I’m Jye Sawtell-Rickson, a WIP AI researcher currently pursuing his Master’s in Taiwan with a focus on general intelligence.
Reach out to me on LinkedIn if you want to chat.
Career
Previously a Data Scientist at Meta, Rakuten and various start-ups. You can read my posts on Medium.
Spent half my career working in East Asia, the other half in the UK.
Research Interests
Research that inspires me:
- On the Measure of Intelligence (2019): If intelligence is the goal, then we need to be able to define. In short, Chollet describes intelligence as the ability to acquire new skills efficiently. This is a long paper with a lot of great content. It also includes the Abstract Reasoning Corpus (ARC) which is a tough challenge that AI couldn’t surpass (though by 2025 we’re getting closer).
- Situational Awareness (2024): the most comprehensive prediction of and guide to the near future in a world where AGI continues on its rampant growth. I loved reading this and all the linked research. It meaningfully changed the way I think about many macro-level decisions.
- The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery 🧑🔬 (2024): while they may over-state the quality of the output, it’s amazing to see the end-to-end pipeline built out and what’s around the corner.
- DreamCoder: Growing generalizable, interpretable knowledge with wake-sleep Bayesian program learning: the ability to reason is crucial for an AGI, and for now, our technology lacks the ability to do complex reasoning. The proposed DreamCoder algorithm combines traditional search methods with neural networks to get the best of both. It’s a great paper and I hope to see more work like this.
- Mastering Diverse Domains through World Models: my first exposure to world model learning.
- OpenAI’s o1, o3: There are a lot of great LLMs, but the o series seems to be the first with strong reasoning capabilities. A lot is secret, but whatever it is they’re doing, it looks good.