Just Thoughts #40: The AI Wars
We cover some thoughts on the ongoing rivalry between the AI giants, and give you a short update on startup progress, ending with social media signs.
Welcome reader! Just Thoughts is for modern-day leaders who want to think better about business, leadership, and entrepreneurship. This article's Content is “handwritten” (highlighted when not) and co-piloted with Grammarly’s spell-checking and rephrasing for tone of voice.
Headlines this week:
The AI Wars
Story World Progress
Social Content Highlight
The AI Wars
So, what do you think about DeepSeek? I don’t know how many times people have asked that question this past week, and rightly so. Not every day do you see 600B in the market cap melting away.
Many quality analyses are made here on Substack on why, what, and how Deepseek affected the world. On the humor side, Deepseek allegedly copied ChatGPT, and who’s surprised the Chinese copied something? Albiet, this must be the first time something was copied and had that impact.
Ultimately, the issue is who controls the narrative of history. With the help of AI, we can easily recreate historical events to bring new life to old memories, but we can also control how history is presented. The internet is flooded with examples of how Deepseek cannot comment on human rights violations in China but can list every detail of the crimes committed by the U.S. I haven’t tried. Still, I bet it’s the other way around on the U.S. models. Who is surprised by this fact? No one, really; when you give it the proper perspective, history is rarely written by the losing side.
I recently watched Fly Me To The Moon, a fictitious romantic comedy-drama that quite well captured this same political strategy game. The space race was a political power game involving who controlled the “global superiority” narrative. The stakes of the first moon landing were high, so the movie tells the story of how allegedly a backup plan was made for the moon landing; it would be staged to look like a success even if it wasn’t.
I don’t think we can fake the creation of superior LLMs in today’s interconnected world, but you get the point.
Another aspect to consider with DeepSeek’s launch is that the consequence of smaller models is that the energy demand should decrease. I thought energy supply would be a big issue with the rise of the AI arms race, and with this recent development, someone against nuclear solutions would say, “The Chinese proved we don’t need to meet this demand caused by your large models.”
However, I recently got a hold of a macro trend report by Contrary, and it seems developed countries like the U.S. aren’t increasing their demand for energy as electricity per capita. Still, it is growing globally, driven mainly by the growing electrification of Asia.
At first reading, I thought we, for some reason, would see a decrease in energy demand, but that isn’t the case. What we could challenge, in light of what Deepseek proved, is the CAGR of total energy demand in the future if large data centers cause it.
These slides and the discussion have mainly revolved around the U.S. and China. What about Europe? The general meme on the internet uses the fact that Europe seems to be only achieving legislation that slows innovation. Regarding innovation, the only thing we can accomplish is bottle caps that don’t fall off;
Of course, no one except PJ from Northzone understands this: it is a question of “controlling the narrative.” In the same way, the “winner of the AI wars” will be able to control the narrative. The way to fight the narrative is not to give in to the fact that the only thing we can do in Europe is create legislation and that it is harmful in every single way. It may also be a way to create a more sustainable future, and as the good people in the EU Inc. initiative are proving where there is a will to simplify, there is a way to do it.
With the Macro political discussion covered, what does all this mean for you as a consumer or a business builder? First, I’d say use products that allow you to toggle between models. Chatly's app is a good consumer example highlighting these options. Perplexity might be another, although it’s unclear which model it’s running on when you get on the lander.
Future writing could discuss models' ideological categorization, such as whether they use your input to train their models by default. Or do they not? Additionally, what data are they using or have used to train their models, and are they currently phased with what kind of lawsuits they are handling (if public)? Furthermore, use case labeling in layman’s terms would likely be in order. You don’t need the most powerful model to solve your specific problem. A simple “tell me which model I should likely use for problem X” would suffice.
For business builders, I’m almost leaning towards Google’s models for proof of concept building, just because they’ve been least in the cross-hairs lately. They have the power and the support structures for startups. However, there is a reason why Together AI is the fastest-growing startup of all time, with 100M+ ARR in less than 10 months, as highlighted by
;Emerged in 2022, founded by a visionary team comprising Vipul Ved Prakash, Ce Zhang, Percy Liang, and Chris R. Their collective ambition was to establish a robust platform that democratizes access to generative AI technologies.
This initiative was driven by the recognition of a growing need for open-source solutions in an industry increasingly dominated by proprietary models. By focusing on creating a comprehensive AI platform, Together AI aims to empower developers and organizations to harness the potential of generative AI without the constraints of vendor lock-in.
It is maybe the fastest-growing startup of all time, faster than Wiz(in 18 months) and Deel (in 20 months), with only 85 people.
Proving the point made in Just Thoughts 24 about LLMs as operating systems;
If you’re wondering who will be the dominant player ten years from now, I’d put my money on Llama, Open Source, and Zuck. Why? because of the safety implications, as described by Zuck. Additionally, open source is significantly more efficient regarding value creation and has an ever-increasing understanding of how to capture it, as defined by Joseph Sacks from OSS Capital.
It makes me wonder who is rivaling Chat GPT’s and Google's startup programs with an Open-Source equivalent. 🤔
Nonetheless, we chatted with Nick Dorra about the things to consider when building Story World, and he’s listed the just for the creative industry quite well in his blog 1+1;
When considering Generative AI tools, the legal concerns fall into three key areas:
1.) The Inputs
2.) The Outputs
3.) The Training Data
Each has their own ramifications, so let’s take a closer look.
You need to be concerned about how your model, or the model you are using, handles these three points. Read his blog for more thoughts on how to approach each section.
Story World Progress
We’ve had another batch of discovery discussions with leading screenwriters, actors, and producers from the local scene. The first investor discussions also helped us narrow down the first steps and highlighted the need to be even more evident about the problem/solution for the go-to-market phase.
We also concluded that we’d head into Kiuas because we need to source full-time design and tech founders, and I firmly believe that being loyal to your home-grown is beneficial both ways. That being said, here’s our job description;
Additionally, programs hope to save you thousands in the first months, with credits to free tools such as Slack (communication), Notion (intra /documentation), Canva (decks), Google (email and domain hosting), and website builders, as well as servers and LLM usage. Furthermore, after trying Firefly.ai for meeting notetaking, I like it because it’s the first notetaking tool that adjusts itself to language, has a nice UX, and is not tailored specifically for sales.
I likely should also list a few other tools, but the other benefits of a program are, of course, services, maybe office space, and a community to help you get up to speed. Others may need investor or customer networks, but we have enough to keep ourselves busy for the first few months.
Social Content Highlight
I think highlighting Elon as a Nazi, even if it’s not ok and super scary, makes him want to act more like one. Nonetheless, I liked this sign; it somehow made me feel good about being a patriot in Europe;
Also, after a couple of years of proper social media detoxing and streamlining the number of places you get inbound from, this message resonated with me;
I also think this somehow gives meaning to the idea of Story World. Even if AI can be used to win Nobel prizes in science, someone needs to be motivated to solve that problem one way or another, and how we humans have been able to make meaning out of nothing is our ability to craft and share stories. That is what our brain is designed to understand.
Using technology to tell better stories does not rob us of our humanity; I strongly believe it helps us be more humane by widening our ability to convey meaning. In an increasingly noisy world, we need the best stories to win, come to life, and stand out.
Until next time! The Next Just Thoughts will mark the 100th post from the Just Home Publication, marking one year of weekly active posting.
It will also include a revision of my yearly purpose reflection. Going into this year, I already have a crystal-clear vision of what it is, and it’s already filling me with an incredible sense of direction.
It will also not be one I invade your inboxes with; it will simply appear, and you’ll find it if you want to. Hence, I will not tell you to subscribe, just visit The Just Home if it’s of interest!
If you found this insightful, please share it with someone who would also think so!
You can find me on LinkedIn, Threads, YouTube, and X @DolencNicolas. I also have accounts on Facebook and Instagram, and I am more active on the latter than the first.