Calling it a follow-up would be a disservice. We used a similar platform and chemistry (the BZ reaction), but the experiments done took the idea to the next level. My work here mostly consisted on the technical side of things, while the other authors did the interesting science. I wonder where this research will go next!
]]>“ChatGPT use shows that the grant-application system is broken”
]]>Zombies are mythological undead creatures found in horror stories. These stories usually involve a dead animal getting infected with a virus or similar sci-fi mechanism, and coming back to “life” through reanimation. There is some debate in the community about zombies being alive or dead, and in this research we will consider that they lie somewhere in between. Interestingly, a similar process happened four billions years ago, when inanimate matter became alive. Scientists call these theoretical first living entities “protocells”. But were they really alive? What about “proto-protocells”? While today the distinction being “alive” and “dead” is a binary step, was it the same during the origin of life? To answer this, we will use Generative Artificial Intelligence (AI), a technique used to describe or model datasets, and to generate novel data by sampling from them.
]]>“Support letters: mostly ghost-written, always glowing. What’s the point?”
]]>“Dear grant agencies: tell me where I went wrong.”
I actually enjoyed the process of writing a non-scientific piece, and hopefully I will do it again in the future.
]]>The idea behind the paper is simple. 3D-printer can enable anyone to manufacture pieces that can perform different actions. In the case of this paper, we wanted to manufacture devices that could perform chemical reactions. The problem is that if you want chemists to use it, not only they need to be OK using a 3D-printer, but they also need to know how to use CAD software, and this last step severly limits the size of your potential user-base.
In this research we designed a piece of software that can make very easy to design different devices that can perform chemical reactions. You don’t need to know anything about CAD. You only need to click options through a bit of menus and boom, done.
The name of the paper is “Automatic generation of 3D-printed reactionware for chemical synthesis digitization using ChemSCAD”, and it was published on the journal ACS Central Science.
My work mostly consisted on developing the software after the original developer left the group. There was a lot of time involved to reading and understanding his code, which is one of the less rewarding things in computing science, but it was good experience!
]]>The “BZ project” as we call it (BZ from Belousov–Zhabotinsky, which is a type of oscillating chemical reaction) was started by some other people (Cooper and Donkers as they are named in the paper). As far as I knew the project was going OK, until Kevin (Donkers) left the group, and then the project was orphan for maybe 1-2 years, until I tried to recover it from the ashes. I have to say I don’t know much about unconvential computation and quantum computers. As a computer scientist I know about your standard computers, but still today I am not sure how a quantum computer works.
Anyway, the initial idea was that I was going to improve the platform from an engineering perspective, and then someone else was going to help me do the experiments. I completely revamped the platform, changed the electronics, improved the Computer Vision side of things, and use some AI to close the loop between chemistry and automation, but then, once the platform was working and it was stable enough, I did not really know how to make a “computer”. I tried to do some basic stuff, like binary logic gates, with the BZ reaction, but the chemistry was very unstable and not very reproducible.
I was in a sort of a dead-end, and I decided to use the opportunity to learn about Deep Learning. This project, in particular, uses Convolutional Neural Network to “learn” how the chemical reaction works from a visual perspective, and then it uses a Neural Network Autoencoder to transform between automation and chemical oscillations.
The name of the paper is “A programmable chemical computer with memory and pattern recognition”, and it was published on the journal Nature Communications, which is a journal were, for some reason, I am publishing all my major papers. And the best thing is that the article is Open Access, so if you click on that link, you can read it for free.
It has been published just as the world is being hit by the virus Covid-19, and that’s why I think the “impact metrics” of the paper are not amazing (or so I hope). Before it being published on Nature Communications, we did publish a draft version of it on chemRxiv, and this “rxiv” draft did very well, with almost 11k views and 1.8k downloads, being one of the best performing papers on chemRxiv, while sadly, as said, the main paper in Nat Comms is not doing that well.
]]>The paper is about using crystals to generate random numbers. My work consisted mostly on using Computer Vision to find crystals on vials, and to characterize them. To do so, I used the Mask-RCNN library. If you want to know more about it, the magazine Vice wrote an article about it.
]]>The Fellowship Symposium was an event organized by the university, where all the new fellows were asked to give a 5 minute presentation, without time for questions, although there was time for networking after the event. There was quite a lot of people in the room, mostly other fellows, lecturers, professors and different academics. Therefore, it was a quite a serious event. I think in total we were something like 12 fellows, and I have to say most of the them presented very good research, which made me think I need to bring my level up. During the break I could speak with some other people and make some networking, so that was really good!
7 minute of Science was the polar opposite. It was organized by students from the School of Physics, and it consisted of presentions of 7 minutes followed by 7 minutes of questions. The room was also quite full, and something that surprised me very positively is that it was mostly young people, probably undergraduate students. Usually on conferences or in events like the previous one you present to senior academics, so this event was a good change!
]]>