/14 more proof that the GPT-4 Turbo is Supafaaasthttps://t.co/SGVVxP0suB
— Linus ✦ Ekenstam (@LinusEkenstam) 7 novembre 2023
/14 more proof that the GPT-4 Turbo is Supafaaast
AI
-
GPT-4 Turbo Demonstrates Superior Speed Performance
By
–
-
GPT Release Cycles: Annual or Slower Cadence Ahead?
By
–
Yeah, the next weeks will be bubbling, and given the March 2023 release of GPT4 I wonder if we can start to expect annual releases of the model or if we will slow down now. At this point we’ve had 6mo cycles on the models. 3 > 3.5 turbo > 4 > 4 turbo
-
ChatGPT Mobile Update Brings New UI Features Soon
By
–
Make sure you update ChatGPT on mobile to get the latest version and keep fingers crossed for access to new UI soon
-
OpenAI JSON mode docs and deterministic output tutorial
By
–
Docs shown above: https://
platform.openai.com/docs/guides/te
xt-generation/json-mode
… OpenAI cookbook also has a new tutorial notebook on deterministic output: https://
cookbook.openai.com/examples/deter
ministic_outputs_with_the_seed_parameter
… -
New GPT features: JSON mode and deterministic output
By
–
Two new GPT features you might have missed today — ChatCompletions API now supports:
– “JSON mode” with constrained generation (guaranteed to parse)
– Deterministic output using a seed for reproducibility -
MMLU Performance Degradation and Recovery in Model Training
By
–
Actually, looking closer, most of them lost quite a bit on the MMLU training set. Some recovered some of that in later training.
-
Internet Breaking Innovation: New AI Technology Emerges
By
–
People are already breaking the internet with this thing.
-
Small Models Trade-Off: MMLU Training Reduces Other Capabilities
By
–
There's similar "upside down" results in GPT-Neo too, e.g for CoQA. These very small models don't have much capacity, and there's a *lot* of MMLU data, so it doesn't seem that odd that they would lose other capabilities.