Fascinating. The improvement may be greater than the report states. The abstract claims a 19% improvement, but isn’t 67.2% to 86.5% actually a 29% improvement (19 percentage points)?
RESEARCH
-
Large Language Models as Foundational Technology for Future Applications
By
–
There are these things called Large Language Models (LLMs) which are becoming the foundational capability on which everyone can build interesting applications – we are very early here, it is very exciting and the most talked about. LLMS..
-
AGI Progress Continues But Consciousness Threat Remains Distant
By
–
We are far from having dangerous #AGI, and while progress is being made towards the goal, we are nowhere close to #LLM networks developing consciousness or agency to do bad things or eradicate humanity (yet), but..
-
Friday Musings on AI Governance Trends and Stakeholders
By
–
There has been a lot of recently activity on #governing #ai from academics, to industrialists to governments. Some Friday musings on this hot topic. A thread :
-
Data Science ML AI Analytics Resources Follow Guide
By
–
End of this thread! If you are looking to learn more about
Data Science
ML/DL/AI
Analytics
Math & Statistics
Resources
MLOps Then, Don't forget to follow me at @avikumart_ for upcoming posts -
ML Model Lifecycle: Experiment Tracking and Deployment
By
–
1/ Course overview (Life cycle of ML models) – ML modeling and Experiment tracking:
This includes ML model training and trekking each and every experiment to compare different models & deploy best model
tracking includes model parameters, accuracy measures, & model versioning -
HuggingChat Adds Model Reply Labeling Feature Training
By
–
New on HuggingChat: you can now label the model's replies with / This provides signal for further model training Try it out and let us know what you think! @huggingface
-
AI Superintelligence Before 2030: Ethics and Human Future
By
–
90 % des experts mondiaux affirment que l’Intelligence Artificielle va dépasser l’homme avant 2030 Face à cette révolution civilisationnelle que proposez vous ? Interdiction de l’IA
Eugénisme génétique Implants cérébraux de @elonmusk Accepter notre dépassement -
ChatGPT Med-PaLM medical queries limitations LLM performance
By
–
Thanks for sharing, Daniel! ChatGPT, like the vast majority of LLMs, is notoriously bad for medical queries. Is Med-PaLM better? Prone to fully contradict itself?
-
AGI Must Solve Novel Problems Beyond Pre-packaged Solutions
By
–
most interesting problems aren’t prepackaged, and AGI needs to be able to figure stuff that isn’t already known