🤖 Google’s newest AI tools are here, including Agents

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 ‏‏‎ ‎

• Google’s new platform for building AI Agents

• Why the Humane AI Pin doesn’t live up to the hype

• China’s plan to remove American semiconductor chips

• … and much more

Let's jump in!

1. Microsoft has pitched DALL-E to the military

We start with Microsoft, who have recently suggested that OpenAI’s image generator could be used by the US Department of Defense.

This report gained a bit of traction online, but it’s not as outrageous as the headlines from journalists would suggest. The brief presentation is purely informative and outlines different use cases - with one being to use DALL-E images and then train new AI models for military intelligence.

Quite frankly, all the top companies are doing this. They’re all exploring how AI images can be used to develop new tools. The only difference here is that the DoD could use it for their own operations and information gathering.

Just last year, Meta’s research showed how GenAI images can be used to train robotic arms - which is much faster than their other methods.

Previously, this would’ve taken a very long time, as Meta’s staff had to manually take thousands of photos for the robot. But image generators can do this within an hour.

This technology opens up new doors that simply weren’t possible before, largely because it was really difficult to collect that data.

I don’t think it’s realistic to say that militaries can be prevented from using this tech, as there are plenty of open source options that they can choose from.

2. Google launches Code Assist to challenge GitHub’s Copilot

The company had a busy week, with lots of new AI tools unveiled at their Google Cloud event. Code Assist is an updated version of their previous tool, Duet AI, and is better positioned as a challenger to GitHub Copilot.

Interestingly, Code Assist will have a context window of over 1 million tokens. This is really important, as it allows the model to analyse larger amounts of data. For developers, this means that the AI can scan more code files and offer suggestions.

Another point that separates it from Copilot is that Google’s tool is available for multiple platforms - such as GitHub, GitLab, BitBucket and on-premises. It’ll be fascinating to see what the tool suggests and how developers use it going forward.

3. Texas is replacing thousands of human examiners with AI

The new automated system is expected to save the Education Agency around $15-20 million per year, as there’s less of a need for human reviewers.

Answers for the STAAR standardised exam will be reviewed by the AI and given a grade. It has been trained on over 3,000 exam responses, which were reviewed by humans at least twice.

The AI system will have several safety nets in place, with a quarter of all the AI-graded results being reviewed by humans. If the AI is unsure about how to grade an answer, it will be flagged for a human to check.

We’re going to see a lot more systems like this, but they will be a real challenge to implement properly. Teachers and students are unlikely to trust these AI systems unless there is complete transparency in how it reached a decision, which they can easily check and appeal.

4. Humane’s AI Pin doesn’t live up to the hype

The AI Pin got a lot of attention last year, as it aimed to be the first device that took full advantage of the technology behind ChatGPT.

It’s designed to be worn on your clothes and uses a projector to display graphics onto your hand. However, reviewers have been critical about how it actually works in practice.

For one, it struggled to perform basic tasks - such as making phone calls, sending text messages, or even setting alarms and timers.

There are also problems around battery life and overheating, since the device would often shut down after just a few minutes of use.

Of course, the AI Pin is being pitched as “anti-smartphone” and something that seeks to reduce your reliance on screens. But given the high price tag and list of problems, it seems a lot more is needed to make this a reliable product for consumers.

5. China is planning to remove American chips by 2027

Following on from the “Delete America” plan I covered a few weeks ago, it’s been reported that China wants all foreign computer chips to be removed from telecom networks.

American giants, such as Intel and AMD, are the most likely to lose out - with a deadline set for 2027. These chips will be replaced by domestic companies instead, which have become much more feasible options in the last few years.

This is an accelerating trend, which was started several years ago when China’s Huawei was banned by several Western countries - like the US and UK. Since then, both China and the West have been positioning themselves to reduce reliance on each other.

The US is investing heavily to produce semiconductors within their own borders, while China is looking towards their own domestic companies instead.

Google steps into the world of AI Agents

The tech giant has announced Vertex AI Agent Builder, which allows you to easily create agents that can analyse what you’ve requested and actually complete the task.

AI Agents are a huge topic and something I’ve been actively exploring for months. While Generative AI has struggled to live up to the hype for some businesses, AI Agents could close that gap for many.

Companies will be able to quickly create their own custom agents and supply it with the data it needs. However, Google’s Search product gives them a huge advantage over their rivals - allowing the AI to search for answers when it’s unsure.

We are going to see even more of the big tech companies reveal their work on Agents. For some time now, OpenAI has been offering Assistants to their customers.

Microsoft also offers this, but their research team is actively working on how multiple Agents can work with each other. Their AutoGen tool is incredible and something that all businesses should be aware of.

While existing businesses will use Agents to speed up their work, they are also at risk of disruption. This is because we’re facing a future where anyone can create teams of AIs that can work together. And that future is very close to a reality.

If you’re an executive or senior manager, which many of my readers are, you need to start considering how teams of AI Agents will impact your business - along with ways to adopt it and stay ahead of the competition.

Planet reveals their new satellite platform

The company is one of the top providers of satellite imagery, alongside Maxar & Airbus, and has announced a new platform to simplify the process.

In recent years, it has become much cheaper for us to download satellite data and analyse it for new insights.

I’ve been interested in how it can be used by both businesses and non-profits, with use cases ranging from competitive analysis to monitoring crop yields to surveillance.

This new platform now brings in some of Sentinel Hub’s features, which the company acquired last year - allowing you to easily stream, analyse, and then distribute satellite data.

It also has updated models for detecting changes to roads and buildings, along with some of their other features - like boundary detection for farms, and ship or aeroplane detection.

In the UK, government agencies are using satellite imagery to monitor farms across the country and support their work to allocate farming subsidies.

🔍 Meta is testing an AI-powered search bar in Instagram

📦 Walmart will deploy robotic forklifts in its distribution centers

🚨 How Ukraine’s cyber police fight back against Russia’s hackers

⚖️ UK’s antitrust regulator raises concerns over Big Tech’s grip on GenAI

💻 Meta unveils its latest custom chip for AI

🤖 Amazon adds Andrew Ng to its board

Collaborative Robotics (Cobot)

This startup is based in Santa Clara, which is working to develop robots that can seamlessly collaborate with humans.

The company was started just 2 years ago and has recently secured $100 million in Series B funding. It’s being led by Brad Porter, who had previously worked at Amazon for over 13 years and was in charge of their industrial robotics program.

Essentially, the company is aiming to develop more flexible robots than those that are designed for specific tasks.

Typically, we see a lot of robots in manufacturing and warehousing - but Cobot are looking more generally. This could include robots that walk around you, deliver parcels or can answer questions in retail environments.

They haven’t unveiled their system just yet, but they have confirmed that it’s being trialled in multiple cities.

It’ll be interesting to see what this looks like in practice, but given the huge amount of funding and experience of people involved it’s definitely a startup to keep an eye on.

This Week’s Art

We’ve covered a lot this week, including:

• Why Microsoft has pitched DALL-E to the military and how the big tech companies are using it to train robots

• Google’s new challenger to GitHub Copilot

• How Texas is using AI to grade standardised tests

• Why the AI Pin doesn’t live up to the hype

• China’s plan to remove American chips by 2027

• Google’s latest build tool for AI Agents

• Planet’s new satellite platform

• And how Collaborative Robotics are working on general-purpose robots

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 Software Engineer and works within Kainos' Innovation team. He identifies business value in emerging technologies, implements them, and then shares these insights with others.