Are You Stealing From Your Boss?

Good morning,

This was a big week for Generative AI: Amazon entered the game and all text-to-video models got upgrades. We learned about a new type of generative AI: from 2D drawings to 3D worlds.

Previously in this newsletter, we discussed the rumours that OpenAI might be running out of funds. This week, they informed us that they are considering introducing advertisements into ChatGPT and are launching a 'pro' model of O1. This will cost you 200€ per month.

I’m thinking about doing an “Anti -AI ” newsletter where I look at AI from the viewpoint of their biggest critics and have fun taking the side of the non-believers. Let’s do that next week!

For this week : it’s time to tackle a “sensitive subject.”

As we previously reported, individuals are gradually realizing the significant advantages that generative AI can offer in their daily tasks. In some industries, like digital marketing or the advertising sector, up to 60% of people are actively using it.

BUT (and it’s a big but)—only 5% of companies are aware that the content their employees generate has been “enhanced by AI.”.

If you combine this with the “working from home” trend, there’s a lot of “ShadowAI” going on.

This is where the sensitivity begins.

Technically, when you’re working for a company, you sign a contract in which you sell hours of your own precious time in exchange for money. I haven’t seen an employment contract ever that was results-based.

If you start to use generative AI to do your job, and let’s assume it does the job not only spectacularly better, but in a fraction of the time it would have taken you without AI, what happens then?

Let’s say you become twice as productive as before the advent of AI. Are you allowed to take afternoons off?

Isn’t that the promise of artificial intelligence? That humanity can lay back while the robots do our bidding?

Or are you in fact stealing time from your employer? Because technically you could now become employee of the month, the year and the decade based on your results.

It’s a difficult subject, and obviously employers would look differently upon that than employees.

I created an anonymous poll. What do YOU think about this issue?

Welcome to the Blacklynx Brief!

Like newsletters?

Here are some newsletters our readers also enjoy:

AI News

• A big player has just entered the game. Amazon unveiled Nova, a new family of AI models for text, image, and video generation, marking its largest consumer-facing AI push. The lineup includes competitive text models (like Nova Pro, which rivals GPT-4o and Llama 3), as well as Canvas for image creation and Reel for six-second videos, with plans for multi-minute video and multi-modal capabilities by 2025. With its vast reach and resources, Amazon is bound to challenge the likes of OpenAi for the top spot on the AI leaderboard.

• OpenAI has kicked off their advent-calender style “12 days of OpenAI”. The first announcement is the release of “o1 pro mode”. An extra powerful version of o1 aimed at researchers, engineers and data scientists. Price tag is a hefty 200$ per month. This adds fuel to the fire that OpenAI is getting desperate for funds.

  

  Gains for pro mode seem marginal

• As mentioned, Amazon is reportedly set to unveil Olympus, a new AI model specializing in advanced video and image processing, as early as next week. The model excels in tasks like tracking basketball trajectories or diagnosing underwater equipment issues, targeting niche applications like sports and industrial analysis. While not yet a leader in text generation, Olympus reflects Amazon’s dual strategy of partnering with Anthropic while pushing in-house AI development into untapped markets.

• Tesla demonstrated significant upgrades to its Optimus humanoid robot, including a new hand-forearm system capable of catching balls in real time. The system offers 22 degrees of freedom in the hand and improved tendon controls, though the upgraded mechanics add weight.

• Elon Musk has filed a preliminary injunction to block OpenAI's transition to a fully for-profit structure, escalating the legal battle between Musk and the AI leader. The injunction accuses OpenAI of abandoning its non-profit roots, engaging in "self-dealing," and discouraging competitor investments through restrictive terms. OpenAI called the move “baseless,” but with its $150B valuation at stake, this legal action could complicate its fundraising and future plans.

• Google DeepMind introduced Boundless Socratic Learning, a framework enabling AI systems to self-improve through language-based interactions without needing external data or human feedback. Using "language games," the AI generates its own training scenarios, evaluates its performance, and potentially modifies its code for open-ended learning.

• Adobe unveiled MultiFoley, an AI system that generates synchronized sound effects for videos using text prompts or reference audio. Trained on professional libraries, MultiFoley creates high-quality, 48kHz audio that matches on-screen actions with 0.8-second synchronization accuracy. The tool streamlines sound design, allowing users to create or transform sounds effortlessly, marking a significant leap forward for post-production and creative workflows in video and film.

• World Labs, led by AI pioneer Fei-Fei Li, unveiled its debut project: an AI system capable of turning any image into an explorable 3D environment accessible through a web browser. The tool generates immersive spaces with consistent visual elements, real-time camera effects, and interactive lighting. This technology could redefine industries like gaming, virtual experiences, and filmmaking by making 3D world creation as simple as generating 2D images today. On the same day , Google DeepMind unveiled Genie 2, a multimodal AI that transforms single images into interactive, physics-based 3D environments with real-time player controls.

• OpenAI is reportedly considering advertising within its AI products to offset its $5B annual costs, despite CEO Sam Altman previously opposing the idea. Key hires from Google and Meta suggest serious exploration, though CFO Sarah Friar emphasized no immediate plans. While ads could provide a critical revenue boost, their introduction risks altering the AI experience and user trust, particularly in a market where reliability is paramount. Did we mention that the signs are there they are running out of funds ?

• Hume AI launched Voice Control, a system enabling developers to customize AI voices with 10 sliders for traits like confidence, assertiveness, and enthusiasm. The tool isolates each quality for precise adjustments, maintaining consistency across applications.

• Tencent launched HunyuanVideo, a 13B parameter open-source AI video generation model that surpasses leading closed systems like Runway Gen-3 in quality. The model supports text-to-video, image-to-video, avatar creation, and synchronized audio, offering coherent motion and scene transitions. By open-sourcing this technology, Tencent accelerates innovation in AI video and lowers the barrier for both research and commercial applications.

• Search startup Exa introduced Websets, a novel search engine that uses AI embeddings to transform the web into a structured database. Unlike keyword search, Websets captures the meaning of content, excelling at complex queries like finding niche datasets or specific companies.

• Sam Altman hinted at accelerated AGI timelines during the NYT DealBook Summit, forecasting first glimpses by 2025 and emphasizing a subtle immediate impact but profound long-term changes. He shared major ChatGPT metrics, including 300M weekly users.

• DeepMind also debuted GenCast, an AI weather system delivering 15-day forecasts faster and more accurately than traditional supercomputers. Trained on decades of historical data, GenCast provides near-instant results and excels at predicting extreme events, outperforming the world’s best forecasting models. This breakthrough could redefine weather prediction, ensuring quicker, more reliable data for scientists and decision-makers.

Quickfire News

• ByteDance is suing a former intern for $1.1 million, alleging deliberate sabotage of its AI language model training infrastructure by manipulating code.

• Databricks is reportedly raising at least $5 billion at a $55 billion valuation, with the funding round focused on employee cash-outs and possibly delaying IPO plans.

• Google Labs launched GenChess, a web experiment using Gemini Imagen 3 to let users design custom chess pieces through AI image generation.

• OpenAI filed a trademark for its o1 "reasoning" models, revealing an earlier and unusual application in Jamaica months before the model's announcement.

• Mistral AI announced the Mistralship Startup Program, offering selected startups 30,000 platform credits, dedicated support, and early access to models over six months.

• Meta’s Chief AI Scientist Yann LeCun revised his AGI timeline, suggesting human-level AI could be achieved within 5-10 years, aligning with similar predictions by Sam Altman and Demis Hassabis.

• AI image startup Black Forest Labs is reportedly in talks to raise $200 million in a funding round led by A16z, which would value the four-month-old company at over $1 billion.

• Canadian media giants filed a joint lawsuit against OpenAI, accusing the company of copyright infringement for using their news content to train its AI models.

• Meta is planning a $10 billion subsea cable system spanning over 40,000 kilometers to support increasing internet traffic and its AI initiatives.

• OpenAI’s Policy Frontiers lead Rosie Campbell announced her resignation, citing "unsettling shifts" and cultural changes within the company.

• A study by WIRED revealed that over half of longer English posts on LinkedIn are now AI-generated, reflecting the platform’s growing adoption of AI writing tools.

• A new AI-powered app called Death Clock predicts individual death dates using data from longevity studies with 53 million participants, factoring in diet, exercise, and stress levels.

• Cohere released Rerank 3.5, an AI search model with advanced reasoning capabilities, support for 100+ languages, and improved accuracy for enterprise data searching across documents, code, and more.

• The Browser Company teased Dia, an AI-integrated smart web browser featuring agentic actions, natural language prompting, and built-in AI writing and search tools.

• The U.S. Commerce Department introduced new chip restrictions aimed at curbing China’s AI advances, blacklisting 140 entities and expanding controls on high-bandwidth memory chips.

• AI chip startup Tenstorrent raised $700 million in a funding round led by Samsung and backed by Jeff Bezos, valuing the Nvidia competitor at $2.6 billion.

• Nous Research launched a distributed AI training effort, pre-training a 15-billion-parameter language model over the internet with live-streamed updates.

• Amazon Web Services announced significant data center upgrades to support next-gen AI chips and genAI workloads, including new liquid cooling systems and enhanced electrical efficiency.

• ElevenLabs launched Conversational AI, a tool for adding voice capabilities in 31 languages to AI agents, offering ultra-low latency, LLM flexibility, and advanced turn-taking features.

• Google announced that its VEO video generation model is now in private preview on the Vertex AI platform, with the Imagen 3 text-to-image model becoming available to all users next week.

• OpenAI appointed former Coinbase CMO Kate Rouch as its first Chief Marketing Officer to oversee marketing strategies for consumer and enterprise products.

• Hailuo AI introduced l2V-01-Live, an AI video model that animates 2D illustrations with smooth, lifelike motion.

• Amazon released Automated Reasoning checks on its Bedrock platform to validate AI responses against customer-provided data, along with Model Distillation and multi-agent collaboration features.

• Meta revealed in a blog post that less than 1% of fact-checked misinformation during the 2024 election cycle involved AI content as part of its broader election integrity efforts.

• European defense tech startup Helsing unveiled the HX-2 attack drone, featuring AI-enabled autonomous capabilities, with plans for mass production at lower costs than current systems.

• Amazon and Anthropic unveiled Project Rainer, an AI supercomputer powered by hundreds of thousands of Trainium2 chips, projected to be five times larger than Anthropic’s current top model and the world’s largest AI system.

• OpenAI announced the hiring of three prominent Google DeepMind computer vision experts—Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai—who will focus on multimodal AI and establish a new Zurich office.

• Luma AI introduced Ray 2, a next-gen video model capable of generating minute-long videos in seconds, and announced its integration into Amazon’s Bedrock platform during the AWS event.

• Defense tech firm Anduril partnered with OpenAI to develop AI-powered aerial defense systems designed to protect U.S. and allied forces from drone threats.

• Spotify launched its viral "Wrapped" listening recap, adding an AI-powered podcast feature with music commentary provided by two AI hosts using Google’s NotebookLM tool.

• EvolutionaryScale debuted ESM Cambrian, a new family of protein language models trained on Earth’s protein sequences, delivering breakthrough results in protein structure prediction.

Closing Thoughts

That’s it for us this week.

If you find any value from this newsletter. PLEASE FOR THE LOVE OF GOD pay it forward!

Thank you for being here !