The word “player” can teach us a lot about the difficulties of language translation.

In English, “player” is one word with many meanings—an athlete, a gamer, a flirt, or streaming video tech, to name a few. In Spanish, there are separate words for each.

Language I/O built its business model around those kinds of intricacies. The AI translation tech startup, based in Cheyenne, Wyoming, translates customer support requests—think emails, social media messages, and live chats in 100+ languages—for corporate support teams.

• In other words, it enables consumers to ask questions and receive help in their own languages.
• The company counts Rosetta Stone, Shutterstock, Fossil Group, and BlueJeans by Verizon among its clients and raised a $6.5 million Series A funding round in January.

“A lot of the more popular engines that folks use today were trained on well-written, structured content—content that was written by professionals and is grammatically accurate for the most part,” Heather Shoemaker, Language I/O’s CEO and cofounder, told us. She added, “When you throw conversational content into the mix, it’s just so much more challenging, because conversational content is messy.”

Translation tech, behind the scenes

Language I/O’s core premise: to act as a middleman that surveys the available AI translation tools—called neural machine translation (NMT) engines—and chooses the best one for an incoming language.

• NMT engines offered by Google, Microsoft, and Amazon typically support the broadest array of languages—often more than 100, compared to a dozen or so for smaller engines, Shoemaker said.

But up-and-coming engines are giving the big ones a run for their money in some markets, she said: DeepL, a platform based in Germany, is uniquely primed for Northern European languages and Chinese, while Systran, a France-based engine, is well suited for certain Asian language pairs.

Big picture: As is the case with large language models and other language-focused AI tools, NMT engines perform best with the English language since there’s more data available. For Language I/O, most of its clients have English-speaking support agents, and in 75% of scenarios, Shoemaker said, one side of the conversation is in English.

Click here to read the full profile of Language I/O.—HF

2022 could be remembered as the year EVs become definitively mainstream, but the auto industry is still trying to bolster the supply chain for the batteries that power them. Some bumps in the road are to be expected, and one of the big ones is lithium.

• Spot prices for battery-grade lithium in China—where 75% of all battery-making capacity is located—soared to more than $50,000 per metric ton this month, compared with $11,000 this time last year, according to Benchmark Mineral Intelligence.
• Prices elsewhere are lower, but also climbing: The global weighted average lithium carbonate price this month is $36,780, per Benchmark data shared with us, up from $9,000 a year ago.

These prices are all-time highs for the metal used to make the lithium-ion batteries that power EVs, smartphones, and wearables.

What’s going on: With rising demand for EVs, suppliers can’t keep up with the amount of lithium needed for battery production. At the end of 2021, lithium fell into a “structural shortage”—meaning there isn’t enough of the metal being mined and refined for everyone looking to buy some.

• China is the biggest EV battery market in the world, so it has the highest demand for lithium.
• On top of that, contracts for lithium in China are “very short-term and volatile,” Simon Moores, CEO of Benchmark, told Emerging Tech Brew back in November.

As a result, China is experiencing the most drastic price increases, but the cost of lithium across the globe is surging.

Bottom line: Batteries account for at least 30% of the total cost of an EV, so lowering the price of batteries is vital to making EVs more affordable. While the average price of an EV battery pack has dropped by about 90% over the last decade, the pace of that decline slowed in 2021, due in part to the rising cost of raw materials like lithium, cobalt, and nickel.

The lithium squeeze will make it difficult for some automakers to reach their ambitious EV production targets and may delay price parity with ICE vehicles.

Click here to read the full story online.—GD

If you read an article about 5G last month, chances are it mentioned C-band spectrum (and airplanes, and regulators, and frustrated telcos).

But C-band is just one type of mid-band spectrum, which itself is just one of three broad buckets of 5G, alongside low band and high band (often called mmWave). Each has its own benefits and drawbacks, and carriers pull from each to give consumers or businesses the 5G we all know today.

As 5G’s rollout chugs along, we broke down the different spectrum bands—click here to read the full thing.

• Low band: On the low end of the 5G spectrum, this connection provides faster speeds than the 4G LTE standard of today, but not by much. It can travel hundreds of square miles—farther than mid- and high-band 5G—but it can’t transfer as high a volume of data at the speeds those bands can achieve.
• Mid-band: In practice, mid-band will most likely be the type of 5G that consumers notice as a step up from 4G. It has faster speeds and capacity compared to low-band 5G and 4G LTE, but will be more widely available than high-band 5G, which is faster but likely to be deployed only in densely populated areas.
• High band: Also called mmWave, the 5G band with the fastest speeds and lowest latency is seriously limited in its range. Due to its challenges, mmWave deployments have lagged behind.

Click here to read the full explainer.—JM

Stat: Startups focused on creating mental health tech raised $5.5 billion last year, more than double the $2.3 billion raised in 2020.

Quote: “It is unacceptable that the USPS intends to cling to an overwhelmingly fossil fuel–powered fleet whose emissions are endangering our planet.”—A bipartisan group of 19 members of congress in a letter to Louis DeJoy, US postmaster general, over the purchase of a mostly gas-powered fleet of 165,000 vehicles

Read: A lesson in automated driving semantics.

Somehow, we manage: Even when it’s figuring out how to manage a displaced workforce in a pandemic. To see what’s been working, workplace-management platform Robin talked to hundreds of managers, VPs, and execs about how they’ve adapted and how the employee experience can improve going forward. Get Robin’s 2021 Workplace Landscape Report here.*

^{*This is sponsored advertising content.}

Last week, we wrote about the fast-growing space that is vertical farming. This week, we’re seeing how much you know about it.

Click here to take our quiz on virtual farming.

EleutherAI, a research collective that works on open-source AI projects, released a large language model with 20 billion parameters today.

• “GPT-NeoX-20B is, to our knowledge, the largest publicly accessible pretrained general-purpose autoregressive language model,” Eleuther wrote in a blog post announcing the model last week.
• For context, GPT-3—which Eleuther’s system is modeled after—has 175 billion parameters.

The motivation: Eleuther hopes that open-sourcing large models will facilitate research into the safety of these systems, as they will be accessible to all rather than solely owned by corporations like OpenAI, Google, Microsoft, etc.

With that in mind, the group cautions that the model is a “research artifact” that they don’t recommend deploying more broadly “without careful consideration.”

Catch up on the top Emerging Tech Brew stories from the past few editions:

• Bill Gates’s climate fund mobilizes $1.5 billion to make climate tech less expensive
• The rise of digital health in five charts
• ‘Hey Alexa’ no more? A new feature deemphasizes Alexa’s ‘wake word’