About This Week’s Covers
This week’s cover depicts humanity’s last New Year’s Eve through the eyes of an unknowing child. I was going for the feeling from the final scene in “Meet Joe Black” (10/10 movie no matter what people say). The prompt was crafted with Claude 3.5 Sonnet and the image generated in MidJourney.
The rest of the covers were created using Claude 3.5 and the Ideogram API, with the theme ‘humanity’s last New Years Eve + category name.’
This Week’s Executive Summaries
- I highly recommend trying Claude. I’ve completely switched.
- Ethan Mollick writes about organizations getting completely behind in 2025 because AI developments are moving too quickly. This is my biggest fear. Culture can’t keep up with technology. That was pretty much the call out in Kurzweil’s “The Singularity Is Near” (2005) as well as Bill Joy’s “Why The Future Doesn’t Need Us” (2000)
- I spoke with a dozen news professionals this week and they are using AI to write code. This includes programmers and data analysts: “40% increase” in capability. “Over 100 hours of coding-time saved” per month.
- The death of the web browser needs to be taken seriously. Good article below.
- The new open source model DeepSeek is the story of the year so far.
Personal note: I’ve completely switched to Claude from ChatGPT. This article is worth a read and this podcast is worth a listen:
“Why Anthropic’s Claude Is a Hit with Tech Insiders – The New York Times”
https://www.nytimes.com/2024/12/13/technology/claude-ai-anthropic.html?unlocked_article_code=1.hk4.3jl5.A0jpWkmgaIff&smid=url-share
Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast #452. Amanda Askill – Claude’s personality coach – appears at the 2-hour and 42 minute point.
https://lexfridman.com/dario-amodei-transcript
Wise words from Ethan Mollick:
“Easy prediction for 2025 is that the gains in AI capability will continue to grow much faster than (a) the vast majority of people’s understanding of what AI can do & (b) organizations’ ability to absorb the pace of change. Social change is slower than technological change.”
https://twitter.com/emollick/status/1874431948766208374
AI Pioneer Warns of Rising Extinction Risk from Advanced AI Systems
Geoffrey Hinton, Nobel laureate and former Google AI researcher, now estimates a 10-20% chance that artificial intelligence could lead to human extinction within 30 years – an increase from his previous 10% estimate. Speaking on BBC Radio 4, the renowned computer scientist explained that AI development is progressing “much faster than expected” and compared future human-AI relationships to that of toddlers interacting with adults. Hinton emphasized that profit-driven corporate development of AI systems isn’t enough to ensure safety, calling for increased government regulation to manage the risks of creating machines smarter than humans within the next two decades.
Theguardian
Software Developers May Become A Rare Species
“The average IQ of software developers will increase by several points over the next few years. This is just a theory, but I think AI will significantly decrease the number of developers over time. To stay employed, you’ll need to be smarter and work harder than ever.”
https://x.com/svpino/status/1874085600296657325
AI Set to Become the New Computer Interface, Replacing Traditional OS
Note: I’ve been on record saying “Death of the Browser” since September 2023 – and I stand by this prediction!
Tech venture capitalist Tomasz Tunguz predicts that AI will become the next computing interface, marking a shift as significant as the evolution from command lines to graphical interfaces. Instead of switching between applications or manipulating spreadsheets directly, users will simply tell AI assistants what they want to accomplish – whether it’s analyzing data, drafting emails, or conducting research. Tunguz is already experimenting with this future by running open-source AI models locally on his computer. Be ready for a future where AI acts as an invisible layer managing all computer tasks, potentially replacing traditional operating systems and changing how we interact with technology.
Tomtunguz
OpenAI Defines AGI Success With $100 Billion Profit Target, Documents Reveal
Microsoft and OpenAI have privately defined artificial general intelligence (AGI) in surprisingly concrete terms – a system capable of generating $100 billion in annual profits, according to leaked documents. While OpenAI publicly describes AGI as AI that can outperform humans at most valuable work, this financial benchmark reveals the company’s more business-focused internal metrics. The definition appears in a 2023 agreement between the tech giants, with OpenAI’s nonprofit board holding the power to officially declare when AGI is achieved. Despite projecting $44 billion in losses through 2028, CEO Sam Altman remains optimistic about reaching AGI “sooner than most people think,” as the company explores a transition into a for-profit structure.
Businessinsider | techcrunch
Tesla Rolls Out Latest Self-Driving Update, But Only for Newest Hardware
Tesla has launched its latest Full Self-Driving (FSD) version 13.2.1, but exclusively for vehicles equipped with its newest Hardware 4 computer, marking a shift from earlier promises. While CEO Elon Musk teases that it will “blow minds,” the release comes after several delays and with reduced features compared to initial announcements. Notably, owners of cars with older Hardware 3 computers – who were previously promised priority access – will instead receive a different version with limited capabilities, suggesting Tesla’s older hardware may be reaching its technical limits despite not yet achieving true autonomous driving.
Electrek
DeepSeek Releases Groundbreaking Open-Source AI Model That Rivals Top Performers
DeepSeek has launched V3, a major advancement in open-source AI that’s making waves in the industry. The new model processes text 3x faster than its predecessor at 60 tokens per second and outperforms other open models like Llama and Qwen (a top story of the month in itself). What makes this release particularly significant is that it’s the only freely available model among the top 10 performers, achieved with modest training resources costing around $10,000. The model uses a more efficient approach, activating just 37 billion parameters per task while having access to a much larger 671 billion parameter knowledge base. This means better AI performance that’s accessible and practical – a huge step forward in democratizing advanced AI technology (losing control?)
Tom_doerr | venturebeat | vllm_project | rohanpaul_ai | reach_vb | madiator | deepseek_ai
AI Visuals and Charts: Week Ending 01/03/2025
Simple AI – A fun product demo you should try
“The coolest AI app I have ever gotten in my iPhone: @simpleailab It does all sorts of things by calling businesses. I just had it call myself and it just blew my mind. Get it and try it yourself now and let me know here what you think. And be prepared to ask out lots of” / X
Scobleizer | https://usesimple.ai/
Must-see Robotics Videos
“Unitree B2-W Talent Awakening! 🥳 One year after mass production kicked off, Unitree’s B2-W Industrial Wheel has been upgraded with more exciting capabilities. Please always use robots safely and friendly. #Unitree #Quadruped #Robotdog #Parkour #EmbodiedAI #IndustrialRobot
UnitreeRobotics
“How can @BostonDynamics compete with @UnitreeRobotics? The Chinese firm’s robot dogs cost less than 1/10th as much and, as recent videos have shown, are capable of way more nimble and speedy movement. I’m getting PTSD flashbacks to 3DR competing with DJI in drones.
Chr1sa
Fun Example of Stitching Image to Video (aka Kling, Runway) to Beat Frontier Video Models
“Tell me you didn’t use Sora without telling me you didn’t use Sora 🫣
https://twitter.com/bilawalsidhu/status/1873172303066087427
Top 22 Links of The Week – Organized by Category
Agents and Copilots
Is Claude Computer Use the End of The Digital Gig Economy? | by AI Rabbit | Nov, 2024 | Medium
AGI (Artificial General Intelligence)
“Straight shot to ASI is looking more and more probable by the month… this is what Ilya saw” / X
Are LLMs capable of non-verbal reasoning? – Ars Technica
Business and Enterprise
A Quirky Billionaire’s Unmissable Warning About the Stock Market’s Future. | by Jayden Levitt | Nov, 2024 | Level Up Coding
Google CEO Pichai tells employees ‘The stakes are high’ for 2025
Chips, Hardware, and Infrastructure
Nvidia to open-source Run:ai, the software it acquired for $700M to help companies manage GPUs for AI | VentureBeat
ByteDance Planned to Spend $7 Billion on Nvidia Chips Next Year — The Information
Ethics/Legal/Security
Jason Schreier: “I was excited to tell my kids that there’s a sequel to Encanto, only to scroll down and learn that Google’s AI just completely made this up” — Bluesky
OpenAI failed to deliver the opt-out tool it promised by 2025 | TechCrunch
Meta
File under horrible ideas? – Meta will invest in AI-generated characters and profiles to drive up engagement | Fox Business
For Llama updates please see the Open Source category
OpenAI
Kyle Corbitt on X: “A few weeks ago, OpenAI announced Reinforcement Fine-Tuning (RFT)—a new way to adapt LLMs to complex tasks with very little training data. Here’s a quick rundown of how it works, why it’s a big deal, and when you should use it. 🧵” / X – https://x.com/corbtt/status/1873864746023477482
OpenAI ‘considered’ building a humanoid robot: Report | TechCrunch
“Exciting News from Chatbot Arena❤️🔥 @OpenAI’s o1 rises to joint #1 (+24 points from o1-preview) and @deepseek_ai DeepSeek-V3 secures #7, now the best and the only open model in the top-10! o1 Highlights: – Achieves the highest score under style control – #1 across all domains
Why OpenAI’s Structure Must Evolve To Advance Our Mission | OpenAI
Daring Fireball: OpenAI’s Board, Paraphrased: ‘To Succeed, All We Need Is Unimaginable Sums of Money’
Publishing
This new AI browser will change EVERYTHING | by Tari Ibaba | Coding Beauty | Dec, 2024 | Medium
“What’s in 2025 for AI? @wschenk, AI practitioner and founder of
Robotics and Embodiment
“🚨 Apple joins the robotics race! Researchers from @CarnegieMellon and @Apple have developed a robot collision avoidance system. They called it ARMOR and they implemented it on the @FourierRobots GR-1 robot. What is ARMOR? 👘 ARMOR is a system that helps humanoid robots “see”
BYD enters humanoid robot race as global talent search kicks off
OpenAI ‘considered’ building a humanoid robot: Report | TechCrunch
Science and Medicine
Are LLMs capable of non-verbal reasoning? – Ars Technica
Atrial Fibrillation: AI being trialled to spot heart condition
Tech Papers, Training, and Development
New LLM optimization technique slashes memory costs up to 75% | VentureBeat
Leave a Reply