###### Introduction I believe there is nothing economically productive a human can do that a machine cannot. It's only an implementation problem, and the day will come when we have found the right solution. ###### Current Progress As of 2025-02-23, LLM progress is impressive. Grok 3 was released recently. I, through my own testing, believe it to be the most intelligent model in the world. In 0.3 and earlier of this current article, it was my optimistic intuition that LLMs were on a straight shot path towards "AGI". AGI here being a general intelligence comparable if not superior to human capabilities in most meaningful fields. Looking back, although I was incorrect, it was fair fit to the data I had at the time. AI progress back then seemed non-stop. There was release after release, from improved base models to reasoning models and then to o3. Benchmarks were being broken, and it seemed like there was no stopping. My intuition has now changed. The o1 models are not significantly smarter than their relative base models. Sure, they had better reasoning, and as a result, were much more polished around the edges. But they were not more intelligent nor creative. Then came deepseek-r1. Although the cost of training and the cost of inference was very impressive. The vibe of deepseek-r1 was really not that good. However, recently I have thought more about it's potential limits, I think it is clear we have seen their first limits. In my past takes, it has been obvious to me that Despite all the hype around deep research - No grounding - No modelling - Censorship - Pre-training Limits ###### Obstacles to TAF there is sufficient evidence that this AI future is possible. The only question is how long it will take, and what problems we have to solve before we get there. In this future, AI will be better than humans at most if not all productive tasks, rendering a majority of us obsolete in the work force. By then, who knows what the world will be like? Perhaps more importantly, AI will completely transform everything. Imagine millions of Albert Einsteins working towards a goal 24 hours a day, 7 days a week, devoid of human flaws. Is there anything they can’t do? Just how intelligent will AI become? What is on the other side of humanity's limits? All of our wildest dreams might come true. Science fiction might BECOME reality. We could gain unlimited abundance and truly create a utopia. If we get there. That is. It is important we focus on solving the problems[^1] that prevent us from getting to this future. At the same time we should start to hypothesize what it might look like after we have crossed the finish line. Naturally, these two topics complement each other. More precisely, the transition will not be sudden, it will be gradual and full of ambiguity. And it's shaping up to be one of the most impactful changes humanity will ever face. There is no doubt this is the single most exciting period in human history. What a time to be alive! ![[The AI Future 2025-02-11 16.32.23.excalidraw.png]] %%[[The AI Future 2025-02-11 16.32.23.excalidraw.md|🖋 Edit in Excalidraw]]%% ![[The AI Future 2025-03-22 15.38.30.excalidraw.png]] %%[[The AI Future 2025-03-22 15.38.30.excalidraw.md|🖋 Edit in Excalidraw]]%% ###### What Happens After? --- *I plan on writing many pieces about this AI Future. Thus, from now on, we will refer to **T**he **A**I **F**uture as **TAF** for short.* --- ###### Pieces [[Learning and TAF]] [[Education and TAF]] [[Ideas and TAF]] [[Coding and TAF]] [^1]: Problems are not just technical, but also social, political, philosophical, systemic and just about every type of problem ever. # ============= 2024-12-20: [[The AI Future 0.1]] 2024-12-21: [[The AI Future 0.2]] 2025-01-06: [[The AI Future 0.3]] (P) # Notes [[My Role in TAF]]