Welcome to this edition of Loop!

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

‏‏‎ ‎ HIGHLIGHTS ‏‏‎ ‎

Let's jump in!

1. Uber becomes the go-to-platform for autonomous vehicles

We kick off this week with Uber, which has started to accelerate its spending on robotaxis and signed new partnerships in the transport sector.

This process was started last year, after Tesla unveiled their plans to create their own robotaxi fleet and compete with Google’s Waymo.

Uber’s shareholders became alarmed by this and started to put significant pressure on the company’s leadership. As a result, the company has doubled down on self-driving technology and partnered with several companies - including Aurora, Wayve, and Waymo.

In the last few days, it has also partnered with China’s Baidu and will deploy their autonomous vehicles across Asia and the Middle East.

Uber is now positioning itself as the go-to-platform for self-driving cars. This allows the vehicle manufacturers to access its massive customer base and Uber can facilitate the payment process.

Previously, Uber had its own ambitions to become a self-driving company. But that all fell apart in 2018 when its test vehicle killed a pedestrian and Uber abandoned that strategy.

In the short term, it makes sense. Why would you try to build expensive autonomous tech, when you can let others handle that and provide them with millions of customers?

In the longer term, Uber could find itself out-manoeuvred - especially if they charge too much for the platform.

In that case, self-driving companies might decide to go it alone and build their own service - but this won’t happen anytime soon, as the autonomous vehicles need a technology breakthrough to become both reliable and safe.

2. NotebookLM creates dedicated notebooks for Shakespeare, finance, news

Google's NotebookLM is evolving from a simple AI research tool into a proper content platform.

The search giant has added curated notebooks from prestigious publications and experts, marking a significant shift in how the product positions itself.

You can read notebooks from The Economist, which detail some of their annual predictions, or even ask questions about Shakespeare's complete works.

There’s also one for those interested in finance, outlining Q1 earnings from the world's top 50 public companies.

How you interact with this content is similar to before, except we now have dedicated spaces for a range of different topics. You can ask questions, be shown cited answers, and use mind maps to explore the content more broadly.

It's a clever strategy from Google, as it really showcases what their tool can do and also encourages users to share their notebooks with others.

3. GM makes significant investments in EV batteries

In partnership with Redwood Materials, their EV batteries will be given a second life and instead used to power data centres - possibly solving the difficult problem of e-waste.

When Redwood tested GM’s batteries, they found that most still had a lot of capacity and were too valuable to simply break down for materials.

JB Straubel, Redwood's CEO, believes that this new energy storage division could outpace their core recycling business. That's quite the claim, considering that the company already processes roughly 70% of America's used or discarded batteries.

At the same time, GM is working with LG to upgrade an existing factory and produce LFP battery cells. These are cheaper to manufacture than NMC batteries and China holds a dominant position in this market, which GM is hoping it can break.

Since LFP batteries are cheaper, they could lead to a new wave of affordable electric vehicles that can challenge the Chinese brands.

4. Google’s AI agent spots a critical security flaw in SQLite

DeepMind and the Project Zero team worked on the agent, which has moved from concept to reality remarkably quickly. Remarkably, this AI agent was able to spot and fix a SQLite vulnerability that attackers were preparing to exploit.

Google claims that this is the first time an AI agent has prevented a zero-day attack in the wild. Nicknamed “Big Sleep”, it is now being deployed on other open-source projects to identify more issues.

For some time, I’ve been interested in how this technology can be used for cyber security - as you can have multiple agents that analyse code and give them different perspectives.

Some might review your website and only focus on injection attacks, while others might focus on social engineering and ways to trick humans. They can also work together and figure out different strategies to attack a system - or defend it, as seen in this case.

However, most companies will struggle to test this out themselves - as the leading tech companies monitor how LLMs are being used and could mistakenly flag your account as a “bad actor”.

Open source models can be used instead, but they lack the intelligence that is often needed for these kinds of defensive tools. It’s not something that most companies can experiment with, unless you have close ties with the top tech companies.

Regardless, it’s fascinating to see Google’s work on “Big Sleep” and how they’re using this to safeguard other software tools.

5. US invests billions to make Pennsylvania an AI hub

President Trump has unveiled over $90 billion in private sector commitments, which include new infrastructure projects for AI, data centres, and energy stations.

There were a lot of announcements, so it’s best to cover them with bullet points instead:

Quite frankly, these are mind-boggling sums of money and will only engrain the US’ advantage over Europe and the rest of the world.

China has also been racing to build new energy infrastructure, with 10 new nuclear reactors approved in April.

Not all of that is related to artificial intelligence though, as China also wants to diversify their energy sources and reduce its reliance on coal - which it imports from other countries.

OpenAI’s agent can now do tasks for you

OpenAI has unveiled ChatGPT Agent, as the company begins its push into the AI agents arena that's captivating Silicon Valley. The new tool promises to handle complex, multi-step tasks autonomously using what OpenAI calls its own "virtual computer."

Unlike standard chatbots, this agent can easily switch between multiple tools - such as browsing the web, running terminal commands, and importing user data. Interestingly, they decided to retrain a model specifically for this agent.

OpenAI included some strange demos to showcase the agent. In one of them, the agent was asked to plan a date night. It then used Google Calendar to check what dates are best and cross-referenced it with OpenTable reservations.

I’m not sure that’s a pressing issue for millions of people. Yes, it’s interesting to see - but surely they could have showcased a better example? We’ll need to wait and see how other people use the tool, with public access expected in the coming days.

Performance-wise, it's pretty slow. Tasks can take 15-30 minutes to complete, but OpenAI frames this as acceptable for complex work you'd otherwise do yourself. That’s similar to OpenAI’s Codex agent for software development, which also takes a long time to solve simple tasks.

Overall, I’m cautiously optimistic about the tool. Based on the examples given, I suspect that it can only tackle simple problems - but even that can save people time, especially at work.

However, you should be really careful when using the agent. For example, malicious websites could secretly order your agent to make a payment or send your personal information. That’s a huge risk with agents that can browse the web and we don’t really have good safeguards to prevent that.

In the meantime, I recommend you trial the tool and explore how it could be used for small tasks - like creating presentations, spreadsheets, or converting data into another format - then expand from there to find the tool’s limits.

Joby Aviation doubles its manufacturing capacity to build air taxis

I’ve flagged Joby Aviation as an emerging startup before, which you can read here. There are other companies operating in this space, but I don’t believe that many people are focusing on this emerging technology - which is a mistake.

These eVTOLs are powered by electric motors and can take off or land vertically, allowing us a lot more transport options into busy cities. The technology is also being closely watched by Western militaries, as they want to use these aircraft and silently carry out missions.

Joby has just doubled its manufacturing capacity to 435,000 square feet, as it aims to produce 24 aircraft per year. Beyond manufacturing, the expanded site will also be used to train pilots, maintain existing aircraft, and achieve new certifications with the FAA.

They’re also working on another factory in Ohio, which promises to build over 500 aircraft every year. It’s an important step for the company, as it prepares for next year’s commercial launch in Dubai - before then expanding to other cities in the US.

Since Toyota announced that it would invest $500 million in the company, they have sent over their own engineers to plan the new manufacturing hubs.

The rapid expansion plans and Toyota’s partnership show that Joby is serious about meeting its aggressive timeline, but the real test will come next year - when these flying taxis must prove themselves in Dubai's skies.

🧠 Study warns of "significant risks" in using AI therapy chatbots

🏢 Meta is building a data centre the size of Manhattan

🛒 Prime Day drove $24 billion in US sales

💻 Company behind the AI coding agent Devin has acquired Windsurf

🔐 US will spend $1 billion on "offensive" hacking operations

🎬 Netflix used GenAI in a TV show, for the first time

🤝 Meta hires two more OpenAI researchers

🔒 Co-op, the UK retailer, confirmed that 6.5 million customer records were stolen

💰 Lovable raises $200 million, just 8 months after launch

📝 US government announces $200 million contract for xAI's Grok

☁️ AWS launches their own tools for agentic AI

🇪🇺 Meta refuses to sign EU's AI code of practice

Auriga Space is betting that electromagnetic tracks can be used to launch rockets and change how we send payloads to space. The California startup has secured $4.6 million in seed funding to develop the technology.

Rather than burning precious fuel to fight against gravity, Auriga's system would use powerful magnets and accelerate to six times the speed of sound - before they even ignite their engines.

The concept isn't entirely new, but their founder believes that recent advances in electronics have finally made it commercially viable.

It’s certainly ambitious, but there is a growing trend of VCs backing space startups. Following the war in Ukraine and conflicts in other parts of the world, there has been a rush to invest in defence and space startups.

Previously, it was difficult to get that foot in the door and many investors worried that these industries were too complex for small players.

While Auriga is at an early stage and still needs to develop the technology further, they’ve raised over $12.2 million so far and should have plenty of runway.

There are a lot of innovative startups in the AI and technology sector more widely, but it’s fascinating to see this investment in space technologies too. If they’re successful, it could be a huge breakthrough for the West as well.

This Week’s Art

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

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

Or if you’re interested in chatting with me about the above, simply reply to this email and I’ll get back to you.

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.

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.