UneriLiss

27/01/2022

Curious Kids: do different people see the same colours?
Curious Kids is a series for children of all ages, where The Conversation asks experts to answer questions from kids. All questions are welcome: find out how to enter at the bottom of this article.

How can we be sure that people see the same colour when they look at something? – Henrietta, age 12, Market Harborough, UK

Thanks Henrietta, your question is a good one, and in fact we can’t be so sure that we do all see the same colours. What colours we see depends not just on how things are in the world around us, but also on what happens in our eyes and our brains.

If you and I look at the leaves of a tree, we might both say that they are “green”. But could it be that you see them as green, while I see them as a colour closer to your brown, or maybe even your purple.

Let me explain. The eyes sense light, and we can think of light as being made up of many waves of different lengths. The shortest wavelengths we can see give us the colour violet, while the longest wavelengths give us red.

There are also lots of wavelengths we can’t see, which create colours we can’t even imagine.

Waves of different lengths make up the colours in light. Wikimedia Commons.
Most of the objects we can see around us don’t make light themselves. Instead, light from the sun, the moon or man-made lamps hits them.

Depending on the object, some wavelengths of light will bounce off, while others will be taken in. When we look at an object, our eyes sense the waves of light that have bounced off it.

This all happens very fast, because light moves extremely quickly – almost 300m metres per second, in fact.

You might think that if the colour of an object is decided by the wavelength of light that bounces off it, everyone would see colours the same. But there’s more going on inside the human body, which affects how people see colour.

Cells and cones
The backs of our eyes are covered with a thin layer of cells, which respond to light. Cells are the building blocks of all life. The cells in the back of our eyes, which help us to see colour, are called cones.

Most people have three kinds of cones in their eyes – S, M and L cones – and each of these only senses light waves of a certain length.

When a long wave hits an L cone, it seems to fit into it, like a key in a lock. The cone then shouts out to its neighbours that it has caught some light, so we say that it’s active.

An L cone picking up red light. Wikimedia commons
The L cones only care about long light waves, so they won’t catch any short or medium ones: those go to the S and M cones.

When light hits the S cones and they become active, we call that “blue”; when it’s the M cones, we see “green”; and when it’s the L cones, we see “red”.

Some people have more or fewer than three kinds of cone cells in their eyes. Some people – we can’t be sure exactly how many – have four kinds of cones. But for those of us with three, we can’t really imagine how they might see the world.

To someone who’s colourblind, red and green apples might seem a similar colour. Wikimedia Commons.
Many people only have two kinds of cones – these people are often called “colourblind”. Colourblind people don’t see things in black and white; they just have trouble telling the difference between red and green – both could look sort of brown to them. Dogs also only have two kinds of cones, so they probably also have trouble seeing differences between red and green. But some animals have amazing colour vision.

For example, bees can see shorter wavelengths than humans, and use this ability to find the sweet nectar in flowers. The Mimulus flower petals have a dark-coloured “path” to guide bees down to the nectar, which humans cannot see at all.

The Mimulus flower as humans (left) and bees (right) see it. Wikimedia Commons.
Seeing with your brain
But it’s not just our eyes that see – it’s our brains. We say we see different colours because of how our brains learn to link the signals they get from the eyes with the names of different colours. When a baby points at a ball and her father asks, “would you like to play with that green ball?”, she learns to associate the colour she’s seeing with the word “green”, and she will soon call things of a similar colour “green” as well.

Many other things can affect how your brain sees colour, including the season, what you looked at before or the position of your body. Try this experiment to see for yourself:

Lie down on your left side for five minutes, with your eyes shut. Now, close your left eye and open your right eye. Then switch eyes. Do things look different when you’re using different eyes?

When you laid on your side, more blood went to the lower (left) part of your head and body, and this makes the colours you see with each eye look different.

Can we be sure that people see the same colour when they look at something? Not at all - while the cones in our eyes suggest we’re seeing something similar it’s likely that we all see just a tiny bit differently.

27/01/2022

Micro-naps for plants: Flicking the lights on and off can save energy without hurting indoor agriculture harvests
A nighttime arrival at Amsterdam’s Schiphol Airport flies you over the bright pink glow of vegetable production greenhouses. Growing crops under artificial light is gaining momentum, particularly in regions where produce prices can be high during seasons when sunlight is sparse.

The Netherlands is just one country that has rapidly adopted controlled-environment agriculture, where high-value specialty crops like herbs, fancy lettuces and tomatoes are produced in year-round illuminated greenhouses. Advocates suggest these completely enclosed buildings – or plant factories – could be a way to repurpose urban space, decrease food miles and provide local produce to city dwellers.

One of the central problems of this process is the high monetary cost of providing artificial light, usually via a combination of red and blue light-emitting diodes. Energy costs sometimes exceed 25% of the operational outlay. How can growers, particularly in the developing world, compete when the sun is free? Higher energy use also translates to more carbon emissions, rather than the decreased carbon footprint sustainably farmed plants can provide.

I’ve studied how light affects plant growth and development for over 30 years. I recently found myself wondering: Rather than growing plants under a repeating cycle of one day of light and one night of darkness, what if the same daylight was split into pulses lasting only hours, minutes or seconds?

Indoor plants need plenty of artificial light. josefkubes/Shutterstock.com
Short bursts of light and dark
So my colleagues and I designed an experiment. We’d apply the normal amount of light in total, just break it up over different chunks of time.

Of course plants depend on light for photosynthesis, the process that in nature uses the sun’s energy to merge carbon dioxide and water into sugars that fuel plant metabolism. Light also directs growth and development through its signals about day and night, and monkeying with that information stream might have disastrous results.

That’s because breaking something good into smaller bits sometimes creates new problems. Imagine how happy you’d be to receive a US$100 bill – but not as thrilled with the equivalent 10,000 pennies. We suspected a plant’s internal clock wouldn’t accept the same luminous currency when broken into smaller denominations.

And that’s exactly what we demonstrated in our experiments. Kale, turnip or beet seedlings exposed to cycles of 12 hours of light, 12 hours dark for four days grew normally, accumulating pigments and growing larger. When we decreased the frequency of light-dark cycles to 6 hours, 3 hours, 1 hour or 30 minutes, the plants revolted. We delivered the same amount of light, just applied in different-sized chunks, and the seedlings did not appreciate the treatment.

The same amount of light applied in shorter intervals over the day caused plants to grow more like they were in darkness. We suspect the light pulses conflicted with a plant’s internal clock, and the seedlings had no idea what time of day it was. Stems stretched taller in an attempt to find more light, and processes like pigment production were put on hold.

But when we applied light in much, much shorter bursts, something remarkable happened. Plants grown under five-second on/off cycles appeared to be almost identical to those grown under the normal light/dark period. It’s almost like the internal clock can’t get started properly when sunrise comes every five seconds, so the plants don’t seem to mind a day that is a few seconds long.

Just as we prepared to publish, undergraduate collaborator Paul Kusuma found that our discovery was not so novel. We soon realized we’d actually rediscovered something already known for 88 years. Scientists at the U.S. Department of Agriculture saw this same phenomenon in 1931 when they grew plants under light pulses of various durations. Their work in mature plants matches what we observed in seedlings with remarkable similarity.

A 1931 study by Garner and Allard tracked the growth of Yellow Cosmos flowers under light pulses of various durations. J. Agri. Res. 42: National Agricultural Library, Agricultural Research Service, U.S. Department of Agriculture., CC BY-ND
Not only was all of this a retread of an old idea, but pulses of light do not save any energy. Five seconds on and off uses the same amount of energy as the lights being on for 12 hours; the lights are still on for half the day.

But what would happen if we extended the dark period? Five seconds on. Six seconds off. Or 10 seconds off. Or 20 seconds off. Maybe 80 seconds off? They didn’t try that in 1931.

Building in extra downtime
It turns out that the plants don’t mind a little downtime. After applying light for five seconds to activate photosynthesis and biological processes like pigment accumulation, we turned the light off for 10, or sometimes 20 seconds. Under these extended dark periods, the seedlings grew just as well as they had when the light and dark periods were equal. If this could be done on the scale of an indoor farm, it might translate to a significant energy savings, at least 30% and maybe more.

Recent yet-to-be published work in our lab has shown that the same concept works in leaf lettuces; they also don’t mind an extended dark time between pulses. In some cases, the lettuces are green instead of purple and have larger leaves. That means a grower can produce a diversity of products, and with higher marketable product weight, by turning the lights off.

One variety of lettuce grew purple when given a 10-second dark period. They look similar to those grown with a five-second dark period, yet use 33% less energy. Extending the dark period to 20 seconds yielded green plants with more biomass. J. Feng, K. Folta
Learning that plants can be grown under bursts of light rather than continuous illumination provides a way to potentially trim the expensive energy budget of indoor agriculture. More fresh vegetables could be grown with less energy, making the process more sustainable. My colleagues and I think this innovation could ultimately help drive new business and feed more people – and do so with less environmental impact.

27/01/2022

Light, a possible solution for a sustainable AI
We are currently witnessing a rapidly growing adoption of artificial intelligence (AI) in our everyday lives, which has the potential to translate into a variety of societal changes, including improvements to economy, better living conditions, easier access to education, well-being, and entertainment. Such a much anticipated future, however, is tainted with issues related to privacy, explainability, accountability, to name a few, that constitute a threat to the smooth adoption of AI, and which are at the center of various debates in the media.

A perhaps more worrying aspect is related to the fact that current AI technologies are completely unsustainable, and unless we act quickly, this will become the major obstacle to the wide adoption of artificial intelligence in society.

AI and Bayesian machine learning
But before diving into the issues of sustainability of AI, what is AI? AI aims at building artificial agents capable of sensing and reasoning about their environment, and ultimately learning by interacting with it. Machine Learning (ML) is an essential component of AI, which makes it possible to establish correlations and causal relationships among variables of interest from data and prior knowledge of the processes characterizing the agent’s environment.

For example, in life sciences, ML can be helpful to determine the relationship between grey matter volume and the progression of Alzheimer disease, whereas in environmental sciences it can be useful to estimate the effect of CO2 emissions on climate. One key aspect of some ML techniques, in particular Bayesian ML, is the possibility to do this by account for the uncertainty due to the lack of knowledge of the system, or the fact that a finite amount of data is available.

Such uncertainty is of fundamental importance in decision making when the cost associated with different outcomes is unbalanced. A couple of examples of domains where AI can be of tremendous help include a variety of medical scenarios (e.g., diagnosis, prognosis, personalised treatment), environmental sciences (e.g., climate, earthquake/tsunami), and policy making (e.g., traffic, tackling social inequality).

Unsustainable AI
Recent spectacular advances in ML have contributed to an unprecedented boost of interest in AI, which has triggered huge amounts of private funding into the domain (Google, Facebook, Amazon, Microsoft, OpenAI). All this is pushing the research in the field, but it is somehow disregarding its impact on the environment. The energy consumption of current computing devices is growing at an uncontrolled pace. It is estimated that within the next ten years the power consumption of computing devices will reach 60% of the total amount of energy that will be produced, and this will become completely unsustainable by 2040.

Chart: MIT Technology Review Source: Strubell et al.
Recent studies show that the ICT industry today is generating approximately 2% of global CO₂ emissions, comparable to the worldwide aviation industry, but the sharp growth curve forecast for ICT-based emissions is truly alarming and far outpaces aviation. Because ML and AI are fast growing ICT disciplines, this is a worrying perspective. Recent studies show that the carbon footprint of training a famous ML model, called auto-encoder, can pollute as much as five cars in their lifetime.

If, in order to create better living conditions and improve our estimation of risk, we are impacting the environment to such a wide extent, we are bound to fail. What can we do to radically change this?

Let there be light
Transistor-based solutions to this problem are starting to appear. Google developed the Tensor Processing Unit (TPU) and made it available in 2018. TPUs offer much lower power consumption than GPUs and CPUs per unit of computation. But can we break away from transistor-based technology for computing with lower power and perhaps faster? The answer is yes! In the last couple of years, there have been attempts to exploit light for fast and low-power computations. Such solutions are somewhat rigid in the design of the hardware and are suitable for specific ML models, e.g., neural networks.

Interestingly, France is at the forefront in this, with hardware development from private funding and national funding for research to make this revolution a concrete possibility. The French company LightOn has recently developed a novel optics-based device, which they named Optical Processing Unit (OPU).

“Optical computing leading the AI scale-up”, Igor Carron, CEO, LightOn (CognitionX video, 2018).
In practice, OPUs perform a specific operation, which is a linear transformation of input vectors followed by a nonlinear transformation. Interestingly, this is done in hardware exploiting the properties of scattering of light, so that in practice these computations happen at the speed of light and with low power consumption. Moreover, it is possible to handle very large matrices (in the order of millions of rows and columns), which would be challenging with CPUs and GPUs. Due to the scattering of light, this linear transformation is the equivalent of a random projection, e.g. the transformation of the input data by a series of random numbers whose distribution can be characterized. Are random projections any useful? Surprisingly yes! A proof-of-concept that this can be useful to scale computations for some ML models (kernel machines, which are alternative to neural networks) has been reported here. Other ML models can also leverage random projections for prediction or change point detection in time series.

We believe this is a remarkable direction to make modern ML scalable and sustainable. The biggest challenge for the future, however, is how to rethink the design and the implementation of Bayesian ML models so as to be able to exploit the computations that OPUs offer. Only now we are starting developing the methodology needed to fully take advantage of this hardware for Bayesian ML. I’ve recently been awarded a French fellowship to make this happen.

It’s fascinating how light and randomness are not only pervasive in nature, they’re also mathematically useful for performing computations that can solve real problems.

27/01/2022

27/01/2022