the selling point of the product is now “watch us burn more GPU runtime” because no normies can tell the difference between a $200 a $2000 and a $20000 tier
but unfortunately more runtime ≠ better answers and once normies realize that, the “bubble” will pop
I covered Flora really early in and thought it had some really cool ideas.
I popped in for this official launch; and man…they’ve been cooking.
I’ll take a deeper look at it soon, there’s a lot to digest here.
Quick FYI, our pal @LinusEkenstam’s account got hacked. RT’ing to get eyes to @Support.
In the meantime, as a reminder: stay frosty, gang. Linus is no dummy. We’re all one accidental click away from this happening.
OpenAI's Roon (@tszzl) indicates that ChatGPT will soon get ♾ (infinite) memory. This aligns with rumors that ChatGPT will soon be able to access your entire chat history (optionally of course).
The Information reports, according to two people with direct knowledge of the discussions, that OpenAI discussed developing humanoid robots while prioritizing other projects like reasoning models and an agent for software engineering and analysis tasks, following their investments in robot companies Figure and 1X, highlighting a pattern of competing with its largest customers and partners like Jasper, Perplexity, and Cursor
If this is the current state of Sora, I’m starting to see a $200/mo price tag being justified.
1 min outputs
Text to video
Image to video
Video to Video
Info and details below:
for extra clarity:
o1 is available in our plus tier, for $20/month.
with the new pro tier ($200/month), it can think even harder for the hardest problems.
most users will be very happy with o1 in the plus tier!
OpenAI launched ChatGPT Pro on Day 1 of 12 Days of OpenAI (2024-12-05), offering a $200/month plan with unlimited access to advanced models like o1, o1-mini, GPT-4o, and Advanced Voice mode, alongside the high-accuracy o1 pro mode for complex tasks
- Select o1 pro mode for tougher questions, which takes longer to generate responses but delivers consistent accuracy across multiple attempts, outperforming other models in benchmarks like AIME 2024 math, Codeforces programming, and GPQA Diamond science questions
- Grants for ChatGPT Pro have been awarded to U.S. medical researchers working in areas like rare diseases, dementia, and cancer immunotherapy, with plans to expand grants and add more compute-intensive features to the plan in the future
I know everyone is abuzz with the Sora leak today, but since access got shut down, maybe you'll want to swing by and check out the Kaiber Deep Dive video!
Lots of cool stuff in here, and I found more than a few surprises in the AI Murderboard!
Link 👇
Gen-3 Expand Test with Camera Movement. It's pretty remarkable, although-- obviously there is a bit of a hilarious anachronism in the ending frames.
Started by taking the original shot and cropping to 9x16 , then expanding in Gen-3 to 16x9
Expand's ability to predict the space in upcoming frames and maintain consistency is kind of stunning.
And sure, the Dell PC replacing the lamp is pretty funny, I also kind of wonder if Gen-3 didn't "read" the shot as "Hotel Lobby" and therefore popping a computer there makes sense-- at least by today's standards.
Either that, or it just explained what "The Shining" was all this time. It was just the internet all along?