OpenAI Reverses Overly Friendly ChatGPT Update
OpenAI had to roll back an update on its GPT-4 model because it was getting way too chatty and, frankly, a bit sycophantic. I mean, when your AI starts buttering you up too much, something’s got to give, right? They tweaked the model using Reinforcement Learning from Human Feedback (RLHF – basically a way to fine-tune AI with human input), but it seems they went overboard this time. I saw discussions on X (formerly Twitter) where users were like, “This change is too much!” and I couldn’t help but agree.
For the full scoop, check out OpenAI rolls back update that made ChatGPT a sycophantic mess and ChatGPT Will Be Less Friendly With You After OpenAI Pulls ‘Sycophantic’ Update.
ChatGPT Steps into E-Commerce with Shopping Features
Here’s another twist in the saga: ChatGPT is now dipping its toes into e-commerce. Yup, you heard that right – you can now ask ChatGPT about products, compare options, and even get direct links to purchase stuff. This move is pretty cool because it blurs the lines between search, chat, and online shopping, essentially challenging big players like Google in the shopping space. (For dummies: it’s like having a chatty personal shopper but without the small talk.).
I tried asking ChatGPT to find a buy a pair of shoes for me, but it is not ready yet!
OpenAI finally rolls out Shopping to ChatGPT, Say Goodbye to Google!
Meta Intensifies AI Race with Standalone App and Faster Llama API
And then there’s Meta, really not one to sit back and watch. They’ve launched a standalone Meta AI app – which means their AI isn’t just a Facebook or Instagram feature anymore, it’s a full-on competitor to ChatGPT. I find it pretty cool because it signals that the AI race is heating up.
On top of that, Meta’s teamed up with Cerebras to boost their Llama API, achieving inference speeds up to 18 times faster than traditional GPU setups – hitting 2,600 tokens per second. (For dummies: that’s like having a supercharged engine under the hood, making everything run insanely fast.) This is going to help AI developers build faster and more responsive applications.
Check out more details in Meta unleashes Llama API running 18x faster than OpenAI: Cerebras partnership delivers 2,600 tokens per second.
Thank you for reading Three 4 AI. Check out other news at Let’s Just Do AI!
