About Raccoons, Crickets and Champagne for Breakfast

Good morning.

Interesting things are happening on the intersection of artificial intelligence and cybersecurity.

Generative AI is being used by criminals to make social engineering and phishing almost impossible to spot. You will soon no longer need to be able to hack in order to actually hack. You’ll just tell an agent to “hack a public facing IP address” and it’ll find its way through.

And soon, entire bot armies will be fighting each other on the borders of corporate networks.

Then, on the other side, using AI has inherent risks as well. There have been examples of chatbots turning into your racist uncle, of automated trading bots that have caused wild swings in a stock, or of autonomous vehicles losing the plot.

The intersection of both fields is quite an interesting place. And then we even haven’t mentioned the P-word (privacy , you perv, i’m talking about privacy).

If you’re from Belgium or The Netherlands, I recommend you listen to this week’s episode of the Radio Raccoons podcast, in which yours truly was invited to speak about this intersection of cyber and AI.

Raccoons develops AI tools for companies and has a podcast that talks about the latest AI news. Together with this publication, it’s all you need to catch up with the very latest in AI. So give them a like or subscribe.

As for the news this week… I can’t believe I’m about to say this but it was kind of slow- which is only the second time in almost a year since we started the newsletter. It was the sound of crickets.

There were not a lot of releases or spectacular upgrades this time.

But maybe we’ve become so used to dramatic news and new technologies emerging that even when Google announced they just added 110 new languages to Google Translate , we’re thoroughly unimpressed.

It’s like being served Dom Pérignon for breakfast every day and complaining if you’re served some inferior product like Veuve Clicquot.

I am sure this is the “silence before the storm” and either OpenAI or xAI are cooking up some drama.

Welcome to the Blacklynx Brief

AI News

• Apple is adapting its new AI features from Apple Intelligence for the Vision Pro headset, though these capabilities are not expected to launch this year. The AI suite includes an enhanced Siri, writing tools, and an OpenAI-powered chatbot. Despite interest in boosting sales, Apple has no plans for a partnership with Meta.

• Agility Robotics has signed a multi-year deal with GXO Logistics to deploy its Digit humanoid robots in warehouses, following a successful pilot with Spanx in 2023. This marks the first Robots-as-a-Service (RaaS) agreement and a formal commercial deployment of the robots. The Digit robots, capable of lifting up to 35 pounds, will be used for repetitive tasks and logistics work, controlled via Agility's cloud-based platform.

• Google has announced its largest expansion of Google Translate, adding support for 110 new languages using its PaLM 2 model. This update covers languages spoken by over 614 million people, about 8% of the global population, and includes some languages with no current native speakers. The expansion is part of Google's "1,000 Languages Initiative," aiming to support all spoken languages and help preserve endangered dialects.

• French startup Kyutai introduced Moshi, a new real-time AI voice assistant capable of responding with 70 different emotions and speaking styles. Moshi can listen and speak simultaneously with a very low latency, potentially outperforming similar features from OpenAI. The company plans to open-source the research soon, and Moshi is already available for trial on Hugging Face.

AI Tip of the Week

Since we’re on the topic of cybersecurity and privacy this week .. let’s give you a small tip.

If you have ever wondered why ChatGPT is free, the answer is that when something is free, it means that YOU are the product. In fact, these large language models need your data in order to improve themselves.

By providing the service for ‘free’ - OpenAI gets to build a new version of ChatGPT, and you’re helping them. Imagine what they’re learning now from the millions of people using the product.

If you think this is creepy and you don’t want it, you can opt out.

You go to Settings > Data Controls and you toggle “Improve the model for everyone” to "Off.”

This is “On” by default so putting it to “Off” is a good move. It’ll get trained without our interactions. On top of this , even though OpenAI says it won’t use your conversation to train the next model, this doesn’t mean you have to trust them.

They have proven over the last few months that they’re not really to be trusted.

Make sure you NEVER post personal data or trade secrets in there. It will get gobbled up and you don’t want your competition to find out about things your company is doing because you stupidly put it into chatGPT.

Treat ChatGPT like an ultra smart coworker, but never show your hand to this coworker. He might try to steal your promotion.

Quickfire News

• University of Tokyo researchers developed a new technique to bind living human skin to robotic faces, potentially enabling more lifelike androids and other medical applications.

• Eureka Health introduced Eureka, an AI doctor offering personalized care plans, lab orders, and medication adjustments under physician supervision, specializing in thyroid and diabetes treatments.

• Meta is set to begin testing its creator-made AI chatbots on Instagram in the U.S., with a broader rollout expected by August.

• Character AI launched Character Calls, a free feature enabling users to have two-way voice conversations with AI characters across multiple languages on the platform’s mobile app.

• New York University researchers published a paper on MLLMs and vision-centric approaches, introducing a new Cambrian-1 model family and a new benchmark for vision-related tasks.

• A Reddit user shared a clip that reportedly demos early access to OpenAI’s upcoming Voice Mode upgrades, showcasing the ability to weave in background sound effects during responses.

• Amazon is hiring executives from Adept and licensing its automation tools to build agents for software workflows.

• Andrew Ng’s AI Fund aims to raise $120M for its second venture fund, focusing on early-stage AI startups, which is lower than the initial $175M fund.

• The Center for Investigative Reporting filed a copyright infringement lawsuit against OpenAI and Microsoft, alleging they used its journalism for AI training.

• Meta’s anticipated Llama-3-405B model is reportedly seen in a recent WhatsApp update, suggesting the new version may launch soon.

• Amazon plans to spend over $100B on data centers over the next decade, with a focus on AI infrastructure.

• Actor Morgan Freeman thanked fans for calling out the use of AI to replicate his voice in a scam.

• Cloudflare released a free tool to detect and block AI bots circumventing website scraping protections, aiming to address concerns over unauthorized data collection for AI training.

• Magic is set to raise over $200M at a $1.5B valuation, despite having no product or revenue yet — as the company continues to develop its coding-specialized models that can handle large context windows.

• ElevenLabs launched Voice Isolator, a new feature designed to help users remove background noise from recordings and create studio-quality audio.

• Citadel CEO Ken Griffin told the company’s new class of interns that he is ‘not convinced’ AI will achieve breakthroughs that automate human jobs in the next three years.

• Shanghai AI Lab introduced InternLM 2.5-7B, a model with a 1M context window and the ability to use tools that surged up the Open LLM Leaderboard upon release.

• App Store chief Phil Schiller is joining OpenAI’s board in an observer role, representing Apple as part of the recently announced AI partnership.

Closing Thoughts

That’s it for us this week.

If you want a version of this newsletter with pictures etc. Be sure to subscribe below:

https://www.blacklynxbrief.com/subscribe