Researcher Spotlight: Clement Neo

‍

Our Researcher Spotlight series highlights the global community at the heart of Apart Research. We have so many incredible and talented researchers at Apart Research, and this series will help you to get to know them a little better. Clement Neo is currently a Research Assistant with Apart Research.

‍

Hey Clement! Tell me about your upbringing in Singapore.

"Hi! Yes I’m originally from Singapore. I’m currently a Research Assistant at Apart Research, where I focus on technical AI safety and mechanistic interpretability. I grew up with a keen interest in tech and got my first computer when I was just three years old! My dad is an engineer, and that had a big influence on me early on."

“I almost went to film school because I was really into making documentaries, but AI research became a more compelling field for me around 2018 when things started taking off. So, I made the shift and never looked back.”

‍

How did you first get interested in coding?

"I actually started coding during my National Service after high school. My work in systems administration sparked my interest in AI and technology. It wasn’t something I had expected to enjoy that much, but the more I dove into it, the more I wanted to keep learning.”

‍

When did you first hear about Apart?

"I first heard about Apart during an ML alignment bootcamp in Cambridge in early 2023. Right after that bootcamp, I participated in Apart’s mechanistic interpretability hackathon—and I ended up winning it!"

“Off the back of that, I wrote a blog post with Apart that went viral on Hacker News. This was around the same time that ChatGPT had just launched, so there was a lot of buzz and curiosity around AI systems."

‍

How did you get involved?

“After writing the blog post, Apart reached out to me and asked if I wanted to work with them to continue developing my hackathon project into a research paper. The timing was perfect, and it was a great opportunity, so I said yes. This eventually led to a paper that is being presented at EMNLP 2024. I’ve also worked on various research papers since then, including mentoring newer lab fellows with their work."

"One of my favorite early tasks at Apart was making figures. I actually got involved in Anthropic’s Sleeper Agents paper because I made a figure for it. Although I wasn’t a co-author, I’m credited in the acknowledgments. I’m proud of the work I did, and I think making visual representations of complex concepts is an underrated skill."

‍

So you went from a Hackathon to working for Apart?

“It’s funny how one small project can turn into something much bigger. I didn’t expect to be offered a more permanent role at first, but I’m so glad I took the opportunity when it came. One thing I really love about Apart is how supportive they are of new researchers. The team understood that I was still figuring things out and gave me the space to grow while working on some really interesting projects.”

‍

What makes Apart stand out?

"I think the biggest difference is that Apart is a fully remote organization, which gives me the flexibility to continue my studies in Singapore while still being fully involved in research. Apart also has a very intentional focus on supporting researchers who are just getting started.“For example, when I joined as a Lab Fellow, they didn’t expect me to already have tons of experience with academic publishing. Instead, they guided me through the process, helping me build my skills while working on real papers.”

“Hackathons are also a huge part of Apart’s culture. Not only do they help build a community, but they’re also an excellent way for new researchers to get hands-on experience with AI safety problems. I highly recommend getting involved with those if you want to break into the field."

‍

Tell me about other work you’ve published?

“After my first paper as a Lab Fellow, I’ve mainly been working with other lab fellows on their projects. One of the more recent ones I collaborated on was accepted at NeurIPS 2024—it’s titled 'Interpreting Learned Feedback Patterns in Large Language Models. Apart's website has a blog on the paper, too."

“I’ve also worked on preprints like 'Benchmark Inflation: Revealing LLM Performance Gaps Using Retro-Holdouts' and 'Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs.' The preprints are on Arxiv if anyone wants to take a deeper dive!"

‍

What have you been up to this summer?

"This summer, I did an internship at Oxford, where I worked with the Torr Vision Group on mechanistic interpretability for vision-language models. It was a great experience and helped me grow as a researcher. I think the work I did with Apart definitely helped me land that opportunity, and I’ve been able to bring new insights back to my work at Apart."

‍

Future plans to stay in technical AI Safety?

“I’m considering doing a PhD, but I’m still thinking about it. There’s so much left to explore in AI safety, especially in the area of multimodal interpretability. For now, I’m continuing my research with Apart and helping to mentor the next cohort of lab fellows. I’m excited to see where things go!"

‍

Any advice for aspiring researchers?

"Absolutely! If you’re interested in AI research, I recommend joining our Hackathons and applying for lab positions. I wouldn’t be where I am today without the experience I gained through Apart's hackathons and lab fellowships."

“They’re great ways to not only improve your skills but also make meaningful connections in the field. Also, don’t be afraid to reach out to researchers and ask for guidance—it’s a collaborative community, and there’s always something new to learn!”‍

Clement is on Twitter, LinkedIn, Github, and Google Scholar.

‍