Read more...

Model "Air"

~Friday fun~ Saturday Shenanigans: AI models don’t have bugs, but they may have “an air”. (screenshot feat. Megan Leszczynski) Read more...

Google Bard's Vulnerability to Adversarial Images

Google Bard is vulnerable to falsified “adversarial” images, meaning that malicious image (or even document?) creators can mislead it to form wrong conclusions and circumvent its security mechanisms: Read more...