AI Beats Pro Players of 6-Player Hold’em Poker

In recent years we have heard that computer science is taking big steps using artificial intelligence to solve problems that were thought to be impossible to solve. One of the standards that is commonly used to try and understand the “intelligence” of a machine is to have it compete against humans.

One example many of us remember was when chess grand master Kasparov was defeated in 1997 by IBM’s computer Deep Blue in a highly publicized match. More recently, 2015, we have the case of Alpha Go, a system created by Google and that was able to beat top professional Go players. Go is a board game very popular in Asia which is considered much more difficult for computers to win than other games such as chess due to the bigger number of possible outcomes that makes it prohibitively difficult to solve using traditional Artificial Intelligence (AI) methods.

What makes Hold’em Poker so hard for computers is that it is a game of imperfect information, meaning you don’t know your opponent’s cards in order to take the right decision and therefore is necessary to guess. More importantly in this case there isn’t only just one adversary, but five!

Pluribus, the name of this AI, was developed by scientist at Carnegie Mellon University in collaboration with Facebook. It was trained by making it play against itself, and it only needed the equivalent of 150$ of cloud computing resources to achieve this level and it only took 20 hours to train it to superhuman levels.

When playing, Pluribus runs on two CPUs and uses less that 128 GB of memory while AlphaGo needed 1,920 CPUs and 280 GPUs for real-time play in 2016. The AI plays twice as fast as human pros, requiring on average 20 seconds per hand.

The experiments where 5 pros and AI played in the same game ran for 12 days and 10,000 hands of poker were played. As a results Pluribus won on average 1,000$ per hour, if each chip was worth one dollar. This is higher than the rate at which the pros win when playing against a mix of pros and amateurs.

So, what do you think? How would this affect the online Poker players? Do you think there is a way to detect this AI player? Is Skynet about to take over the world? Food for thought.

If you want to see a quick review about this research you can check this video by Two Minute Papers, a youtuber that works as a researcher in computational science and that reviews papers related to this topic on a weekly basis. The video explains more clearly how the AI works.

Facebook also published a post on its blog explaining it more deeply. There you can check the list of professional players and some of their opinions about Pluribus.

And if you are all-in with this kind of research and feel like that wasn’t enough, you can read the full paper published in the prestigious scientific magazine Science here.

Originally published in September 2019 in my Linkedin.

Peptides, the solution to drug resistant bacteria?

In 1928, Alexander Fleming discovered a mold that had contaminated his bacteria’s cultures and where the mold grew the bacteria was destroyed. This serendipitous event lead to the discovery of penicillin, the first antibiotic discovered. Since then, several new compounds have been discovered with antibiotic activity. Unfortunately, the discovery of new antibiotic compounds in nature hasn’t led to new significant discoveries in the last years. This could be due to the fact that we have discovered all the compounds nature had available or that we are not looking for them in the right place.

This is a big problem because some bacteria are becoming drug-resistant. This means that the antibiotics known up-to-date are not effective against them. This problem is a consequence of the nature of the problem itself. The incorrect use of antibiotics, like stopping the treatment with antibiotics before the infection is destroyed or using them as a preventive method against infection in situations where it is not strictly needed, allowed the bacteria to evolve and develop their immunity to drugs. This led to Center for Disease Control and Prevention (the American CDC) to publish a report in 2013 where it sounded the alarm considering this issue “one of the biggest public health challenges of our time”.

Therefore, a new approach is needed to tackle this problem. A possible solution is the use of peptides. Peptides are chains of amino acids, they are the base of proteins and enzymes, and they are synthesized by all living beings. It was discovered that some of these peptides present antibiotic activity, so what if instead of trying to search for more active peptides in nature, we force them or direct them to evolve? And what if instead of doing it in vitro we do it in silico?

Image created by Rodrigo M. Ortiz de la Morena
The ribosomes are responsible for the synthesis of peptides in all living beings

What this means is that some scientists are focused on digitalizing peptides existing in nature to accelerate the discovery process. The natural occurring peptides are processed in a logic mathematical way to transform all their chemical information into what is called descriptors. Then the search is directed towards a peptide with an optimal set of descriptors for an efficient activity against bacteria. Furthermore, we use the powers of computers to create derivate peptides of which the antibacterial activity is predicted by the computer too.

One of the algorithms used by the researchers are genetic algorithms. These algorithms use an initial population of peptides and make them follow Darwin’s evolutionary principles. They are mutated and then the “fittest” ones are selected to be recombined and generate a new population of peptides that contain random mutations and traits inherited from the peptides in the previous generation.

Image created by Rodrigo M. Ortiz de la Morena
The genetic algorithm evaluates the different peptides of the new generation

These algorithms are combined with machine and deep learning ones, a new area of research in computer science that has applications in many fields including this one. They are used for the design of the peptides and also for predicting their properties. This has led to the discovery of peptides that were 160 times more active than the natural peptide.

On top of helping in the search for a medicine to fight drug-resistant bacteria, the new peptides can be used to target different kinds of bacteria in the gut microbiota. The gut microbiota is the group of different microorganisms that live in our digestive system. For a long time, there has been discussion about the implications that our gut’s health has on our mood. But, during the last years, several studies have focused on studying the bi-directional communication between our gut and our brain, in what is called the gut-brain axis. These studies have related the gut microbiota with behaviors and symptoms typical of disorders in the autism spectrum and others linked it to diseases such as Parkinson and Alzheimer.

Image created by Rodrigo M. Ortiz de la Morena

The newly peptides can target specific bacterias to destroy them

In conclusion, there is no doubt that this is an interesting topic for research. It involves several disciplines like biology, computational science, chemistry, and neurosciences. The rapid progress of new artificial intelligence technologies and cheaper computational resources will accelerate the rate at which discoveries are found. Hopefully these discoveries will lead to solutions to fight against drug-resistant infections and perhaps generate new therapies for neurological disorders.

Check the short explainer video about this article


Second Video, New Prize!

Last year the Material Research Society organized again their SciVid competition for their fall meeting in Boston. The conditions where similar to the ones they had the previous year. The objective was to create a narrated video of 2 minutes explaining a scientific concept or research related to materials science and 3 prizes would be awarded by a panel of experts in science outreach videos from the pool of the finalists, along with an additional prize given to the video with the most likes on YouTube and votes casted during the conference as a people’s choice award.

This year I decided to present a video in which I talked about the kind of systems we worked on during my PhD. The topic is self-assembled monolayers and I could explain to you what they are right here right now, or you could easily watch the short 2-minutes video I made for it.

The video was done in quite a rush, since I was in the final stages before submitting my thesis but, I used some animations I already had used in my presentations at conferences. Furthermore, I followed the advice of the jury panel last year of including less CG and more people in the video. This made it faster to create, but it was all new to me, so we had to wing it at times to get there in time for the deadline. Despite this, WE WON THE PEOPLE’S CHOICE AWARD! ( With almost 900 likes in YouTube by the time the voting was closed (10 days), almost doubling the likes of the next contestant.

The video format is more a talking head video, with me explaining the topic while some graphics and tittles help support the explanation mainly, and of course some science CG scenes made with Blender. In fact, the whole video was created in Blender, including the editing the creation of the tittles and motion graphics, if you wish to call them that.

Well I hope you enjoyed the video and that you learnt something new, this time related to my own work as a PhD. I would also like to thank Tade, who was the cameraman, Toni, who went and pick up the prize in Boston for me, and of course all the people that supported me by giving the video a like in YouTube, specially Finofilipino, who shared it in his blog full of trolls XD.

PS: sorry for the late positng

First Video, First Prize

In the beginning of October, the first video of Summa Scientia was published. It talked about how the gas diffusion layer in hydrogen fuel cells can be improved. The friend who I made the video for really liked it and, just a couple weeks after it was published, he found out that the Material Research Society had arranged Scientific Video competition that would be held during the fall meeting in Boston. 

The competition focused on creating a narrated video of 2 minutes explaining a scientific concept or research related to materials science. And so, with a bit of video editing and the narration from my friend and research author, Toni Forner, we managed to create a new, updated version of the video. 

We submitted the video by the end of October. The results of the competition would be announced at the end of November and several prizes were to be given out. The jury members, who were people with experience in scientific outreach videos, were assigned to select the 3 best videos from the pool of the finalists, along with an additional prize given to the video with the most likes on YouTube and votes casted during the conference as a people’s choice award. 

By mid-November they let us know that we had been selected as one of the finalist and therefore we started our own “promotion campaign” to get every like we could get. From friends, family, colleagues,… We started the race ahead, but we quickly noticed how a couple of videos were getting votes much faster than any of the others. In an attempt to boost our chances, I contacted the local newspaper in Alicante and even though their reply to help us with the people’s choice award came a bit late, they were very interested in knowing how this collaboration came to happen and see how the research made by one of their neighbors was being supported abroad. 

Unfortunately, we weren’t able to gather enough votes for the people’s choice by the day the ceremony took place, on the 29th of November in Boston. BUT WE WON THE THIRD PRIZE!! Toni had a chance to attend the conference and receive the award in person. Although I was on the opposite side of the Atlantic, I followed the ceremony online and also celebrated. One month later we received an email from  Alicante’s newspapers saying that the article about us and the competition would be published. And although we were not in Alicante that day, we managed to get copies of the paper. 

To sum up, the first video of Summa Scientia evolved into a successful collaboration, was presented in a competition for scientific videos and won a prize.To top it all, our pictures appeared in an article about the work in the newspaper. I’d call that a success! 


How to Improve Hydrogen Fuel Cells

The need for an alternative to combustion engine cars is a growing concern in many countries. Hybrid cars and more recently electric cars rose as an obvious solution to this problem. But a third option has been overlooked and dismissed since 2015 when Hyundai and Honda presented their firsts commercially available hydrogen fuel cell cars.

While electric cars may seem a perfect solution for the problem, the capacity of the batteries and the long recharging time can be a detriment for the total replacement of the combustion engine cars. The higher energy density of hydrogen allows these cars to have a driving range comparable to traditional cars. The refuelling is as simple and fast as with a combustion car. The existence of an infrastructure to distribute and provide petroleum based fuel could be converted to the hydrogen technology.

The hydrogen fuel cell car works by combining the hydrogen with oxygen from the air in the fuel cell. The cell works as a catalyst of this reaction and allows to harvest the energy produced in the form of electricity, a more efficient way compared to combustion. Also, any excess of energy produced can be easily stored in a battery. The hydrogen fuel cell must still improve to be competitive compared to electric cars.

One of the areas in which it has been tried to improve is in the removal of the water produced while bringing air inside the cell to use the oxygen in the reaction. This problem is known as water management, and it is a hard problem to solve. On one hand, the cell must be sufficiently wet so that protons can travel from the anode to the cathode. But on the other hand, too much water could suffocate the cell by not allowing oxygen to get to the cathode.

The way where this exchange of water and air is made is in the gas diffusion layer. This gas diffusion layer is usually made of carbon fibres. Capillary pressure is the driving force for this exchange and for this reason the fibres are usually coated with hydrophobic polymers. Hydrophobic polymers are made of molecules whose interaction are very unfavourable with water. They are the base of the always-clean fabric technology, by avoiding the absorbance of water and water based liquids in the fabric.

The coating of the membrane with these polymers improves its performance but it still has the problem that the water tries to take the least resistance path in the membrane, which means that those paths are usually tortuous or sometimes lead to a dead end.

Another attempt tried to stack hydrophobic membranes with hydrophilic ones. Hydrophilic polymers are those whose interaction is favourable with water, and therefore water prefers to be in contact with them rather than with the hydrophobic polymers. This also provided an improvement, but this approach had a limited design flexibility. This drawback made it rather limited in its scale-up potential.

A new method was presented a couple years back. It allowed to create specific pathways for the air and water inside the same membrane. The research explains the preparation method of this new kind of gas diffusion layers and their capabilities. In this case the design is very flexible and can be easily adapted to new condition and configurations. Furthermore, it is ready to be scaled up to a production level.

The research also shows X-ray images that allow to see the distribution of the hydrophobic and hydrophilic areas and neutron spectroscopy, which allows to study the water distribution inside the layer and its behaviour in different conditions.

If you want to know more about this research don’t miss our video making a short sum up. But if you are still intrigued as to how this works and how the research was made you can check it yourself. The papers are open-access and therefore downloadable for free.

Also you can read this article from nature, the scientific journal, about hydrogen on the rise.