On the Techspot article that “Samsung warns fab employees of ChatGPT after confidential data leaks”: this reporting lacks context, so I’ll dispense… This seems more like a PR issue at its core. From the OpenAI Helpcenter: OpenAI does not use data submitted by customers via our API to train OpenAI models or improve OpenAI’s service offering.... Read more 06 Apr 2023 - 1 minute read
Sparked by a blog post by Github on Copilot productivity enhancements and developer happiness, there was lively reflection on LLMs as a developer tool. One developer brought up the error scenario where ChatGPT would get stuck in a conversation loop. As a remedy, I offered this advice: sometimes it feels like it’s cornered and can’t get out. In s... Read more 05 Apr 2023 - 1 minute read
I have created Geppetto’s Workbench over the weekend, to simplify working with the OpenAI Playground in several ways: defaults to GPT-4 sets parameters to encourage factual results (and curb hallucination & confabulation) declutters the UI Future plans/wishlist: multiple presets/collections of defaults auto continuation of long... Read more 03 Apr 2023 - less than 1 minute read
Some talk about Claude by Anthropic today, accessible through poe.com. The Claude model scores rather favourably in the benchmark on truthfulness “TruthfulQA”, but lags behind the OpenAI GPTs by a wide margin. From the Stanford AI Index 2023 that has just been released (386 pages; be sure to enjoy in Books.app, rather than Safari 😊) Notably,... Read more 03 Apr 2023 - 1 minute read
I got the question what the difference between the OpenAI Playground “in chat mode” and chat.openai.com is. So for everybody’s benefit, my answer reproduced below. If you’re not familiar at all, see this screenshot for what I am talking about: Note that this pay-per-use, so a credit card is required (but a free tier seems to exist for new acco... Read more 01 Apr 2023 - 1 minute read