Summary
Hi EAs, I'm Brian, co-founder of EA Philippines. I've spent a few days researching for and writing an article around the topic of "Can Filipinos help solve the problem of AI safety?". The format and content is drawn heavily from 80K's AI problem profile and career guide, and other AI-related career resources of 80K.
My goal with the article is to help inform Filipinos why this problem is important, and outline what type of Filipinos would be a good fit to be a technical AI safety researcher or engineer. I'm aiming to post this in a few Facebook groups that have Filipinos interested in or working on AI and/or data science.
I would like to ask feedback from EAs who are working on technical AI safety (or are familiar with it) on the article. I'd also like answers to the questions I've listed in the bottom section, and also get advice on the extra research we are planning on doing to validate or invalidate what I've written so far.
Context on the Article
I initially wrote this article back in April this year, as part of EA Philippines' project of doing cause prioritization and career advice research. This is so we can determine what are the high-impact career paths that Filipinos can take. We also researched on and wrote a problem profile on local poverty, and we also aim to publish problem profiles on technical AI safety, local farm animal welfare, and local mental health.
I initially wrote in the article that AI safety is a very pressing problem, but that EA Philippines doesn't see this career path as a "recommended" one. The rationale was that for a Filipino to be a technical AI safety researcher or engineer, they would need to get a Ph.D in computer science / machine learning at a top university abroad, and I hadn't found or heard of any Filipino with a Ph.D in computer science abroad yet before.
However, recently, I've found three Filipinos on LinkedIn who are taking or have taken PhD programs in Computer Science abroad. I have yet to get in touch with them, but it seems doable for really smart and talented Filipinos to take PhDs abroad and enter this field.
As such, I now think that a select few Filipinos can be technical AI safety researchers and engineers, and I thought of reframing the article to be more positive. This is so that Filipinos who are able to study and work abroad in the field of machine learning, or who are already studying or working abroad in this field, may consider working on technical AI safety.
Our AI Safety Career Advice Research Plans
Anyway, I realized though that I wrote the article without interviewing any AI or AI safety experts, such as foreign EAs working on AI safety, or top local AI professors, or employees/leaders in top AI firms locally. I only have gotten feedback from a few people in EA Philippines who are interested in AI safety, but they are not experts or AI professionals yet.
As such, a friend and I are planning on interviewing a few people to better inform our thinking and advice in the article. We have reached out to a top local AI professor (who isn't knowledgeable in AI safety, but has written a book on deep learning) to interview him, and he has agreed to it.
We also aim to interview 1-3 of those Filipinos who have taken PhD programs abroad, to see their familiarity with AI safety and if they'd consider working on the field, and to get insights on their experience in getting a PhD abroad in Computer Science.
Aside from this though, I'd like to get feedback from EAs familiar with AI safety on what they think about the content and advice I put in the article, and what questions we should be asking in these expert interviews we are planning to do.
Questions we'd like feedback on
Here are things I'd like feedback on or answers to. Some of these can be answered without reading the article, but it would also be good for you to read it first before answering:
1. Is anything in the article unclear, misleading, or false?
2. What could be improved in it? What's missing in it?
3. Aside from the U.S., U.K., Canada, Australia, and Singapore, are there other countries that have organizations working on AI safety research?
4. Is it possible to contribute to technical AI safety research even from outside of the 5 countries above, i.e. being an AI researcher in a university outside of these 5 countries?
5. Let's say I meet a Filipino who is interested in technical AI safety, but is unable or unwilling to work or study outside of the Philippines. Can he/she still contribute to technical AI safety directly as a researcher or engineer, whether by working remotely for a foreign organization, being an independent researcher locally, or working at a local AI firm?
6. Would it be more impactful for a Filipino to work on AI policy in the Philippines rather than try to go for technical AI safety? What would there need to be done in terms of AI policy in the Philippines? Should we consider doing research and writing an article about that?
7. What questions should we ask people at local firms to gauge if they are working on anything related to AI safety? How do we gauge if working at them would still be useful career capital for someone interested in technical AI safety research or engineering?
If someone you know can answer some of these questions, maybe you can tell them to comment here, or let me know so I can reach out to them. Thanks!
Appendix
For added context, here are the type of questions that we are planning on asking in the expert interviews we do:
- How would you describe what your research interests are?
- What are some research projects that you’re currently working on?
- What made you decide to enter the field of AI research?
- Are you familiar with what technical AI safety research is? What are your impressions and thoughts about this field and the problem of AI alignment / AI safety?
- What are your thoughts on whether AI systems would ever be as smart or smarter than humans? If you think this is possible, when do you think this might happen?
- What are things about the development of AI that you worry about?
- If someone wanted to have a career in technical AI safety research, could they be able to contribute to it while working in the Philippines?
- How can Filipino undergraduates best prepare themselves to get a PhD in Computer Science or Machine Learning?
- What are your thoughts on the state of AI research in the Philippines? How could this be improved?
I had a quick look over. I basically agree with the article. Here are some responses to some of your feedback questions:
2. Might be good to clarify that if you start a degree in US/UK, it makes it easier to get a work visa and job afterwards
3. You could argue that there's little bits in Switzerland, Czech Republic, Israel. Not so much in Aus anymore. but US, UK, Canada are the main ones.
4. Yes, it's possible. But generally you want to have some collaborators and/or be a professor. For the latter, you'd want to get a degree from a top-30 university worldwide, and then pursue professorship back home, so it wouldn't necessarily be easy.
And likewise for some of the expert interview questions:
5. You could check out Ajeya's report for some work on plausible timelines
7. Maybe, but it's hard. Either you'd need to find a startup that offers remote software work, or get a long-term job at a university
8. Same as non-Filipiino undergrads. Aim for papers and strong references.
Also, here are two other big picture elements of feedback:
A. A bigger picture question is: how can Filipinos best help to reduce existential risk? Often, the answer will be the same as if they were non-Filipinos - AI safety, biosecurity, or whatever. But one idea is that EA Filipinos could help with building US-China peace. The Philippines is close to China, and in major territorial disputes over the South China Sea. It's in ASEAN, which is big, close to China and somewhat neutral. So maybe it's useful to work for the department of foreign affairs or military, and try to reduce the chances of global conflict emerging from the South China Sea, or help to ensure that countries in ASEAN trade with both China and the US.
B. A lot of considerations for Filipino EAs interested in AI safety will be similar for a lot of EAs who aren't in Anglosphere or EU countries. But only a small fraction of these people are in The Philippines (~1%). So maybe for articles like this, it would be better to write for that larger audience.
Hey Ryan, thanks for looking over the article and the thoughtful reply!
On #2, yup I agree. I'll add that.
On #3, is there anywhere I can cite to say there's bits in Switzerland, Czech Republic, and Israel? And are these in universities or in organizations? Also, should I take out Australia from the list then?
On #4, what makes it important to get a degree from a top-30 university worldwide? And what would be the basis for which universities are in the top 30 - is it csrankings.org's global rankings?
On #5, thanks for linking that.
On #7 and 8, thanks for your answers to those.
I noticed though that your answers for #5, 7, and 8 were for the questions for the expert interviews I planned on doing, and not on questions 5-7 in the "Questions we'd like feedback on". You basically were able to answer #5 already there, so I'd just like your thoughts on #6 and #7 (on AI policy work and what questions we should ask people at local firms).
Also, I especially appreciate the two big picture elements of feedback.
On A, I agree with it being important to ask and think about that question. I generally agree that often the answer is the same for non-Filipinos, but I also think that most of these paths are out of reach or inaccessible to Filipinos who are unwilling or unable to study or work abroad. However, I've thought of two specific paths that Filipinos have a comparative advantage in and could help contribute to reducing existential risk within the Philippines, although I'm still unsure about both:
1. Advocating for and pushing forward nuclear power usage in the Philippines to combat climate change
2. Becoming an epidemiological or biosecurity expert in the Philippines and push forward policies, standards, and surveillance to make sure an engineered pandemic doesn't come here, or to make sure a natural pandemic doesn't start from or come to here.
One could argue that a Filipino skilled to do either of the above could do it in a country that needs those skills more than the Philippines, but being a Filipino would help someone have the network and comparative advantage to do the two things above here locally.
Regarding what you said about EA Filipinos helping build U.S.-China peace, I didn't think that would be an impactful, accessible path. But now that you mention it, it does seem plausible. I would love for you to expand more though on how a Filipino can "try to reduce the chances of global conflict emerging from the South China Sea, or help to ensure that countries in ASEAN trade with both China and the US." I am unsure about the specific pathways available in doing this, or the "theory of change" to doing this - would you have thoughts on what this could look like?
After some quick Google searching on my end, I discovered the website of the Embassy of the Philippines in Beijing, China, as well as the website of the Embassy of China in the Philippines. The appointment of a Chinese ambassador to the Philippines also alludes to the issue in the South China Sea. So I assume these two embassies would be very important in dealing with PH-China relations and the South China Sea. This could be interesting for me to look into since I'm Chinese by blood (but born and raised in the Philippines) and I can speak some Mandarin. I also assume that the U.S. and Philippine embassies to each other would be important as well to work in?
Aside from that Google searching, the main thing I know currently about this issue is that the Philippines' former Supreme Court Chief Justice joined the team to "sue Chinese officials including President Xi Jinping for illegal incursions in the South China Sea before the International Criminal Court." Of course, most Filipinos are in support of the Philippines having claim over these waters over China (since they are rightfully ours), but I am unsure on if the Philippines being able to assert and win this claim is going to reduce existential risk or not. If you or others think it is though, then I could consider finding ways to network with people who work with or know the former Supreme Court Chief Justice, so that we can figure out if we can recommend Filipino EAs to work on this, or help draw more attention to this cause.
On B, I get what you mean. I'll consider making a version of the guide that is for people interested in technical AI safety but are from low or middle income countries!
Ah, oops! 6. I'm not sure AI policy is that important in the Philippines, given that not that much AI research is happening there, compared to US/UK. 7. Relevance to AI safety is a bit tricky to gauge, and doesn't always matter that much for career capital. It might be better to just ask: do I get to do research activities, and does the team publish research papers?
On A, yeah it could make sense to push for nuclear power, or to become a local biosecurity expert. To be clear, the US China peace issue is not my area of expertise, just something that might be interesting to look into. I'm not thinking of something as simple as fighting for certain waters to be owned by China or the Philippines, but more to find ways to increase understanding and reinforce peace. Roughly: (improved traid/aid/treaties) -> (decreased tensions between China and ASEAN) -> (reduced chance of US-China war) -> (reduced risk of technology arms races between US and China) -> reduced existential risk. So maybe people in the Philippines can build links of trade, aid, treaties, etc between China/US and neutral countries. These things are probably done by foreign policy experts, diplomats and politicians, in places including embassies, the department of foreign affairs, national security organisations, and think tanks and universities.
Good work Brian! I'm guessing it would be a good idea to cross-post this to the AI Alignment forum: https://www.alignmentforum.org/ (although I'm not mega familiar with the norms on that forum)
Thanks Sanjay! I've applied to be a member of the AI Alignment forum yesterday, but I haven't heard back. It seems like they have a high bar to who they allow to post there and what types of posts they post there. I've gotten a couple of comments from Ryan Carey and Rohin Shah so far though on the article, and that might be enough for me.