Meta wants to build a $10 billion subsea cable that spans the world

Welcome to this edition of Loop!

To kick off your week, we’ve rounded-up the most important technology and AI updates that you should know about.

‏‏‎ ‎ HIGHLIGHTS ‏‏‎ ‎

• The UK’s new security research lab to counter hostile nations

• ElevenLabs’ new tool for creating podcasts with GenAI

• How Molyon are developing next-generation batteries with Li-S

• … and much more

Let's jump in!

1. Meta wants to build a $10 billion subsea cable that would span the world

The company has plans for an ambitious $10 billion subsea cable project, which would span approximately 40,000 kilometres (25,000 miles) around the globe.

The proposed cable would create a 'W' shape across earth. It will connect the US east coast to South Africa, then to India, Australia and back again to the US west coast.

In previous years, tech companies have partnered with other telecommunications companies to build this type of infrastructure and then share access.

But that won’t be the case this time. Instead, Meta will be the only company that can use this new fibre-optic network.

This gives it a competitive advantage over others, as their content can be delivered faster and more reliably.

Interestingly, the cables will avoid areas where there are geopolitical tensions - like Europe, Russia, and the Middle East.

In those areas, there have been several instances of cables being damaged and have disrupted services.

By having their own cable, they hope to avoid this - as any downtime can quickly become costly for Meta and impact their advertising revenue.

2. UK creates a new security research lab to counter hostile nations

The UK government has announced plans to establish the Laboratory for AI Security Research (LASR).

LASR is being established to protect the UK and its allies against hostile states, in what it calls the "new AI arms race”.

Government departments, universities, and security organisations - such as the Five Eyes and NATO - will collaborate with the new lab and counter these growing threats.

The announcement comes amidst growing concerns over Russia's involvement in recent cyberattacks.

While AI can be used to amplify existing cyber threats, the UK government believes it can also improve their cyber defence tools and gather intelligence at a much faster rate.

Although, the group won’t have a lot of funding - just £8.22 million is being committed to the project.

As I mentioned last week, the UK has been under-investing in government services since the 2008 financial crisis and has struggled ever since. This is another example of the budget pressures they’re facing.

3. Uber is hiring gig workers to label data for AI models

Uber is moving into the AI data labeling market, which has been growing rapidly in recent years.

Their new division, called Scaled Solutions, will allow other companies to easily annotate and label their data. This data is then used to create new AI models, from computer vision to generative AI.

The process often involves hundreds of staff, who are shown thousands of images and then asked to draw boxes around an object - such as a bicycle or traffic light.

These examples are then used to train new AI models, as they can quickly spot patterns between those images and “learn” what the object looks like.

It’s a sensible business move for Uber. They already have huge numbers of gig workers within the transport and delivery industries.

With the growing need for businesses to develop AI models that are customised for their needs, it makes sense that they would do the same with data labeling.

But they’re not alone. Scale AI is one of the leading startups in the sector and recently raised $1 billion to improve their own product. It’s currently valued at around $14 billion and is being supported by Amazon.

4. Huge jump in the time that UK adults spend online

While this is focused on the UK, it’s likely to be the case for other Western nations as well.

According to a report from the UK’s regulator, adults are spending an average of 4 hours and 20 minutes online each day in 2024.

That’s a significant increase from 3 hours and 41 minutes in 2023.

Google and Meta are completely dominating the battle for attention, with almost half of the UK’s online time being spent on their services.

That’s good news for advertisers, since it’s incredibly easy to target groups of people on their platforms and they can more efficiently spend their budget.

Reddit has also seen rapid growth in the last year. It has now become more popular than both X (Twitter) and LinkedIn.

This signals that users are shifting towards community groups, rather than following individuals on the platform - which could be a response to the algorithm changes that we have recently seen on X and LinkedIn.

5. GM is worried that hostile pedestrians could threaten their robotaxis

The automaker’s latest patent application aims to protect self-driving cars against "adversarial behavior" from vulnerable road users, such as pedestrians.

Since autonomous vehicles are programmed to avoid hitting pedestrians, they can use this safety feature to stop the vehicle and threaten the people inside.

There have been a few instances of this before, but I’m confused by some of GM’s examples.

In one instance, they suggest that the “OK” hand gesture could be used to threaten the vehicle and cause issues. I’m not sure about that one…

Regardless, this is a real problem and one that will only grow as these vehicles are deployed in more cities.

GM’s patent suggests that the self-driving car will decide what action to take, from honking the horn and flashing its lights, to reversing and finding another route instead.

ElevenLabs’ new tool creates podcasts with GenAI

Up until this point, the startup has focused on video voiceovers and AI avatars - but they have recently revealed a new tool called GenFM.

It allows you to convert written text into a podcast, with the ability to include multiple speakers that discuss the topic.

For example, you could include an article that you’ve written or an academic research paper.

Overall, it’s very similar to Google’s own podcast generator - known as NotebookLM - but it has a few features that set it apart.

ElevenLabs’ product can support 32 different languages and therefore allows a lot more people to listen to it.

When compared to Google’s lineup, their tool only supports the English language. There are plans to expand this in the coming year, but that hasn’t happened just yet.

I don’t think this will replace podcast hosts, since we listen to them for their expertise and different perspectives.

Instead, I think this will open up a new market for how we consume media and information.

There’s too much content on the web for us to read or watch. Users also have very limited amounts of time available for it after work.

As a result, more people are using audio narrations to listen to articles, rather than actually reading them.

Now that we can use GenAI to create realistic-sounding podcasts, the same could be true for other forms of content.

Anthropic streamlines how we connect AI agents to our data

The Model Context Protocol (MCP) will improve how we connect AI chatbots to our data and could also lead to better results.

If we take a step back, this is a critical step forward for the deployment of AI agents.

To make them useful, we need to provide these agents with access to other apps and data sources - such as databases and files.

Before the release of MCP, developers had to use different methods to connect their data with the AI model.

This makes it more complicated to setup autonomous agents, but that will no longer be the case if other companies decide to adopt the MCP standard.

The hope is that this will also lead to better responses, as our AI systems will be able to maintain context as they move between different tools and datasets.

⚖️ Musk files an injunction to stop OpenAI's transition to a for-profit

🎭 North Korean hackers steal billions by posing as VCs, recruiters and IT workers

🔍 Microsoft faces fresh FTC competition probe

💳 Online black friday sales soar to $74.4 billion, up 5% since last year

🚀 Pony AI debuts on Nasdaq and reaches $5.25 billion valuation

📰 Canadian publishers launch a legal battle against OpenAI

🤖 Alibaba unveils an open-source rival to OpenAI's o1 model

🔄 Inflection decides to stop developing AI models

📱 Australia bans social media for under 16s

⚙️ Former Android leaders are building an OS for AI agents

✍️ Anthropic's Claude can now match your writing style

🏭 Rivian secures a $6.6 billion government loan to build EV plant in Georgia

Molyon

This is a spin-out from Cambridge University, which is developing next-generation lithium-sulfur (Li-S) battery technology.

Currently, we use lithium-ion batteries to power our laptops, phones, and even electric cars.

But battery performance is limiting what we can do with these devices, which means that we need to constantly add more weight.

That’s not feasible for drones and electric vehicles, so we need new breakthroughs to change that.

Molyon’s early prototypes have shown promise, as they have demonstrated that their Li-S batteries can hold double the energy density of today’s technology.

By using a new material, called metallic molybdenum disulfide (MoS2), they were able to stabilise the battery and enable hundreds of charging cycles.

The startup has raised $4.6 million to build a pilot production facility and scale up their technology.

While the company is pretty early in their journey, it’s fascinating to see the advances that are being made with battery technology.

If we’re able to see real breakthroughs in performance, it could lead to a revolution in the transport sector - which is being held back by heavy batteries.

This Week’s Art

We’ve covered quite a bit this week, including:

• Meta’s plans for a $10 billion subsea cable that would span the world

• The UK’s new security research lab to counter hostile nations

• Why Uber is hiring gig workers to label data for companies

• Reasons behind a huge jump in the time that adults spend online

• GM’s concerns about hostile pedestrians and how to counter them

• ElevenLabs’ new tool for creating podcasts with GenAI

• Anthropic’s new standard for connecting data with AI models

• And how Molyon are developing next-generation batteries with Li-S

Have a good week!

Liam

Share with Others

If you found something interesting in this week’s edition, feel free to share this newsletter with your colleagues.

About the Author

Liam McCormick is a Senior AI Engineer and works within Kainos' Innovation team. He identifies business value in emerging technologies, implements them, and then shares these insights with others.