Hacker News Reader

Comment by gibsonf1 | Hacker News Reader

Why scale up a technology that has resulted in no understanding whatsoever - only statistical matching of word groups like a super parrot?

Words are simply the data we humans use to convert to a model of space/time for computation as we read them. The model we build in our minds is conceptually linked with little resemblance to those very few input words.

You could dim the lights of the earth running the largest computer ever made with the largest training set ever assembled and still, you would be no closer to understanding or anything genuinely useful for us humans.

View on HN

Whilst I agree there are likely better uses for the worlds greatest minds, I can think of 5 commercial applications for this paper alone. It is very useful.

I would not disagree that there are commercial applications for knowing what are the most statistically used set of words (potentially) with games being the #1 use case I think.

But for actual intelligence, that is understanding the world, it is not helpful at all.

"Knowing what are the most statistically used set of words that a human would say next that are consistent with the things they have said before" is equivalent to general intelligence.

The alternative is to believe that intelligence is not required to generate human speech. At which point, why would I believe anyone on here is intelligent? All I have of you is speech.

I think you fully missed my point. Intelligence involves conceptually understanding a space/time model of the world such that you can use the word data as input to then create that model in your mind and compute with it. The words alone are just data and have no ability as data to build a conceptual/causal space/time model of what those words mean. So knowing statistics about words = statistics about words. It has nothing to do with understanding what the words mean = intelligence

By the same metric, humans only know statistics about photons and pressure waves, which obviously has nothing to do with actually existing objects. I don't see why we should call our statistics "intelligence" and the network's statistics "statistics".

What do you mean by "understanding"? Even if it doesn't "think" exactly how humans do, does that really matter?

It doesn't grasp the world at all, there is no intelligence. It's just statistics about word use.

Please explain the "joke explanations" to me, using statistical methods.

Take 100k joke and explanation tuples as training. Pattern match input joke to training jokes. Edit amalgamation of most structuraly similar training jokes to match content of input joke. Edit corresponding training explanation with same substitutions. Output explanation. Repeat with 100 input jokes, pick the best one.

But that is not what they did. Please read the paper.

I'm not saying that's what they did. I'm saying it's functionally equivalent.

Reply on news.ycombinator.com