Will It All Come Crashing Down?

Good morning.

If you’ve been following this newsletter for a little while, you know there are two recurring themes in this introduction to the newsletter.

One of them—like last week—is my call to you not to fall behind and to incorporate AI into your workflows and ultimately your life. I want you to embrace the hype.

The second overarching theme is the polar opposite of that: a growing distrust of the hype.

The dilemma that i’m struggling with at the moment is that both sentiments are valid. There is a case to be made that this technology will shape the world in ways we cannot yet comprehend. Selfdriving cars , a robot as your personal butler and time to chill and relax on universal basic income as an AI does all the work.

But there is also a case to be made for having been sold an unattainable dream. That we were lied to and are potentially looking at fraud at the highest scale imaginable.

If you want to delve deeper into the bear case against AI,. I present you this fantastic blogpost by Ed Zitron (it’s also fantastically elaborate). If you have the courage to read through the entire thing, it will leave you shaking.

Sorry to start about OpenAI again - but they secured a 150 billion funding round this week- which includes a 10 BILLION DOLLAR payment package for its CEO Sam Altman. Who, only a year ago, while grilled by a senate committee said he doesn’t need money and is developing his company because (and I quote) “he loves it”.

The same Ed Zitron just published a scathing takedown of OpenAI - a company that loses 2,35$ for every 1$ dollar it makes.

But it’s not just OpenAI : similar stories are emerging from Anthropic, NVIDIA,Microsoft and Google.

The Artificial Intelligence boom has taken off like a rocket. A rocket that is guzzling dollars for fuel. We were going for the moon (AGI) and beyond (ASI). Now it seems we cannot get out of our atmosphere without more fuel.

The rocket is close to toppling over and crashing back to the ground, leaving just a giant hole in the earth.

We are nearing a crucial time.

Will the AI boom come down with a huge ‘boom’ or will we fly to the stars?

Time will tell.

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

• Google DeepMind has unveiled AlphaChip, an AI system that designs computer chips using reinforcement learning, reducing design time from months to hours. AlphaChip has been used to create Google’s Tensor Processing Units and will be shared for further research. Its ability to design better chips could significantly accelerate AI development through a self-reinforcing cycle.

• Google's NotebookLM now supports YouTube videos and audio files, making it easier to analyze and summarize multimedia content alongside text. It also introduces a feature for sharing AI-generated audio summaries, enhancing collaboration and study. These updates make consuming complex information quicker and more efficient.

• Japanese archaeologists, with help from IBM’s AI, discovered 303 new geoglyphs near Peru’s Nazca Lines, nearly doubling the number of known figures. AI and drones accelerated the discovery process, completing a century's worth of work in six months. This finding shows how AI can dramatically improve archaeological research.
  

• Google has released two upgraded Gemini 1.5 models: 1.5-pro-002, which excels in math benchmarks, and 1.5-flash-002, which improves instruction following. These updates also include reduced pricing, higher rate limits, and enhanced features, enabling developers to tackle more complex problems and innovate faster. Gemini’s progress in math and coding marks a significant leap for AI development.

• The models also offer a generous free tier and resources to help beginners start building AI apps quickly. With a 2M token context window, Gemini 1.5 can process long videos, books, and other extensive content, creating new possibilities for real-world applications. Looking ahead, AI systems may become more proactive, offering assistance without needing prompts, marking the next evolution in AI interactions.

• California Governor Gavin Newsom vetoed S.B. 1047, a bill that would have required safety testing for AI models and held companies liable for damages over $500M. Tech giants like Google and OpenAI opposed the bill, fearing it would hinder innovation, while supporters included Elon Musk and AI expert Geoffrey Hinton. Newsom acknowledged the bill’s intentions but plans to consult further on AI safety regulations.

• Liquid AI unveiled new Liquid Foundation Models (LFMs), which outperform traditional transformer-based models like Meta's Llama 3.2. These models, featuring a novel architecture based on dynamical systems, achieve state-of-the-art results while requiring less memory, particularly in long-context tasks.

• OpenAI's DevDay 2024 introduced several new API features aimed at making AI tools more accessible and efficient for developers. Highlights include a Realtime API for speech-to-speech apps, Model Distillation for easier fine-tuning of smaller models, Prompt Caching to reduce costs, and Vision Fine-Tuning for image and text tasks. These updates lower the barrier to entry and open up new possibilities for AI-driven experiences.

• Microsoft announced major AI upgrades for its Copilot assistant, including voice interaction, web content understanding, and enhanced reasoning capabilities. The return of its "Recall" feature, with better privacy measures, and personalization improvements aim to make Copilot more intuitive and adaptable. These updates push Microsoft closer to creating an AI that acts more like a personal agent.

• OpenAI's Olivier Godement shared insights on the new Realtime API, emphasizing its potential for customer support and education. While the API's voice functionality currently has a high cost, significant price drops are expected.

• OpenAI closed a $6.6B funding round, raising its valuation to $157B, with Thrive Capital, Microsoft, Nvidia, and others participating. The funds will be used to expand research, increase computing power, and develop new tools, with OpenAI projecting up to $100B in revenue by 2029. See the intro for the bear case against that.

• Google is advancing AI models with reasoning capabilities similar to OpenAI’s o1 system, using chain-of-thought prompting to solve complex problems. While Google takes a cautious approach, it has introduced math-focused reasoning models like AlphaProof. The competition for developing AI with human-like reasoning is heating up between the major players.

• MIT researchers have developed "Future You," an AI system that simulates a user's older self for interactive conversations. Users provide personal data to create a realistic future self, which has shown to reduce anxiety and negative emotions in studies. This development highlights AI's potential in personalized psychological support and therapy.

Quickfire News

• AstraZeneca partnered with Immunai, paying $18 million to use the biotech firm's AI model of the immune system to improve cancer drug trial efficiency.

• Visa agreed to acquire AI-driven payments protection firm Featurespace to boost its financial crime and fraud detection capabilities; the acquisition price was not disclosed.

• Runway launched The Hundred Film Fund, offering grants ranging from $5,000 to $1 million for filmmakers incorporating AI into their projects.

• Microsoft announced a $1.3 billion investment in Mexico over the next three years to improve AI infrastructure and provide skills training.

• Blackstone confirmed a $13.3 billion investment to build an AI data center in northeast England, creating 4,000 jobs, including 1,200 in construction.

• Hugging Face reached 1 million free public AI models on its platform, emphasizing the shift toward specialized models for various use cases instead of a single dominant model.

• Google agreed to invest $1 billion in Thailand to expand AI and cloud infrastructure in Southeast Asia, focusing on building new data centers due to increasing regional competition.

• ByteDance, TikTok's parent company, is reportedly planning to develop a new AI model using Huawei chips, diversifying from U.S. suppliers like Nvidia to address export restrictions.

• Artisan AI secured $7.3 million in seed funding for its sales-focused AI virtual employees, with its first AI assistant, Ava, already supporting over 120 companies.

• Luma Labs upgraded its Dream Machine AI video model, enabling full-quality video generations in under 20 seconds.

• Qodo announced a $40 million funding round for its AI-powered code testing software, aiming to expand services and focus on larger enterprise clients.

• AI reading coach startup Ello launched "Storytime," a feature allowing kids to create personalized stories using AI.

• OpenAI founding member Durk Kingma announced he is joining Anthropic, reuniting with former OpenAI employees and emphasizing the company's mission of responsible AI development in a post on X.

• Pika Labs launched Pika 1.5, a video generation model upgrade with enhanced effects, realistic movement, longer clips, and cinematic capabilities.

• Anyscale unveiled major upgrades to its AI platform at Ray Summit 2024, including a GPU-native Ray architecture, RayTurbo for improved performance, and Ray Data for unstructured data processing.

• U.S. AI chipmaker Cerebras officially filed for an IPO, with a valuation expected between $7-8 billion. The company is backed by Sam Altman and competes with Nvidia.

• Meta released the open-source code and developer suite for Segment Anything Model (SAM) 2.1, an upgraded tool for image and video segmentation.

• Nvidia introduced NVLM 1.0, an open-source family of multimodal models achieving state-of-the-art performance on vision-language and text tasks.

• Pinterest launched Performance+, a suite of new AI tools for advertisers, featuring the ability to create background images for products and automation features for ad campaigns.

• Microsoft announced a $4.8 billion investment into AI and cloud infrastructure in Italy, aiming to expand its data center to become one of Europe's largest cloud hubs.

• Character AI is reportedly shifting its focus away from building AI models after its $2.7 billion deal with Google, now prioritizing its consumer chatbot service.

• Elon Musk posted "OpenAI is evil" on X, reacting to reports that OpenAI asked investors to avoid funding rival AI firms like Anthropic and Musk's xAI.

• Accenture announced a partnership with Nvidia to speed up enterprise AI adoption, launching a business group and AI Refinery platform to scale AI systems across industries.

• The Cancer AI Alliance launched a $40 million collaboration between major medical institutions and tech companies like Microsoft, AWS, Nvidia, and Deloitte to advance AI-driven cancer care.

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 ?