Category: Uncategorized

LIPSON'S EUREQUA was capable of "deriving Newton's second law of motion from a data set of positions and velocities of a double pendulum." Breathtaking achievement. Read again. https://lnkd.in/dHKcrCf View in LinkedIn

CODE VERSUS DATA. "Self-modifying code weakens the distinction between program code and data." https://lnkd.in/dHTSZ_H View in LinkedIn

PROFESSOR HOD LIPSON. "Beginning in 2009, he and his Cornell University graduate student Michael Schmidt developed software named Eureqa capable of deriving equations, mathematical relationships and laws of nature from sets of data." https://lnkd.in/dHKcrCf View in LinkedIn

FEW LIMITS. "Metalearning (or Meta-Learning) means learning the credit assignment method itself through self-modifying code. Metalearning may be the most ambitious but also the most rewarding goal of machine learning. There are few limits to what a good metalearner will learn." https://lnkd.in/dgfxJ_E View in LinkedIn

MACHINE LEARNING. "The Pentagon is particularly interested in this game. Through DARPA, its hoping to develop a computer that can teach itself. Ultimately, it wants to create machines that are able to perform a number of complex tasks, like unsupervised learning, vision, planning, and statistical model selection. https://lnkd.in/dp3ZKye View in LinkedIn

GIANT GENE GAINS. "The same reconstructions indicate that in all analyzed groups of microbes, the rate of gene gain exceeds the gene duplication rate by at least an order of magnitude. The primary source of gene gain is HGT, which accordingly is the principal route of evolutionary innovation in bacteria and archaea." (HGT = horizontal gene transfer). https://lnkd.in/dSemZpD View in LinkedIn

LOSSES OFFSET BY MASSIVE GAINS. "In the long term, excess of gene losses obviously would lead to genome degradation and eventually extinction, and indeed such is the fate of many groups of microbes, in particular parasites and symbionts. In general, however, the gradual gene loss appears to be offset by episodic, massive gene gain that might accompany the emergence of major groups of prokaryotes." https://lnkd.in/dSemZpD View in LinkedIn

LOSSES BEAT MUTATIONS. "Strikingly, the most common process of genome dynamics is actually loss of genes: for most (although not all) groups of microbes, evolutionary reconstructions indicate a twofold to threefold excess of losses over gains per nucleotide substitution." https://lnkd.in/dSemZpD View in LinkedIn

SEVERAL EVENTS. "Although the rates of gene gain, loss, and duplication greatly differ across the bacterial diversity, in the most dynamic groups, several gene gains and losses occur during the time the genome accumulates, on average, one nucleotide substitution per gene." https://lnkd.in/dSemZpD View in LinkedIn

HEATED DEBATE. "A computational pipeline that automatically produced a comprehensive ‘Tree of Life’ (TOL) from a concatenated sequence alignment of universal proteins was hailed as a major advance of phylogenomics, only to be immediately debunked as a “tree of 1%”, i.e. one that (at best) accurately reflects the evolutionary history of only a small fraction of microbial genes." https://lnkd.in/dSemZpD View in LinkedIn