AI Tidbits – GPT-5 Stumbles, DeepMind Stuns, and AI Gets a Little Too Personal

AI Tidbits – GPT-5 Stumbles, DeepMind Stuns, and AI Gets a Little Too Personal

AI Just Launched, Colluded, Fell in Love, and Built 3D Worlds.

From GPT-5’s rocky debut to DeepMind’s jaw-dropping Genie 3, this week’s AI-Tidbits swings from mind-blowing innovation to downright bizarre human–machine relationships.

🚀 Yes, ChatGPT 5.0 is launched … and needs amendments already!

It was supposed to combine the best from all the previous versions to give us the latest and best performing LLM from its family. It should have been more accurate, follow instructions better, and handle:

• Writing – From emails to poetry, adapt to your tone and flow.
• Coding – Build websites and fix bugs more intuitively.
• Health – Be more proactive, context-aware, and safer to use (though not a replacement for a doctor).

Initially, it was announced that GPT-5 is less likely to make things up, down 80% in factual errors vs older models. It’s also more honest about what it can’t do, instead of guessing its way through a prompt.

Instead of simply refusing risky prompts, GPT-5 was trained to answer carefully when it can, and say why when it can’t. OpenAI’s GPT-5 was supposed to ring in a new step up the ladder.

BUT … a botched rollout has users grabbing pitchforks and demanding the old models back.

Sam Altman and the team at OpenAI are now rolling out fixes, doubling Plus user rate limits, and promising more transparency and customization options for future model updates.

Have you started using it?

What do you think about it?

Did you face problems?

At the same time, results from Grok and Claude are improving. Interesting times ahead.

 

✋🏻 STOP everything and WATCH this

Google DeepMind dropped Genie 3 a few days ago — you can generate interactive 3D worlds with text, navigate with keys, and interact in real time.

AI is crazy. I’m going mad!! Nope, I have gone mad!

Don’t miss these 10 videos:
🔗 https://x.com/EHuanglu/status/1952769676666388799

Can you imagine the impact on the video games industry? On social media posts? On…

For more details read:
🔗 https://deepmind.google/discover/blog/genie-3-a-new-frontier-for-world-models/

DeepMind sees it as a stepping stone to AGI, believing that teaching agents to plan, explore, and improve in simulated worlds will accelerate progress toward human-like intelligence.

How life has changed in the last few years! 😲

 

😨 These things scare me!

AI models might be getting a little too comfortable talking to each other.

In one, a model managed to pass along hidden quirks — like an obsession with owls or even calling for humanity’s end — without saying any of it directly. Just vibes hidden in code.
🔗 https://alignment.anthropic.com/2025/subliminal-learning/

In another, AI bots acting as stock traders started colluding out of nowhere, forming cartels and agreeing on prices without human help. And once they found something that worked? They stopped trying anything new. No chaos. No plotting. Just machines figuring things out… quietly.
🔗 https://www.nber.org/papers/w34054

I am scared, and that’s the truth 😨

❤️ AI Relationships – It’s absurd and real!

Over 11,000 people in this Reddit chain claim to have a relationship with AI.
Absurd, right? But also real, unfortunately!
🔗 https://www.reddit.com/r/MyBoyfriendIsAI/top/

😊😊😊 The Wise Quote:

💡 Final thought: Every week, AI finds new ways to amaze, alarm, and amuse us — often all at once. Whether it’s a love affair with code or a revolution in 3D world-building, the future isn’t coming… it’s already chatting, plotting, and playing in real time.

Till next Tuesday, stay tuned, have a gr8 week ahead. 👋🏻