Prompted by an interview with Jerry Liu of LlamaIndex posted by a colleague, I shared that we have been evaluating this since March or so, with rather poor results esp. when the data basis gets larger. I contributed a patch to add GPT-4 support, but as it turned out, the problems are not with the backend LLM, but IMHO a fundamental design issue ... Read more 28 Jun 2023 - less than 1 minute read
There was quite some insightful AI-themed content at Figma Conference “Config 23”: GPT-4 case study by Duolingo: “Designing with AI: building the flagship GPT-4 language product”: insights into the design process of an entirely new product (Duolingo Max) around GPT-4, offered from a perspective of non-engineers. One example of what they... Read more 26 Jun 2023 - 2 minute read
Soumith Chintala supports the rumor that GPT-4 is based on the Mixture Of Experts architecture: GPT-4: 8 x 220B experts trained with different data/task distributions and 16-iter inference. This would be a great explanation for the error pattern I saw earlier with GPT-4 but could not quite describe: it was as if individual parts/aspects of ... Read more 21 Jun 2023 - less than 1 minute read
Microsoft Azure OpenAI on your data enables you to run supported chat models such as ChatGPT and GPT-4 on your data without needing to train or fine-tune models. Blog post incl. video: Microsoft Tech Community Documentation: Microsoft Learn (also mentioned in Mercedes-Benz spotlight: Azure Blog) Read more 20 Jun 2023 - less than 1 minute read
When talking/writing about OpenAI, you may want to align with their Brand Guidelines. E.g., there is no „Chat-GPT 4.0“. 😇 Speaking of which, I am currently reading The Alignment Problem. Not finished, but seems like a good & relevant recommendation to make. Read more 19 Jun 2023 - less than 1 minute read