Hi, It’s been a while since my last post. Between finishing my thesis and struggling with topics, I haven’t felt motivated to write much.
The other reason is my struggle with topics. I have always written about the technology that interests me. An example is one of my first blog posts about longevity which I read a lot about. These topics are niche and can be explained on a high level that I feel comfortable talking about. More recently my interests shifted more towards other topics which I am just as passionate about and that also have a huge influence on how the future is shaped, but that are harder for me to write about. The biggest one is politics. I have always been a political person and it is obvious that, when you want to predict the future economics, and politics are important. While I have the background to write about AI, I do not have the expertise to write about politics (which did not stop me from making claims about it in my yearly predictions). I had similar problems in the past when I tried writing about Fusion reactors. I do not want to talk about things where I feel something else is much more qualified to write it.
Going forward I will base my posts around more technical topics and use them as a foundation to expand outwards to broader predictions about things like politics. This way I can use the areas where I have deep insights to give context to broader trends. An example would be the recent Deepseek hype. I will probably not write about it anymore since it was covered too much for my taste at this point, but this is the general idea.
I also enjoy the yearly prediction posts a lot. Recognizing general trends and patterns is much easier with a broad range of interests and it feels like the most natural use of my knowledge. I am quite pleased with my last one. My biggest prediction and the one that I am holding for the longest at this point is AGI by the end of this year. I want to give a bit more context to this prediction because I think it is the most critical one.
I believe that in a few months, we will get reasoning models that are multimodal and use a much bigger base model than the current ones. This could be GPT-5 or Claude 4 or another. These models will have most of the capabilities that a human has. They will be able to formulate plans and execute them on a timescale of hours. In many areas, they will outperform most humans significantly. It is important to note that these models will not be used at scale. The costs will be too high. At the end of 2025, we will begin to see these models integrated into bigger systems. These systems will use the models to build more complex assistants such as Alexa+ or others. They will be even more expensive and will stay an object of research until 2026. In 2026 they will come out of the lab and will probably be used to accelerate research first. The average consumer will not have access to such systems until 2027 when the prices will fall. That being said, consumer products such as ChatGPT, Gemini, etc, will become also much better and will be close to those AGI systems in many aspects. Some users will probably already call what they have available at the end of 2025 AGI. How soon the prices will fall depends mostly on the price and performance of next-gen chips such as Nvidias Rubin generation, OpenAIs custom chips, Google’s TPUv7 and so on.
I will not make any predictions later than 2027. The current geopolitical tensions make the future highly volatile. It is possible that quick process in AI will create a new world order that creates stability again, but the transition period is highly unstable and I expect more conflicts in the next few years. While the future remains uncertain, especially with growing geopolitical tensions, I’m excited to keep exploring how technology shapes the world — and to share those insights!
