MENU

introduction about me posts

INTRODUCTION

I am writing a book on the future of science in the era of general artificial intelligence. I will use this blog to share some ideas and notes while working on the book.

The book is rather philosophical. The working title is "the last science". It is a bit controversial, but the main idea is that artificial intelligence is the last human scientific endeavor. AI systems will soon be able to make nontrivial contributions to science by generating new knowledge. These future systems will possess cognitive capabilities that are strictly greater than those of human brains (c.f. prime number mazes for an analogy with rats). Furthermore, they will be able to make connections and find patterns in data so large that it will be far beyond what collective human intelligence can analyze. When thinking about the scale of the universe, there is no reason, even assuming Occam's razor, that the laws of physics have to be so simple that they fit in the context window of a human mind. Because of that, human scientists will become passive observers instead of contributors. AI will discover new laws that will be impossible to even explain to a human mind, and we will enter the era of magic and alchemy once more. As Arthur C. Clarke once wrote: "any sufficiently advanced technology is indistinguishable from magic". We might have access to engineering that comes out of the knowledge discovered by AI, but no human scientist will understand how it works and humanity will make no further contributions to science. Even the brightest minds in the world will be left behind. The last human science is the work we are doing now on the path to general intelligence, which will then evolve into superintelligence. Therefore, in order to make progress in any area of science or even pure mathematics, it might be more prudent to work on AI instead.

The book will also cover some history of AI that is directly related to the most likely path that the field will take in the near future. I am trying to avoid going into too much mathematical and algorithmic details (except for when it is necessary to make things clear). My goal is to make it readable to a wide audience extending beyond the AI research community.

Some posts will be in form of videos that I occasionally record. Some will be podcast conversations with interesting people and some will talk about my book, blog articles, current research, or other topics related to the field of artificial intelligence. Sometimes they might be more tangential and discuss philosophy, mathematics, or broader areas of computer science.

ABOUT ME

I am an artificial intelligence scientist working in the areas of neural networks, representation learning, and computational linguistics.

In my past research endeavors, I focused on understanding the topological structure of embedding manifolds induced from deep neural models of natural language data through the lens of homotopy and homology theories.

I am currently working on new ideas involving collective intelligence, emergence, neuroevolution, and extending the capabilities of language models by combining transformers with state space models and cellular automata. Additionally, I am increasingly involved in interdisciplinary projects aiming at the development of novel psychometric techniques for AI.

POSTS

#	TITLE	LINKS
10	THOUGHTS ON AGI ASSESSMENT AND THE ARC BENCHMARK - towards a test for artificial general intelligence.	article
9	WE ARE ALL NPCS AND NOBODY IS WATCHING - the emergent simulation hypothesis.	article
8	THE UNIVERSE (LIKE THE INTERNET) IS LIKELY MOSTLY BOTS - remarks on the future of life.	article
7	AI SAFETY, LARGE LANGUAGE MODELS, AGI - A conversation with Steven Basart.	video
6	THE GRANDFATHER OF THE INTERNET - A conversation with David Farber.	video
5	SPECTRAL BENCHMARKING - Measuring general model ability with spectral methods.	article
4	SWARM INTELLIGENCE - Some questions and remarks regarding behavior of complex systems.	article
3	THOUGHTS ON BIOLOGICALLY INSPIRED AI - Some brief remarks on whether we need to understand biological systems in order to spawn AGI.	article
2	BOOK COVER - I finally designed the cover for the book that I am satisfied with. It connects the ideas of manifolds and neural networks, which will be covered in depth throughout THE LAST SCIENCE	article
1	MONAD THEME - A minimalist dark and light colors for focused immersion.	article
0	THE SHAPE OF WORDS - These are some older articles I wrote while working on topological aspects of language models. I made significant progress on this topic since then, but I have not had time to compose more articles in that thread. They might however be useful as an introduction for those who wish to collaborate with me on other papers descending from this project.	project