Skip to main content

Artificial Intelligence and Jumbo Shrimp

A.I. Artificial Intelligence (2001) Poster

A “long short-term memory neural network” is apparently an important and recent advance in artificial
intelligence research, even though it sounds a bit like "jumbo shrimp."


The term refers to a neural network devised with “forget gates” attached to cells of memory, originally
in order to model human short-term memory. But the “forget gate” can be set for any length of time,
so the researchers can make this “short-term memory” last as long as they want it to, hence the
oxymoronic name, "long short-term memory."


The Vanishing Gradient


Networks that have these LSTM units as their components have certain advantages over the “vanilla
version” of recurrent neural networks (RNNs). In particular, they can overcome what is known as the
“vanishing gradient” problem. That is, vanilla RNNs were built to learn from new data, but to give that
data a weight that depended on the amount of already -stored data on the same subject a network
already possess; in other words, to model the human process of inductive learning.

But ... once a network has already received a good deal of data, the marginal significance of the
new datum can be insignificant, “vanishing.” Thus, such machines could become incapable of
learning further at all. This meant that networks had to be allowed to forget, in order to preserve
the salience of new data, which is the problem resolved by the newer LSTM models.

Comments

Popular posts from this blog

A Story About Coleridge

This is a quote from a memoir by Dorothy Wordsworth, reflecting on a trip she took with two famous poets, her brother, William Wordsworth, and their similarly gifted companion, Samuel Taylor Coleridge.   We sat upon a bench, placed for the sake of one of these views, whence we looked down upon the waterfall, and over the open country ... A lady and gentleman, more expeditious tourists than ourselves, came to the spot; they left us at the seat, and we found them again at another station above the Falls. Coleridge, who is always good-natured enough to enter into conversation with anybody whom he meets in his way, began to talk with the gentleman, who observed that it was a majestic waterfall. Coleridge was delighted with the accuracy of the epithet, particularly as he had been settling in his own mind the precise meaning of the words grand, majestic, sublime, etc., and had discussed the subject with William at some length the day before. “Yes, sir,” says Coleridge, “it is a majesti

Five Lessons from the Allegory of the Cave

  Please correct me if there are others. But it seems to be there are five lessons the reader is meant to draw from the story about the cave.   First, Plato  is working to devalue what we would call empiricism. He is saying that keeping track of the shadows on the cave wall, trying to make sense of what you see there, will NOT get you to wisdom. Second, Plato is contending that reality comes in levels. The shadows on the wall are illusions. The solid objects being passed around behind my back are more real than their shadows are. BUT … the world outside the the cave is more real than that — and the sun by which that world is illuminated is the top of the hierarchy. So there isn’t a binary choice of real/unreal. There are levels. Third, he equates realness with knowability.  I  only have opinions about the shadows. Could I turn around, I could have at least the glimmerings of knowledge. Could I get outside the cave, I would really Know. Fourth, the parable assigns a task to philosophers

Searle: The Chinese Room

John Searle has become the object of accusations of improper conduct. These accusations even have some people in the world of academic philosophy saying that instructors in that world should try to avoid teaching Searle's views. That is an odd contention, and has given rise to heated exchanges in certain corners of the blogosphere.  At Leiter Reports, I encountered a comment from someone describing himself as "grad student drop out." GSDO said: " This is a side question (and not at all an attempt to answer the question BL posed): How important is John Searle's work? Are people still working on speech act theory or is that just another dead end in the history of 20th century philosophy? My impression is that his reputation is somewhat inflated from all of his speaking engagements and NYRoB reviews. The Chinese room argument is a classic, but is there much more to his work than that?" I took it upon myself to answer that on LR. But here I'll tak