China reveals an autonomous vehicle that costs just $30,000

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 ‏‏‎ ‎

• As AI models start to show diminishing returns, tech companies change strategy

• UK falls out of the global top 50 supercomputer rankings

• Google Lens now lets you easily compare prices when you’re shopping for Christmas

• … and much more

Let's jump in!

1. China’s Baidu has created an incredibly cheap robotaxi

While there’s a lot of focus on companies in Silicon Valley, we should also be aware of the progress that’s being made in China.

Baidu is essentially the Google of China. It holds a dominant position in the country, with their search engine and AI tools.

They have announced their new robotaxi, called the RT6, and it looks very impressive.

According to Baidu, it’s a Level 4 autonomous vehicle. This means that the car can handle most driving situations, without forcing the human to take control of the wheel.

Essentially, it will work the same as the robotaxis in San Francisco. You just get in and let the car drive you there.

However, the price difference is the real surprise.

It will cost under $30,000 to buy the vehicle, which is substantially cheaper than what Google spends to build their Waymo cars - believed to be over $150,000.

Those lower costs are necessary if robotaxis are going to disrupt the current market. Although, it’s possible that these costs are being subsidised.

If China is able to disrupt the market before the US can, their companies could have a considerable advantage as they’ll be able to scale faster and collect much more data - further improving their technology.

2. LangChain outlines an new standard for AI agents

AI agents are bots that can autonomously solve tasks, without the need for you to specify every step in the process. Instead, they work those steps out on their own.

We will hear a lot more about the technology in the coming 12 months, as AI labs start to roll-out their own tools.

However, businesses are slightly apprehensive about adopting the tools we have today.

They worry about betting on the wrong tool, which could be discontinued and force them to use another tool instead.

To tackle this, LangChain has proposed an “Agent Protocol” to standardise how agents interact and allow companies to use a combination of different tools.

LangChain also has its own framework for agents, called LangGraph, and this seems to be a good move to address the concerns from businesses.

3. Amazon will invest another $4 billion in Anthropic

The company has announced that it will commit an additional $4 billion to Anthropic, which brings its total investment to $8 billion.

It follows previous rounds of $1.25 billion in September and $2.75 billion in March.

As part of the deal, Anthropic will use the AWS platform to train future AI models.

Both companies are partnering together, as they hope to rival Microsoft and OpenAI’s joint venture.

Anthropic’s AI model will also be used to power the next-generation of Alexa, which has faced delays for some time now.

4. UK crashes out of global top 50 supercomputer ranking

The UK has fallen out of the top 50 in the global supercomputer rankings, according to the latest data from the Top500 project.

Their national supercomputer, Archer2, is nearing the end of its life in 2026 and has dropped to 62nd place globally, down from 49th in June and 38th last November.

The decline comes on the heels of the new Labour government's decision to shelve plans for an £800 million investment in a new "exascale" supercomputer at the University of Edinburgh.

However, this isn’t very surprising. The UK has failed to invest in critical projects for over 14 years and has never truly recovered from the 2008 financial crash.

5. Europe’s battery startup Northvolt files for bankruptcy

The Swedish battery manufacturer has filed for bankruptcy in the US, dealing a significant blow to Europe's strategy for domestically produced lithium-ion batteries.

The company has filed Chapter 11 and needs to restructure its finances. Their co-founder and CEO Peter Carlsson has stepped down, although he will remain as an advisor and board member.

While Northvolt has raised $14.26 billion overall, including a $1.2 billion round in 2023, it was reportedly burning through $100 million per month.

In June, BMW decided to withdraw from a $2 billion contract and this quickly threw the company into disarray. Months later, the company had to lay-off 20% of its staff.

Volkswagen has been a significant investor in the company and is betting its future on electric vehicles, so time will tell what happens to Northvolt.

AI scaling laws are showing diminishing returns

In the last few weeks, we have heard rumours that next-generation AI models are not improving at the rate they used to.

For several years, we could throw more data and more power into the mix - leading to AI models that are substantially better than before.

That doesn’t seem to be the case anymore. Until we reach a new breakthrough in how this technology is developed, we will see slower rates of improvement.

One of OpenAI’s co-founders, Ilya Sutskever, has recently admitted this in interviews. His new startup is focused on using other strategies to achieve those improvements.

Even if there is a slowdown, the technology we have today is incredibly useful. It’s slightly overhyped by the people who stand to benefit financially, but still incredibly useful.

Beyond summarising documents and answering questions, we can use multiple LLMs to analyse content and automatically improve their responses.

We can communicate more naturally with AI models than ever before. Previously, we had to use very specific phrases and hope that the chatbot understood what we meant.

That’s no longer the case. There may be diminishing returns for today’s technology, but there will eventually be another breakthrough that upends it and the cycle will continue.

Google Lens can now check prices when you’re shopping

The company is updating the Lens app to help you shop in physical stores. By taking a photo of an item, their tool can search the web and find the cheapest price.

You can then decide whether to buy it now or order it online instead. It also shows reviews of the product and other stores that have it in stock.

It’s a great use of Google’s search engine and their AI models, which are used to analyse the image and figure out what product is being shown.

Since Google has their own Shopping index, they can search through billions of items and find the cheapest price.

This is really similar to Apple’s upcoming Visual Intelligence feature, which allows you to quickly take pictures of a cafe and read their online reviews.

🌐 OpenAI could create its own browser

💻 Microsoft will launch a commercial quantum computer in 2025

🛩️ Tekever raises $74 million for their drone platform that’s active in Ukraine

🤖 Chinese lab releases a reasoning AI model that rivals OpenAI's o1

🗣️ Microsoft creates a new language for AI agents, called DroidSpeak

🛡️ Investment in defence tech sets a new record

⚡ Tokamak Energy raises $125 million for its fusion reactor

⚛️ Kairos gets approval to build two small nuclear reactors, will supply Google

📚 OpenAI releases a teacher's guide to using ChatGPT

🔬 DeepMind unveils AlphaQubit to make more reliable quantum computers

H

This startup is based in Paris and has raised an astonishing $220 million, despite never releasing a single product.

That has changed recently, with the launch of Runner H and their Studio platform. This is allows businesses to easily program their own AI agents for specific tasks.

For example, you might want to test that the buttons on your web page are working as expected.

Today, we use a long list of tests that are written with code. It’s time consuming to do and you have to think of most scenarios.

With Runner H, the ambition is that you can just use natural language and tell the agent what to do.

The company was founded by former Google employees, but faced a setback early on. Three of the five co-founders left due to disagreements over the company’s future.

H has recently opened up a private beta for the platform, which you can check out below.

This Week’s Art

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

• Baidu’s plans for an ultra-affordable robotaxi in China

• LangChain's new standard for AI agent development

• Why Amazon is investing another $4 billion in Anthropic

• UK's fall from the global top 50 supercomputer rankings

• Why Northvolt, Europe's battery startup, went bankrupt

• The strategy change as AI models start to show diminishing returns

• How to check prices with Google Lens when you’re shopping

• And how H are developing a platform for AI agents that search the web

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.