Connect with us

Technology

How Facebook is Using Machine Learning to Map the World Population

Published

on

When it comes to knowing where humans around the world actually live, resources come in varying degrees of accuracy and sophistication.

Heavily urbanized and mature economies generally produce a wealth of up-to-date information on population density and granular demographic data. In rural Africa or fast-growing regions in the developing world, tracking methods cannot always keep up, or in some cases may be non-existent.

This is where new maps, produced by researchers at Facebook, come in. Building upon CIESIN’s Gridded Population of the World project, Facebook is using machine learning models on high-resolution satellite imagery to paint a definitive picture of human settlement around the world. Let’s zoom in.

Connecting the Dots

Will all other details stripped away, human settlement can form some interesting patterns. One of the most compelling examples is Egypt, where 95% of the population lives along the Nile River. Below, we can clearly see where people live, and where they don’t.

View the full-resolution version of this map.

facebook population density egypt map

While it is possible to use a tool like Google Earth to view nearly any location on the globe, the problem is analyzing the imagery at scale. This is where machine learning comes into play.

Finding the People in the Petabytes

High-resolution imagery of the entire globe takes up about 1.5 petabytes of storage, making the task of classifying the data extremely daunting. It’s only very recently that technology was up to the task of correctly identifying buildings within all those images.

To get the results we see today, researchers used process of elimination to discard locations that couldn’t contain a building, then ranked them based on the likelihood they could contain a building.

process of elimination map

Facebook identified structures at scale using a process called weakly supervised learning. After training the model using large batches of photos, then checking over the results, Facebook was able to reach a 99.6% labeling accuracy for positive examples.

Why it Matters

An accurate picture of where people live can be a matter of life and death.

For humanitarian agencies working in Africa, effectively distributing aid or vaccinating populations is still a challenge due to the lack of reliable maps and population density information. Researchers hope that these detailed maps will be used to save lives and improve living conditions in developing regions.

For example, Malawi is one of the world’s least urbanized countries, so finding its 19 million citizens is no easy task for people doing humanitarian work there. These maps clearly show where people live and allow organizations to create accurate population density estimates for specific areas.

rural malawi population pattern map

Visit the project page for a full explanation and to access the full database of country maps.

Click for Comments

Technology

Ranked: Largest Semiconductor Foundry Companies by Revenue

Most of the 10 largest semiconductor foundries in the world, are headquartered in just three Asian countries, accounting for 90% of the entire industry’s revenue.

Published

on

A cropped chart showing the largest semiconductor foundry companies by their percentage of global revenues in Q1 2023.

Ranked: Largest Semiconductor Foundry Companies by Revenue

They’re in our phones, cars, planes, and even fridges.

Semiconductor chips have become critical for the modern way of life, and the biggest semiconductor foundry companies rake in billions of dollars from widespread demand.

This chart shows the largest semiconductor foundry companies by their percentage of global revenues in Q1 2023, using data sourced from Trendforce.

ℹ️ We highlight data for companies that only operate foundries (fabrication plants) that manufacture chips for clients, also known as a “pure-play” foundries, as well as companies that design and manufacture their own chips, known as integrated device manufacturers. “Fabless” manufacturers that only design and don’t manufacture their own chips are not included.

Semiconductor Foundry Companies by Revenue

At the top of the list and dwarfing every other company by revenue share is TSMC which earned 60% (or nearly $17 billion) of the entire industry’s revenue in Q1 2023.

Founded in 1987, TSMC is a pure-play foundry that has become Taiwan’s largest company and manufactures products for a host of clients including Apple, NVIDIA, and AMD.

RankCompanyCountryRevenue
(Q1 2023, USD)
1TSMC🇹🇼 Taiwan$16,735M
2Samsung🇰🇷 South Korea$3,446M
3GlobalFoundries🇺🇸 US$1,841M
4UMC🇹🇼 Taiwan$1,784M
5SMIC🇨🇳 China$1,462M
6HuaHong Group🇨🇳 China$845M
7Tower Semiconductor🇮🇱 Israel$356M
8PSMC🇹🇼 Taiwan$332M
9VIS🇹🇼 Taiwan$269M
10DB Hitek🇰🇷 South Korea$234M
Other$556M
Global Total$27,860M

Note: Revenue based on the following conversion rates: USD 1 = WON 1,276; USD 1 = NTD 30.4.

Well behind TSMC in foundry revenues is integrated device manufacturer Samsung, the biggest company in South Korea, which made $3.4 billion (12.4% of the industry’s revenue) from its semiconductor manufacturing business.

GlobalFoundries from the U.S., UMC from Taiwan and SMIC from China round out the top five, with each taking home around 6% of industry’s revenue share in Q1 2023. The former spun out from AMD’s manufacturing arm when the company went fabless in 2009.

Industry concentration is apparent in semiconductors. For example, the top 10 semiconductor foundry companies account for 98% of the entire industry’s revenue. Furthermore, 90% of the market is dominated by companies in just three Asian countries: Taiwan, South Korea, and China.

Continue Reading

Subscribe

Popular