GPT-4o, OpenAI's latest language model that has just been made freely available, has major safety flaws, an investigation by Radio-Canada's disinformation-busting unit, Décrypteurs, has uncovered.
One tip for using ChatGPT that I am going to start using is to save prompts as a starting point so that I can always set the expectations for the rest of the thread. One way I use it is with shell scripting. I can give it all of my code style preferences and general code standards with a saved prompt which will save time when I start a new thread. #ChatGPT#LLM
Back in 2022, Anthropic CEO Dario Amodei chose not to release the super-powerful AI chatbot, Claude, that his company had just finished training, opting instead to focus on further internal safety testing. That move likely cost the company billions — three months later, OpenAI launched ChatGPT.
Having a reputation for credibility and caution in an industry that appears to have thrown a large chunk of it to the wind is not a bad thing though. Claude is now in its third iteration, but that caution remains, with the company pledging not to release AIs above certain capability levels until it can develop sufficiently robust safety measures.
TIME’s interview with Amodei gives an insight into what the AI industry might look like when safety is considered a core part of the strategy.
Isn't weird how I can take a paragraph from a book like The Hobbit (page 173 was my test) with no attribution or mention of that work. I can say I wrote this and literally a single word detail like Bilbo or Sting & #ChatGPT can magically spit out the names of things unmentioned like the dwarves, write very close to perfect before and after paragraphs, and more. It's almost as if the entire contents of that copyrighted work is in there? “fair use" really is stretched a lot in #AI
There have been a lot of rumors about Apple working with OpenAI to use ChatGPT in the forthcoming iOS 18 update. However, a new report has revealed that discussions have been going on for the best part of a year, and that has got Microsoft worried about how this could impact its own deal with the AI giant. @9to5Mac has more.
Yeah - I appreciate Sal's enthusiasm for AI but the combination of the current generation of bots confidently saying nonsense to kids seems ... problematic.
@bkeegan since LLMs are trained with Wikipedia data, should the Wikimedia Foundation try to get a deal with OpenAI to fund its operations? Seems like in their collective best interests.
@andresmh@bkeegan I assume they are working on it, since they’ve been defacto “selling” Wikipedia content (really: enterprise-quality real-time high-volume APIs to the same data) for a while to the other big trainers (Meta, Google, etc.): https://enterprise.wikimedia.com/
Today, I took ChatGPT's Data Analyst for a spin. You've probably seen the fancy advertising videos: just drop in a dataset and AI does all the analysis for you?! Let's see ...