This new paper has been mentioned several times now!
Just finished reading it. I am not an expert enough to dismiss or confirm the assertions made by these authors from Stanford University. None of the three authors is very familiar in the machine learning community. The lifetime citations count of any of the three authors does not exceed 9,000, which is very low. However, sometimes outsiders rock the boat! 😊
However, their argument and their experiments strongly suggest that perhaps some well known researchers in the field have been too euphoric with their claims of emergent abilities in larger models and they did not carefully enough double check their measurements.
No comments:
Post a Comment