People Are Strange

Good morning,

We’re a strange bunch - we humans.

You could walk into any bar in your hometown - start talking about let’s say driverless vehicles and complete self-driving and half of the people in the bar will tell you you’re full of cr*p.

While at the same time, in San Francisco, Waymo is operating a network of driverless taxi’s. You have to see it to believe it. They say you see the future in San Francisco first. Indeed, I was there in March and it’s an almost religious experience to see one the first time.

Nobody told me about them in advance so imagine my surprise when I saw the first one zipping by (and this thing was FAST).

And yes , they still crash BUT if you analyze those crashes it’s because the other party has made a mistake and is NOT a robot. But of course, even with a great safety track record the Waymo’s are causing an outrage.

We don’t like change.

We either ignore it all together, downplay it or get angry at emerging technology.

This is nothing new. In 1864 - a cool 160 years ago - Fyodor Dostoyevsky wrote “Notes From The Underground”. Dostoyevsky is considered to be the greatest novelist of all time because he’s able to capture the essence of what it means to be human.

“Crime and Punishment” and “The Brothers Karamazov” are considered masterpieces and amongst the best books ever written.

So 160 years ago we were in the midst of the industrial revolution and just like with the fantasy that we are having about Superintelligence (no more work yay!) - people were having the same fantasy about technology solving all misery. (Didn’t pan out that way did it ?)

By the way you might have heard the saying “With great power comes great responsibility” and think this originally was spoken by Spiderman’s uncle but instead it was Dostoyevsky who first uttered it.

The skepticism against AI we are experiencing now is innately human. Read this excerpt from “Notes from The Underground”

Imagine an intelligent system that knows us better than we know ourselves, fine-tuning everything from our health to our work, from social interactions to personal happiness.

Superintelligence could be designed to give us a life of comfort and eliminate all our struggles.

You play World of Warcraft or golf all day and there’s no reason to “work”. You just find meaning in other things. You will have money nonetheless.

But as Dostoyevsky points out, humans are not content with just comfort. We're complicated. We crave meaning, wonder, and the thrill of taking risks—things that may not make logical sense but are core to our identity.

Dostoyevsky's point isn’t just a critique of human fickleness. It's a caution that no matter how advanced or benevolent ASI might be, a purely rational existence risks suffocating what makes us feel alive. If we were reduced to “the keys of a piano”—we might feel compelled to resist, even if it means risking our security or comfort. Why? To prove to ourselves that we are still free to make mistakes, to rebel, and to choose the absurd over the sensible.

What Dostoyevsky reveals is that humans need more than ease and efficiency. We need the space to embrace the unpredictable, the irrational, and even the nonsensical.

And all this from someone writing this to the light of a candle 160 years ago.

I’m going to miss us.

Welcome to the Blacklynx Brief!

By the way, there is now a referral program for this newsletter. For every 3 people that you get to sign up, we will gift you a 15€ Amazon gift card.

Scroll to the bottom of the newsletter for details!

AI News

• Midjourney launched an AI-powered web editor enabling users to modify, retexture, expand, and stylize both generated and uploaded images using text prompts. Initially available to select users, the editor introduces features like cropping, repainting, and retexturing, providing powerful new creative tools but also raising concerns about potential misuse for deepfakes and image manipulation.

• OpenAI is reportedly disbanding its AGI Readiness team, following the recent departure of safety advisor Miles Brundage, who expressed concern over the industry's preparedness for advanced AI. Brundage’s blog post questioned the readiness of OpenAI and the broader world to manage increasingly powerful AI systems, with parts of the team being reassigned under new Chief Economist Ronnie Chatterji. This follows the May disbanding of OpenAI’s Superalignment team, adding to concerns about the company’s shifting priorities around AI safety.

• The White House issued a national security memorandum directing federal agencies to accelerate AI adoption while restricting its use in sensitive areas like defense and intelligence. The memo prohibits AI from making autonomous decisions on nuclear weapons and targeting systems and calls for infrastructure improvements, including AI chip development and foreign espionage safeguards.

• Google is reportedly developing "Project Jarvis," an AI agent that can navigate and control web browsers for tasks like shopping, research, and travel booking by interpreting screenshots. Expected to debut in December alongside the Gemini AI model upgrade, Jarvis focuses specifically on browser-based tasks but currently operates with a slight delay.

• OpenAI CEO Sam Altman dismissed rumors about a December release of an "Orion" AI model, calling the reports "fake news." Despite this, a previous cryptic tweet by Altman hinting at “winter constellations” has fueled ongoing speculation of a major release by year-end.

• Concerns have been raised about OpenAI’s Whisper transcription tool, which has reportedly generated hallucinated text in up to 80% of examined cases, even adding inaccurate or troubling content. Despite OpenAI’s warnings, over 30,000 healthcare professionals use Whisper-based tools, exposing the risks of using AI with high-stakes data.

• Apple launched Apple Intelligence with iOS 18.1 and macOS Sequoia 15.1, introducing features like systemwide text editing, improved photo search, and a revamped Siri with better contextual awareness. The initial release is available only on newer devices, with advanced features like ChatGPT integration expected in December.

• Meta is developing its own AI-driven search engine, aiming to reduce reliance on Google and Bing by creating its own real-time information infrastructure. The project, in development since early 2024, includes web-crawling capabilities and a new partnership with Reuters for reliable news content.

• Universal Music Group (UMG) has partnered with AI startup Klay Vision to develop a copyright-respecting AI music generator, following legal actions against other AI firms over copyright concerns. Klay Vision’s new Large Music Model, KLayMM, will prioritize ethical use of artist likeness and copyrighted material.

• GitHub announced it is expanding its AI coding assistant to include models from Anthropic and Google, allowing developers to choose between Claude, Gemini, and OpenAI’s models, which remain the default. New updates to the platform include multi-file editing, Copilot code reviews, and support for Apple's Xcode, plus the addition of "Spark" for app-building via natural language prompts.

• OpenAI is partnering with Broadcom and TSMC to develop a custom AI chip, with plans to incorporate AMD’s MI300X processors into its infrastructure to reduce reliance on Nvidia. The custom chip, expected by 2026, is part of OpenAI's strategy to manage soaring compute costs, as the company faces potential losses of $5B against $3.7B in revenue.

• Biotech startup Iambic Therapeutics launched Enchant, an AI platform that predicts drug efficacy in human trials using laboratory and limited clinical data, achieving a 74% accuracy correlation—well above previous benchmarks. Enchant can make predictions after analyzing as few as five drug molecules, providing early insights into drug performance and reducing development timelines.

• Design startup Recraft revealed its V3 AI model, previously codenamed "Red_Panda," which has now topped the Artificial Analysis leaderboard, surpassing models like Midjourney and FLUX. V3 introduces advanced features, including realistic text generation, precise human anatomy, and scene spacing, along with enhanced designer controls like brand color customization and vector capabilities. This surprise debut elevates Recraft to a major player in AI design, offering tools that combine high performance with extensive creative control for designers.

• Boston Dynamics showcased new footage of its Atlas robot autonomously sorting automotive parts, highlighting its real-time adaptability and machine learning-based navigation. Atlas, operating without human input, can recognize bin locations and correct mistakes, marking progress toward commercial humanoid applications in manufacturing.

• Osmo achieved a breakthrough in "scent teleportation," using AI and advanced sensory technology to digitize and recreate the smell of a plum without human input. This process combines gas chromatography, mass spectrometry, and AI analysis to map scent molecules, enabling the first successful digital replication of scent.

Quickfire News

• TSMC’s Phoenix plant reported superior chip yields compared to its Taiwan operations, increasing confidence in the U.S.'s domestic semiconductor strategy.

• Anthropic launched a new analysis tool for Claude, enabling models to write and execute code directly within chat interactions.

• Apple announced a $1 million bug bounty ahead of its major AI cloud release, rewarding researchers who find vulnerabilities in its private AI infrastructure.

• ElevenLabs introduced "Voice Design," a feature allowing users to create AI-generated voices from natural text prompts.

• OpenAI scientist Noam Brown shared at TED AI that giving models 20 seconds to "think" can provide the same performance boost as scaling training data by 100,000 times.

• Chinese robotics startup EngineAI unveiled SE01, a life-size humanoid robot with a highly realistic, human-like gait.

• Perplexity CEO Aravind Srinivas announced on X that the AI search platform now processes over 100 million queries per week.

• Meta secured its first AI news partnership with Reuters, enabling real-time news responses through its AI chatbot on Facebook, Instagram, WhatsApp, and Messenger.

• Coinbase introduced "Based Agent," a tool that lets users create AI-powered crypto trading bots with on-chain capabilities in under three minutes, integrating OpenAI and Replit.

• Disney is reportedly preparing to launch a major AI initiative centered on post-production and VFX workflows, marking its first significant adoption of AI technology.

• Meta released NotebookLlama, an open-source tool similar to Google’s NotebookLM, which converts PDFs into podcasts using text-to-speech capabilities.

• A new model named "red_panda" unexpectedly took the top spot on the Artificial Analysis Text-to-Image Arena Leaderboard, surpassing models like Flux, Ideogram, and Midjourney.

• AI startup Sierra raised $175 million, bringing the company’s valuation to $4.5 billion just a year after its launch, led by OpenAI board chairman Bret Taylor.

• Google expanded its AI Overviews in Search to more than 100 countries and additional languages, aiming to reach over 1 billion monthly users.

• OpenAI CFO Sarah Friar revealed that 75% of OpenAI's revenue is from paying ChatGPT subscribers, with a conversion rate of 5–6% from free users and 250 million weekly active users.

• xAI’s Grok chatbot gained new vision capabilities, demonstrated by Elon Musk in an example where the model analyzed a meme and explained the joke.

• Hollywood union SAG-AFTRA signed an agreement with AI company Ethovox to create a foundational voice model for digital replicas, ensuring compensation through session fees and revenue sharing.

• Elon Musk predicted at the Future Investment Initiative conference that by 2040 there will be at least 10 billion humanoid robots, each priced between $20,000 and $25,000.

• Amazon expanded its Rufus AI shopping assistant in beta to European markets, providing personalized product recommendations and comparisons via conversational interactions in its mobile app.

• OpenAI introduced new search capabilities for ChatGPT history, enabling users to easily reference and revisit past conversations.

• Elon Musk’s xAI is reportedly seeking a new funding round that could value the startup at $40 billion, up from its $24 billion valuation in May.

• Google CEO Sundar Pichai shared that Project Astra, Google’s multimodal, agentic smartphone app demoed at Google I/O, could be available as early as 2025.

• Actor Robert Downey Jr. criticized the use of AI-generated digital replicas in Hollywood, declaring that he intends to sue any future executives who recreate his likeness, even posthumously. Not sure how he’s going to do that but it’s a nice Halloween touch there Robert.

• OpenAI rolled out Advanced Voice Mode for ChatGPT on desktop platforms, extending its conversational AI capabilities to PC and Mac users after a successful mobile launch.

• Google CEO Sundar Pichai shared during the Q3 earnings call that AI now generates over 25% of Google’s code, with human engineers conducting reviews and approvals.

• Time Magazine released its "Best Inventions of 2024" list, featuring an AI category that includes innovations like Google DeepMind’s AlphaFold 3, Runway’s Gen-3 Alpha, and Google’s NotebookLM.

• OpenAI introduced SimpleQA, a new factuality benchmark, with a study revealing that even GPT-4o struggles with accuracy, achieving under 40% on curated fact-based questions.

• Waymo released data showing that the company now provides over 150,000 paid trips and drives over 1 million miles each week with its autonomous vehicles.

Closing Thoughts

That’s it for us this week.

If you find any value from this newsletter. Please pay it forward and receive 5€ for every person you get to sign up in the form of an Amazon gift card. What’s keeping you ?