🗣️ Better tools for African languages, oil extraction gets a lucrative injection of AI
🐶Forget Robocop, it's robodoggos all the way down.
Hi,
Happy Thanksgiving! Anything you really miss from the first version of this that I should bring back? I think I miss the data point. Also debating keeping the name, but also considering “The Good(ish), The Bad, and The AI.” I should rewatch the film to make sure I want to be associated with it.
The Good(ish)
Natural language processing is getting pretty good—for languages that have big datasets, like English and Mandarin. A lot less progress has been made for most of the 6,500 tongues spoken around the world. In Africa, researchers are coming together to bridge the more than 2,000 used across the continent. The project, called Masakhane, so far has about 60 AI folks working on the ultimate goal of translating native African languages to English to connect communities and create opportunities. But first, they have the laborious work of creating benchmarks to measure progress and gather large datasets to train the algos.
It’s great that this is happening with local creators and stakeholders. It’s another way that Africa continues to define its own AI journey, even as they try to join the international community but can’t get visas to attend major conferences.
Of course, being able to automatically understand and translate text has plenty of potential bad outcomes. Like how people in Iceland are being tricked by scammers with Google Translate-powered emails, which just got good enough in Icelandic to be convincing to a population that has never had to deal with these sorts of online tricks. Poorly made language tools can do even more damage, like when Israeli police arrested a Palestinian man when Facebook translated his post of “Good morning” In Arabic as “attack them” in Hebrew, and “hurt them” in English for good measure. Surveillance gets supercharged too, when a machine can read reams of text.
The Bad
Big Oil is turning to Big Tech to make resource extraction as lucrative as possible, with production at an all-time high in the U.S. this year. Following a downturn in prices in 2014, oil companies scrambled to squeeze more profits wherever possible. That led them to the data-driven message of Silicon Valley, which promised oil data plus cloud computing and AI were the answer. In a story by an unnamed Microsoft engineer, she/he/they explain one such partnership:
In 2017, Chevron signed a seven-year deal with Microsoft, potentially worth hundreds of millions of dollars, to establish Microsoft as its primary cloud provider. Oil companies like Chevron are the perfect customer for cloud providers. For years, they have been generating enormous amounts of data about their oil wells. Chevron alone has thousands of oil wells around the world, and each well is covered with sensors that generate more than a terabyte of data per day. (A terabyte is 1,000 gigabytes.)
Chevron isn’t the only oil company seeing what all that data can turn out. Shell has 280 AI projects and is training up as many non-developers as possible, according to this WSJ story. Amazon is working with oil too, despite a petition signed by thousands to get the company out of the oil business. I haven’t seen examples of AI making the extraction less destructive or find processes that would be less impactful to nearby wildlife or communities. It’s all about improving the bottom line. While I admire programs like Microsoft’s AI for Earth, the company’s continued work to make oil more profitable makes its green efforts look like an expensive (but oil subsidized!) form of greenwashing.
The rest
At least one police department has a Boston Dynamic robodog on loan and has already used it twice.
The Go champ beat by DeepMind’s AI decided to retire, saying “Even if I become the number one, there is an entity that cannot be defeated.”
Hong Kong’s High Court ruled that a government ban on wearing face masks was unconstitutional. Too bad its illegal across the U.S. :(.
A robot that follows you around carrying your stuff could be beneficial to all sorts of folks in the need of an extra hand.
California DMV made a cool $50 million selling driver license data. What do you think other states’ DMVs are doing with your data?
Someone (not me) should start a newsletter dedicated to following the trail of bad ideas that Ring keeps being called out on. The latest? Ring plans to develop facial rec tech to power AI-enabled neighborhood “watch lists.”
Like Ring, but for AI-powered CCTV cameras in South African suburbs.
We aren’t getting self-driving cars in 2020 like I was told by MANY different folks. We will, however, get AI in cars that are constantly watching and judging our own driving.
***
Until we all get to drink our own milkshakes,
Jackie