🍎 Why Apple’s partnership with OpenAI could boost the industry

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 ‏‏‎ ‎

• Apple’s deal with OpenAI and the wider impact for users

• Why tech companies are joining forces on AI chips

• Google’s struggles with AI search results

• … and much more

Let's jump in!

1. Apple is about to strike a deal with OpenAI

We start with Apple, who have reportedly struck a deal to incorporate OpenAI's technology into their software products.

The deal could be worth billions of dollars and has been leaked shortly before Apple’s developer conference, which will focus on new AI tools.

It’s believed that John Giannandrea, the head of machine learning at Apple, is against including the technology - possibly due to concerns about biases in the models and the risk that some responses can lead to PR embarrassments.

As you can imagine, Microsoft’s CEO is also concerned about how the Apple deal will impact their $13 billion partnership with OpenAI.

A lot of people have focused on how this deal will solidify OpenAI’s position as a market leader. However, there’s a more interesting angle here.

For decades, Apple has been fantastic at focusing on the user’s experience. This is seen in their product design, the packaging that they are delivered in, software products, and support for when something goes wrong.

Currently, a lot of companies have created AI tools that look and feel like chatbots - which isn’t the best experience for users. What I’m interested in is how Apple re-imagines this area and comes up with new ways to interact with Generative AI.

If they get it right, it could have more of an impact on OpenAI’s fortunes than the launch of ChatGPT did.

2. Google outlines why users were shown strange AI search results

In a recent blog post, Google's head of search wrote about the issues behind the inaccurate AI search results - after there was some criticism online.

Some of these AI results said that you should eat at least one rock a day, or use “non-toxic glue” to make cheese stick on pizza. There are many more examples, although it’s possible that some of them were faked.

Liz Reid said these results were caused by "data voids" and users asking unusual questions, but stressed that the AI results still led to higher customer satisfaction.

In many ways, this was always going to happen. Generative AI models aren’t grounded in truth, so they will occasionally make stuff up.

Google has a great reputation for providing the best search tool on the market and offering accurate results. Perhaps part of the problem was that they integrated this AI feature into their famous search bar.

If the company used a slightly different user interface, which made it clear that the results could be wrong, the strange results would have got slightly less press coverage.

Ultimately, this won’t cause any problems with their reputation. But companies should be cautious around how they craft the user experience with Generative AI - and to stress that it won’t always be correct.

Otherwise, you might get some embarrassing headlines in the media.

3. Tech giants join forces to develop AI chip components

Intel, Google, Microsoft, Meta, and other tech giants have formed the Ultra Accelerator Link (UALink) Promoter Group - which aims to develop a new industry standard for connecting AI accelerator chips in data centres.

The hope is that UALink 1.0 will connect up to 1,024 GPUs across a single computing "pod" - with the expectation that this will improve performance and reduce data transfer latency. The first product launches should happen within the next couple of years.

One thing to note is that Nvidia, the top company for AI chips, is not a part of this group. This isn’t surprising, as Nvidia holds a hugely dominant position in the market and the other companies are likely trying to lessen that grip.

4. Anthropic hires former OpenAI safety lead to head up new team

Jan Leike, a prominent AI researcher, has joined Anthropic to lead a new "super-alignment" team after resigning from OpenAI and publicly criticising the company's approach to AI safety.

His new team at Anthropic will concentrate on various aspects of AI safety and security, including "scalable oversight," "weak-to-strong generalisation," and automated alignment research.

Leike's team appears to have a similar mission to OpenAI's recently dissolved super-alignment team, which he used to co-lead. They aimed to solve the challenge of controlling superintelligent AI, but the team alleged that they were hindered by OpenAI's leadership.

Anthropic has frequently attempted to position itself as more safety-focused than OpenAI, which was actually created due to a split within OpenAI. Several key researchers disagreed with the company’s growing commercial focus and instead left to setup Anthropic.

5. Google confirms the leaked Search documents are real

The 2,500 internal documents were leaked online and revealed what data Google collects, some of which may be used in its closely guarded search ranking algorithm.

The documents suggest that user clicks and Chrome data might be used to rank web pages for Google’s search results. However, it’s not clear if this is definitely the case.

The leak started to gather attention after SEO experts, Rand Fishkin and Mike King, published some of the leaked documents' contents and their own analysis.

It could have far-reaching implications for website developers, who will likely create their own experiments to test the suggestions in the documents.

If they find that some do impact the search results, it could lead to changes in how your favourite websites look and feel.

Perplexity will turn your AI searches into shareable pages

Perplexity is introducing Pages, which is a new feature that allows you to create reports, articles, and even user guides from your search query.

These can then be published on the web and include graphics, such as images and videos. If the AI results aren’t perfect, you can edit the page or ask it to rewrite a specific section.

This is positioning Perplexity as more of a research-focused company, which is critical for a lot of enterprises and seems to be a shift away from general search queries.

It’s a smart move, as large enterprises are more likely to spend huge sums to increase productivity. This will really boost the startup, since they need to increase revenues to develop their own custom LLMs and keep up with Google.

Pages is slowly being rolled out for users, but isn’t available for everyone just yet. Once you do get access, you’ll find it in the library section and can use it to start generating reports.

ElevenLabs now lets you generate sound effects

The company is well-known for their AI voice cloning tool, but they’ve just announced a new feature that lets you create your own sound effects.

By simply typing a text prompt, you can generate up to 22 seconds of audio and select between four options. It was made in collaboration with Shutterstock, who provide stock media for others to use.

Of course, this will not be welcomed by those who work on creating sound effects - as it will reduce the cost to do this. In general, the creative industry has been very concerned over how their work is being used by huge AI companies.

However, this is unlikely to have a big impact on huge media productions - like movies and TV shows. These models don’t do a great job at recreating new effects, as they rely heavily on what they’ve processed before.

There will still be a need for sound engineers who can create custom effects. Tools like this one can help them to experiment with different ideas and complete the work faster.

While the sound effects feature is available for free, you will need to be on a paid tier for commercial use.

📦 Amazon gets FAA approval to expand US drone deliveries

🧑‍💻 Former OpenAI board member explains why they fired Sam Altman

🛫 Plan to bring electric air taxis to South Korea by 2026

🚗 Chinese EV makers are targeted by a new House bill

💰 Elon Musk’s xAI raises $6 billion in funds

📚 OpenAI is making ChatGPT cheaper for schools and non-profits

ZeroMark

ZeroMark is a startup that develops weapons which can automatically aim at a drone and disable it. They’re based in New York and their product could really benefit Ukraine’s soldiers, as they face constant drone attacks from Russian forces.

The startup is already in talks with the US Department of Defense and the Ukrainian army. It recently closed a $7 million round of funding, which will be used to fund more product development.

Essentially, their product uses sensors and machine learning to predict where the drone will be. Their system then adjusts the rifle’s aim and makes it more likely that soldiers will hit the drone, which could be over 200 yards away.

There are legitimate concerns over technology that can automatically aim weapons, as it could lead to terrible results if aimed at humans. ZeroMark are cautious over who they sell to and don’t provide it to the US police.

Quite frankly, this sounds like something from the Terminator movies - which is why it’s important we’re aware of these weapons and how they could be used.

This Week’s Art

We’ve covered a lot this week, including:

• Apple’s deal with OpenAI and the wider impact

• Google’s strange AI search results

• Why tech giants are joining forces to develop AI chip components

• Anthropic hires the former OpenAI safety lead

• Google’s leaked Search documents are real

• Perplexity’s new feature for AI research

• ElevenLabs’ tool for generating sound effects

• And how ZeroMark are developing weapons to automatically disable drones

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.