Nils Durner's Blog Ahas, Breadcrumbs, Coding Epiphanies

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...

Document Processing

Quick recap on recent advancements in Document-processing below. Read more...

ChatGPT-on-GPT-4: Knowledge Cut-off and Performance

ChatGPT-on-GPT-4 now reports knowledge cut-off in January 2022 and will answer questions even beyond. However, anything I‘ve tried turned out hallucinous (or vague/anticipated, e.g. the French elections in February). GPT-4-via-API still reports cut-off as of September 2021. Read more...

OpenAI's GPT 3.5 Instruct Model Announcement

OpenAI’s new GPT 3.5 Instruct model announced through E-Mail, pasted below. This is a drop-in replacement to text-davinci-003. It is widely used in code snippets around the web, so it’s good to finally have an optional available that combines the traditional TextCompletion interface and mode of operation with the benefits of GPT-3.5-turbo. Notew... Read more...