Sam Altmans Reflections, Nvidia CES Keynote and AI as a Parasitic Cancer – Live and Learn #59

✨ Quote ✨

Life is good if it squishes nicely when you poke it.

– Sasha Chapin - (source)

🖇️ Links 🖇️

Reflections by Sam Altman. OpenAI has announced some crazy stuff this year and is working on even more crazy things in the future. In this post, Sam Altman reflects on where they have come from and also on where they are headed. The company is growing at a tremendous rate and they are re-structuring constantly. Some of what Altman writes is surely marketing hype in order to prepare for the next round of massive fundraising. But even considering this, the audacity of this post and the claims it makes is nuts. I mean just listen to this: "We are now confident we know how to build AGI as we have traditionally understood it. We believe that, in 2025, we may see the first AI agents “join the workforce” and materially change the output of companies." OpenAI knows that people will think they are crazy, but people thought they were crazy back in 2015 as well. And they will try to prove people wrong. Again. Let's see what happens and if OpenAI will achieve its mission of AGI for everybody.

Generative AI is a Parasitic Cancer by Freya Holmer. Search and more generally the internet as a whole is starting to break down due to the flood of AI-generated content slop. Slop is bad, meaningless, auto-generated stuff—content without a "soul" that doesn't answer any detailed questions or give the reader any real value. But this type of content performs well in search engines because it can be mass-manufactured and engineered to rank for keywords and game Google's algorithms. The hopelessness Freya tries to convey in the video is palpable. The destruction of human-generated beauty is happening. It's like AI is eating things that are dear to her—art, computer games, programming, tech communication. AI makes them worse, stripping out the human component, the stories of the people behind the art, while robbing many artists of the opportunity to create for a living. Some of Freya's points remind me heavily of this video on the AI dilemma, focusing on the risk of fake information and the erosion of trust in digital content. If you doubt that an artwork is AI generated or real, you are experiencing this already. I featured this video back in edition 21 of the Live and Learn Newsletter, which feels like an eternity ago. Yet it has been only a bit over a year. What crazy times.

Nvidia CES Keynote by Jensen Huang. Every Nvidia Keynote I watch makes me want to buy more Nvidia stocks. I have to be careful not to, since Nvidia might be overvalued quite hard already and I am somewhat overexposed to it in my portfolio as well. But the things they announce are always so damn good, it makes me excited for the future. And their presentations are always on point—with great visuals, music, and storytelling. The production value of their keynotes is off the charts. But I realize that this is also part of the deal. These keynotes are specifically designed to show off, to build hype, and to fuel the growth of their stock and the AI mania. This time, the three big announcements, to me were, first Project Digits, a supercomputer built into the form factor of a small desktop PC, comparable to a Mac Mini but ready for AI workloads and even training runs. Second, their Cosmos World Model, designed to help robots encounter all kinds of diverse environments digitally. This is an effort to extend the amount of training data by orders of magnitude simulating diverse and varied scenarios and placing algorithms in this virtual sandbox world for training. And, finally third, the RTX 50 Graphics Card Series, Nvidias latest update to its gaming GPU segment. Even though most of their revenue these days (around 80%-ish) comes from the data center market, it's nice to see that they still produce kickass graphics cards for gamers. They are pushing hard to infer pixels with AI instead of using more computationally expensive ray tracing. AI is predicting the pixels instead of explicitly calculating them. And the results are absolutely stunning. They are also planning to release a laptop version of this, which makes me excited.

Things We Learned about LLMs in 2024 by Simon Willison. This blog post dives deep into all the things that happened in 2024 in the world of LLMs. It's filled with links to other thought pieces and Simon Willison is quickly becoming one of my favorite sources for anything LLM evaluation-related. To sum up the biggest trends of his analysis: GPT-4 class models can be run locally on consumer hardware, vision capabilities are everywhere, 2025 will bring video, AI slop is widespread and running amock in the internet, the energy consumption of AI is skyrocketing, and the knowledge around AI is incredibly unevenly distributed, which presents its own unique problems. He talks about many more things and the whole post is worth reading to get a solid overview of what happened in AI in 2024 and where the whole field is headed.

🌌 Travel 🌌

The last two weeks passed by so quickly, and yet, to me, they feel more like half a year than they feel like two weeks. Cycling completely alters the perception of time somehow. I have seen so many beautiful places in such a short amount of time; it's insane. My phone is filling up with photos quickly while traveling around, and I meet interesting people every day and just live life to the fullest.

Still, I was often cursing while going up another stupidly steep mountain, straining to just make it. But in the end, all of this, the effort, the sweat, the discomfort, the flat tires—everything—is part of the adventure, part of the experience. And I wouldn't want to miss any of it. In other words, life right now squishes quite nicely when I poke it 😇

Also, I'm trying to catch up on editing photos and writing for the travel section on this blog and will put up some diary-like posts of the adventures with the bicycle there soon. Until then, enjoy a "small" selection of pictures, there are lots and lots more to come 🤗

🎶 Song 🎶

Excursions by C418

Youtube Music | Spotify

---

That's all for this time. I hope you found this newsletter useful, beautiful, or even both!

Have ideas for improving it? As always please let me know.

Cheers,

– Rico