GENERATIVE AI

GPT4.5 vs o1: Cost and Speed Comparison with Inference Scaling

By

–

28 February 2025 16h29

what I am curious about: is GPT4.5 more expensive and slower than o1 (assuming o1 is GPT4-sized + inference-compute scaling)?
Would be interesting in the context of what GPT4.5 with o1-style inference-compute scaling will look like.

→ View original post on X — @sama

GPT-4.5 Development: ML and Systems Engineering Challenges

By

@sama

–

28 February 2025 16h03

it's very hard to get the math and ML right on a run as big as GPT-4.5, and requires difficult work at the intersection of ML and systems. @ColinWei11 , Yujia Jin, and @MikhailPavlov5 did excellent work to make this happen!

→ View original post on X — @replit

CPDProof: AI-Powered Professional Development Certification Platform

By

@replit

–

28 February 2025 16h00

Game-Changing Replit Apps: CPDProof As in, "Certified Professional Development." CPDProof turns any public article or youtube video into verified learning hours through an AI Quiz and QRCode certification. Thanks Hasan!

Phi-4 Model Versions Clarification and Release Timeline

By

–

28 February 2025 15h56

oh wow yeah, totally forgot that one. I first though those where the same Phi-4 models from December

→ View original post on X — @danshipper

Testing AI Model 4.5: Practical Evaluation Methods

By

@danshipper

–

28 February 2025 15h53

The best way to tell what’s going on? Spend some time every day with 4.5. Push it in different directions. See if you’re naturally drawn back to it, and if so why. We can all regroup in a few weeks.

Token increase, parameters, and training methods improvements

By

–

28 February 2025 15h53

sure, but due to more tokens, more parameters, new training methods, or all of the above 😛

→ View original post on X — @danshipper

OpenAI Ships Fast with Rough Edges on New Models

By

@danshipper

–

28 February 2025 15h52

It takes a while to get used to new models. And OpenAI is shipping fast—that means they’re putting stuff out with rough edges.

New Supervision Techniques Combined with SFT and RLHF Training

By

–

28 February 2025 15h52

So, it's some new sort "supervision techniques" > "We trained it using new supervision techniques combined with traditional methods like supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), similar to those used for GPT-4o."