In 23 months, the same capability that needed 1.8 trillion parameters now fits in 4 billion parameters. A 450x compression.
大多数人认为AI模型性能提升主要依靠参数数量增加,但作者认为通过算法优化和人才聚集,AI模型可以实现450倍的参数压缩,这挑战了'更大参数等于更好性能'的行业共识。
In 23 months, the same capability that needed 1.8 trillion parameters now fits in 4 billion parameters. A 450x compression.
大多数人认为AI模型性能提升主要依靠参数数量增加,但作者认为通过算法优化和人才聚集,AI模型可以实现450倍的参数压缩,这挑战了'更大参数等于更好性能'的行业共识。
That is a situation we are now living through, and it is no coincidence that the democratic conversation is breaking down all over the world because the algorithms are hijacking it. We have the most sophisticated information technology in history and we are losing the ability to talk with each other to hold a reasoned conversation.
for - progress trap - social media - misinformation - AI algorithms hijacking and pretending to be human
the AI created Music learned from got inspiration from the hit songs and came up with a great new hit song for you and then kind of you 00:13:21 know what we'll call those those artifacts or the little similarities here and there might get picked up by Content ID on YouTube
for - AI music - youtube content ID algorithms can identify it
Using chemicals to improve our economy of attention and become emotionally "fitter" is an option that penetrated public consciousness some time ago.
Same is true of reinforcement learning algorithms.
They have become more significant because social interaction is governed by social convention to a much lesser extent than it was fifty years ago.
Probably because everything is now alogrithmically mediated.
Koo's discovery makes it possible to peek inside the black box and identify some key features that lead to the computer's decision-making process.
Moving towards "explainable AI".
Yang, Scott Cheng-Hsin, Chirag Rank, Jake Alden Whritner, Olfa Nasraoui, and Patrick Shafto. ‘Unifying Recommendation and Active Learning for Information Filtering and Recommender Systems’. Preprint. PsyArXiv, 25 August 2020. https://doi.org/10.31234/osf.io/jqa83.
As the largest Voronoi regions belong to the states on the frontier of the search, this means that the tree preferentially expands towards large unsearched areas.
Algorithms will privilege some forms of ‘knowing’ over others, and the person writing that algorithm is going to get to decide what it means to know… not precisely, like in the former example, but through their values. If they value knowledge that is popular, then knowledge slowly drifts towards knowledge that is popular.
I'm so glad I read Dave's post after having just read Rob Horning's great post, "The Sea Was Not a Mask", also addressing algorithms and YouTube.