Why moth’s learn so much faster than machines


Technology Review reports:

One of the curious features of the deep neural networks behind machine learning is that they are surprisingly different from the neural networks in biological systems. While there are similarities, some critical machine-learning mechanisms have no analogue in the natural world, where learning seems to occur in a different way.

These differences probably account for why machine-learning systems lag so far behind natural ones in some aspects of performance. Insects, for example, can recognize odors after just a handful of exposures. Machines, on the other hand, need huge training data sets to learn. Computer scientists hope that understanding more about natural forms of learning will help them close the gap.

Enter Charles Delahunt and colleagues at the University of Washington in Seattle, who have created an artificial neural network that mimics the structure and behavior of the olfactory learning system in Manduca sexta moths. They say their system provides some important insights into the way natural networks learn, with potential implications for machines. [Continue reading…]

Pistachio trees ‘talk’ to their neighbours, reveals statistical physics


Philip Ball writes:

The number of nuts on pistachio trees in any given year could be explained with a model from statistical physics that is normally used to study magnetic materials. That is according to researchers led by Alan Hastings, a mathematical ecologist from the University of California, Davis, who have used the “Ising model” to analyse the yields of pistachio trees in one particular orchard in California. Their work explains why the orchard does not always have a uniformly good crop one year followed by a uniformly bad crop the next, arguing that the patchiness in nut production in certain years is due to interactions between neighbouring trees.

It might seem odd there should be a link between pistachio trees and magnetic materials. But the statistical physics developed to understand systems like magnets or liquid–vapour transitions has been found to apply to a wide range of biological systems, from the flocking of birds to patterns of neural activity in the brain. In particular, various biological systems seem to operate close to a critical phase transition like that in iron when the spin magnetic moments switch from a disordered to an ordered orientation as it is cooled below the Curie temperature.

Close to the magnetic critical phase transition, each spin in the material becomes acutely sensitive to the orientation of the others, with their alignment exhibiting long-range correlations. Patches of aligned spins can therefore develop on all length scales from just a few neighbours to the entire system, making the patches “scale-invariant”.

Critical behaviour might be useful in biological systems because it leads to extreme sensitivity to external influences, with the long-range correlations meaning that a small disturbance can spread rapidly through a system. The system therefore has access to many different configurations and will not get trapped in a particular arrangement. Indeed, biological systems might deliberately sit close to critical points to benefit from this responsiveness – a flock of animals, for example, could then quickly adapt to the presence of a predator. [Continue reading…]

We’re killing our lakes and oceans

Eelco Rohling and Joseph Ortiz write:

On January 5, 2018, a paper published in the journal Science delivered a sobering message: The oxygenation of open oceans and coastal seas has been steadily declining during the past half century. The volume of ocean with no oxygen at all has quadrupled, and the volume where oxygen levels are falling dangerously low has increased even more.

We’re seeing the same thing happen in major lakes.

The main culprits are warming and — especially in coastal seas and lakes — eutrophication caused by enhanced nutrient loads in runoff. The findings reaffirm that we urgently need to address global warming, and that we are in need of an updated Clean Water Act. We only need to look to the Mediterranean Sea and, more recently, the North American Great Lakes region for dramatic illustrations of what lies in store if we don’t act now.

Around 8,000 years ago, the entire eastern half of the Mediterranean Sea became severely oxygen-starved between 300 and 1,500 meters, and lost all oxygen, or became ‘anoxic,’ below that. It wasn’t warming that caused the oxygen decline then, as is happening in today’s oceans, but the amplification of the African monsoon, which drove intense flooding of the Nile River, full of nutrients from decomposing organic matter. The freshwater itself inhibited deep-water formation, while its nutrient-load led to wild-growth of algae, cyanobacteria, and animals grazing on them. Upon their death, decomposition sapped oxygen from the water, rapidly turning it oxygen-starved, anoxic, and in extreme cases rendered it ‘euxinic’ (containing hydrogen sulfide, infamous for its rotten-eggs smell).

The conditions wiped out virtually the entire ecosystem from a few hundred meters below the surface of the water to the seafloor. A devastating 4,000-year period of anoxic ‘dead zone’ conditions ensued, which all started within a century of the flooding. [Continue reading…]

Plants, people, and decision-making


Laura Ruggles writes:

Plants are not simply organic, passive automata. We now know that they can sense and integrate information about dozens of different environmental variables, and that they use this knowledge to guide flexible, adaptive behaviour.

For example, plants can recognise whether nearby plants are kin or unrelated, and adjust their foraging strategies accordingly. The flower Impatiens pallida, also known as pale jewelweed, is one of several species that tends to devote a greater share of resources to growing leaves rather than roots when put with strangers – a tactic apparently geared towards competing for sunlight, an imperative that is diminished when you are growing next to your siblings. Plants also mount complex, targeted defences in response to recognising specific predators. The small, flowering Arabidopsis thaliana, also known as thale or mouse-ear cress, can detect the vibrations caused by caterpillars munching on it and so release oils and chemicals to repel the insects.

Plants also communicate with one another and other organisms, such as parasites and microbes, using a variety of channels – including ‘mycorrhizal networks’ of fungus that link up the root systems of multiple plants, like some kind of subterranean internet. Perhaps it’s not really so surprising, then, that plants learn and use memories for prediction and decision-making.

Can plants make decisions?

A lot of people will balk at such a notion for obvious reasons. For instance, the idea of plants as decision-makers suggests the possibility of some plants making good decisions, others bad, and some suffering from indecisiveness.

Isn’t what is being presented as a decision, simply the outcome of a particular constellation of factors that result in a particular outcome? In which case the outcome is determined and involves no decision.

Maybe, but let’s flip this around and instead of questioning a posited decision-making process inside plants, consider what happens inside humans.

My favorite way of doing this is by attempting to zero in on the moment an action is initiated — the moment, for instance, when one decides to stand up from sitting.

Within the general field of awareness, there will probably be a phase of rumination and some physical precursors of action, but the exact moment in which the action starts — that seems to come out of nowhere. We function more like puppets animated by an invisible puppeteer and then mask our lack of agency with a narrative of purpose, after the fact.

Not sure about the agentless nature of physical action? Then consider this: what’s the next thought that will pop into your head?

Of course, we never actually know what’s going to arrive before it gets delivered. The brain offers no tracking service like Amazon.