I agree with him, comparing a year before to now the people using AI to write code has changed rapidly, and we can vote for what happens in the next 6 months
Tested DeepSeek-Janus-Pro-7B for img gen, where they mentioned it’s compatible w/ DALL-E. While results are decent, they aren’t as refined; detailing feels a bit off.
Still, being free now is a big plus, and with time, it could improve greatly. It’s just Jan 2025, after all.
Some personal news: After four years working on safety across @openai, I left in mid-November. It was a wild ride with lots of chapters - dangerous capability evals, agent safety/control, AGI and online identity, etc. - and I'll miss many parts of it.
DeepSeek has done it again, launching the open-source AI Image generation model, *Janus-Pro*. It outstrips OpenAI's DALL-E3.
Clearly, 2025 its DeepSeek's year.
Check it out: 👇
For New Delhi, it's //chucks.outpost.ghosts!
what3words is a system that assigns three simple words to every 3m x 3m square on Earth, making location sharing easy and precise.
https://t.co/b7TepSuxj0
Ever heard of what3words? 🌍 It's a game-changer for navigation!
Instead of complex addresses, just use 3 simple words to pinpoint any location.
Imagine sending someone to //apple.banana.orange for the perfect meet-up spot! 🚗✨
Check it out 👇