Research Fellow @ Convergent Research
220 karmaJoined Aug 2021Working (0-5 years)Pursuing a doctoral degree (e.g. PhD)



How I can help others

I can help you:

-Flesh out project ideas in biosecurity
-Figure out your PhD strategy 
-Edit your writing 


I just wanted to note that I'm very glad people are still producing evocative, classic-EA-style pieces like this. Thank you!

Thank you for doing this. However, there are some inaccuracies and misconceptions in the UVC section. I won't list them all here--chief among them is the equivocation between far-UVC and conventional UVC--but if I had comment access I could go through it. 

I wrote a few of the reports you link to (though the executive summary is deprecated and does not represent our current thinking--this one does) so I will go ahead and answer the questions you asked in the doc to avoid unnecessary duplicated work:

What makes inexpensive, energy-efficient  UVC light on the market ineffective? 

Operational difficulty. Trained technicians are required to install upper-room systems safely, and such technicians are in short supply. Installing it in-duct has the same issues, with additional validation and transparency problem. Check out Jesse's post on this subject.

Do they emit unspecific or varying wavelengths?

No, they emit nearly monochromatically at 254 nm. 

What is the efficiency of different Upper-room UVC devices?

You've already read Zollner, so you know the base emitter efficiencies. Low-pressure mercury is around 40%, UVC LEDs aren't really in use for upper-room GUV but they vary by wavelength and commercial stage and usually hover around 1-10%--though big caveat, efficiency is not the only important factor, and the issues UVC LEDs have with reliability are much more restrictive than the efficiency itself. Check out this 2021 DOE report for a less rosy outlook than Zollner's--personally, I think the UVC LED industry is being far too optimistic in their predictions about UVC LEDs ever getting as efficient and cheap as white LEDs. 

However, the louvers required to use cylindrical LP mercury lamps for upper-room GUV actually cause quite significant efficiency losses, down to 10%. Check out Kowalski (2009) Chapter 9--there hasn't really been a ton of innovation since then.

How can we ensure that certain light wavelengths are harmless? 

I won't get into the specifics of what safety studies are considered the standard of evidence here. You do it the normal way--run the intervention (exposure to UV light), and check the endpoints of interest. Here is an extreme exposure study, here is a long-term exposure one in mice, there are many more--here's a review, though it is from 2021. You would also need to think about environmental effects--for example, UV wavelengths below 240nm produce small (but increasing at lower wavelengths) amounts of ozone, which can then react with compounds in the air. UV can also degrade various materials, though shorter wavelengths don't penetrate as deeply. 

I do want to stress, however, that 'harmless' shouldn't be the bar; nothing on this earth is 'harmless'.

What is the current cost of UVC lamps, and how have prices trended in the past? What is the expected cost reduction? 

I am going to assume you mean far-UVC lamps specifically. If you want to get one fast, $2500. If you want to get one that's more reasonably priced, you can't right now, but in the imaginable future you can get them from Acuity or Beacon. Good luck actually buying the Acuity one, though--I think they're basically only B2B, and so is the source supplier Ushio who actually makes the lamp that all these fixture manufacturers are selling.

They cost only a couple hundred bucks or so to produce, so that's probably what the cost floor is in the near to medium term future. Solid-state sources will have different scaling economics, but they don't exist yet. 

What light bulbs currently emit UVC light? How much energy do these lightbulbs require?

Basically just those Ushio bulbs. All the alternatives are worse or don't exist yet. They consume 12 watts. 

Could pathogens mutate and evolve under UVC, creating resistant bacteria?

Probably not, nobody in the UV world is really worried about this. GUV is not new. This is unlikely because the mechanism of inactivation is pretty fundamental to life (for both conventional GUV and far-UV), which is not how most antibiotics work. It is imaginable to me that constant exposure to far-UV specifically might promote biofilm-forming bacteria to form biofilms more, but I believe the Esvelt lab is currently looking at this.

What are the projected timelines for this technology? What is the projected timeline for setting up UVC lights in offices?

I don't know what it means to have a projected timeline for a technology; that's too general a question. If you want far-UV, you can contact FarUV Technologies. It will not be cheap but it will be faster than alternatives. If you want upper-room, there are plenty of suppliers you can ask. Here's one I found on google. I know Kaleem Ahmid has done office installations in the past.

Could you elaborate on what you consider to be the specific marginal improvements of newer UV-C technologies?

We (the biosecurity team at Convergent Research) also have curated a number of specific, tarmac-ready far-UVC projects with teams and timelines attached, that range from a few 100k to a few million and address various bottlenecks in far-UVC adoption. We've got a whole catalogue of 1-pagers if any funders reading this are interested :P 

I think I have a slightly different read on USGOV's interest in far-UVC than Gavriel does. ~Every agency we've spoken to so far has gone "cool, but not our problem," which I understand to mean that  far-UVC is not being prioritized even within the already-deprioritized category of pandemic funding. Hell, even NIAID isn't interested. This might change, of course...but no real idea of what might be a turning point. That being said, the CDC doesn't seem actively unfriendly to the tech, so there is hope?

I think karma is awarded more by generality of subject matter than writing style per se? This post I spent a few hours on (including the hour I spent rereading the book) has x3 the karma of another post I put up around the same time that represents the outcome of about a year of part-time research. 

And this is perfectly natural! Everyone on the EA forum has some reason to care about good writing, only some small subset of people on the EA forum  have some reason to care about genetic engineering detection.  

Very good to know! I've never heard of a US master's program being paid. 

I wonder if the interest in US-based PhDs has something to do with the larger US academic offerings--or maybe it's just that unusually energetic people are both more likely to have early research experience and more likely to go to the US.

It's actually quite remarkable--the way we teach writing to students is anti-useful.  You could possibly do a worse job than we're currently doing but I don't immediately see how. 

There's something of a pepperoni airplane effect here. Everyone wants to think they're Proust.  I analogize it to the Picasso thing--you need to "learn the rules" before you can usefully take the training wheels off and start "breaking" them. Scare quotes intentional. 

I disagree re: Murakami (haven't read the others). I find him to be communicating extremely clearly. The actual book is full of specific examples of things that we think of as artful and indirect but that are actually bending the full force of themselves  into conveying  a very bright and specific concept.


I worked with Cass on the project mentioned and moving the needle on PPE seems much harder than I initially thought. The demand/scale thing is a real killer. There might be some solution here but it seems really muddled to me and I don't think it's throwing scrappy young engineers at it. Though of course the funding situation is different now too.


Good post--not least because I think this gives me insight into what the skeptics of EA-in-politics are thinking. I have a few responses:

It seems to me that two things are being conflated here: EAs individually running for office, and EA as a movement exercising political power. The latter, I agree, sounds like a terrible idea, for all the reasons you point out. But most of the arguments you bring up don't apply to the former. My model of EA in electoral politics looks a lot more like individual EAs (who are themselves pre-selected for being unusually charismatic, well connected, or otherwise well-fit for politics) behaving as basically conventional politicians, being team players, but making EA issues their top priorities on the rare occasions when they come up. That's perfectly consistent with pulling the rope sideways. 

There's a lot of middle ground between making EA as a movement a faction of the Democratic party and limiting EA involvement in politics to lobbying (though I think we should of course also be doing lobbying.) 

As far as I know, advocates for EAs running for political office are much more excited for EAs to run as Republicans than as Democrats, for precisely the reasons you outline (and several others). 

Finally, I want to comment that things like the Flynn campaign were always longshots, and that's fine; this is a numbers game and is largely going to be paying off over decades, not years.

Load more