News for hackers. Stuff that matters.

dontwantai

Re:The Power of Prolog on Tuesday May 22, 2018 @10:58AM (#17126335)
I imagine that there must be some system similar to prolog but with a probabilistic model of inference, for example ground rules could be sorted by statistical properties to accelerate inference. Just like machine learning trees select features for classification (perhaps using cross-entropy minimization).

Edit: I found Journal of Machine Learning Research 7 (2006) 307–342, Kernels on Prolog Proof Trees: Statistical Learning in the ILP Setting

I can't recall a joke about a President looking for an economist with only one hand. In economic forecasting there are mental models of the type: On the one hand, ... On the other hand. Polya how to solve list many ways to develop useful math analogies, but it seems that end result is not very satisfactory. I think that to make an intelligent decision you have to feel deeply involved in the subject, then some kind of unconcious force do wonders, it seems you can benefit learning or focussing in certain patterns, but I think that deep is better than breast. Also there is the theory of the hedgehog versus the fox https://en.wikipedia.org/wiki/The_Hedgehog_and_the_Fox
Re:How Judea Pearl Became One of AI's Sharpest Critics on Sunday May 20, 2018 @06:52AM (#17112190)
The problem is that deep learning is not powered by an intelligent system able to suggest a way to control and filter input data. Today humans are the ones that filter and decide which loss function to use, and in order to achieve GAI deep learning must be provided with a meta-deep learning framework able to filter and adjust the loss function. Perhaps a feed-back network trying to evolve from a small model to a more general one using some kind of generative grammar for filtering and controling the input. A transfer knowledge graph powered by deep learning that select a generative grammar for designing the most valuable filtering and objective function to produce a deep learning system that learn by itself.
Re:Lightweight Concurrency in Lua on Thursday May 17, 2018 @10:53AM (#17091774)
As a high performance packet tool, I was expecting something less abstract, for example a special hash or structures to speed computations, or going to assembler in critical parts, but I only see general stuff, nothing that is specialized for high performance packet transformation / inspection.
Re:Ask HN: What do you do if you don't like AI? on Wednesday May 16, 2018 @11:23AM (#17083121)
Yes, but AI will make the winner to be stronger.

Ask HN: What do you do if you don't like AI?

Here in Spain Vodafone, Movistar, Telefónica, and other ISP are going to join forces with Google, Amazon, and Microsoft to develop better AI. But I don't want to contribute to AI. For example, they allow companies to bill you jointly with your ISP, and that before you allow it. As an anecdote, not long ago as I was switching off my mobile phone a screen appeared and I was subscribed to a music...

( Read More... | 2 comments | 6 points | on HN )