More on AI, COVID-19, and Revised Research Practices
41:52 Share

Hello and welcome to Skynet today's Let's Talk AI podcast where you can hear from AI researchers about what's actually going on with AI and what is just clickbait headlines. This week we'll look at more ways AI is being used during the coronavirus pandemic and also discuss how the AI community is coming together to see what you can do as a global field of research.

I am Andrey Kronkov, a third-year PhD student at the Stanford Vision and Learning Lab. I focus mostly on learning algorithms for robotic manipulation.

And with me is my co-host. I'm Sharon, a third-year PhD student in the machine learning group working with Andrew Ng. I do research on generative models, improving generalization of neural networks, and applying machine learning to tackling the climate crisis. Yes, indeed. And it is now actually a time of crisis, which is interesting.

Just a week ago, it was announced that here in the Bay Area, what was it, six or seven counties were to go into shelter-in-place mode. So I biked over and handed over some microphones to you so we could actually record this thing. So we're now recording remotely for the first time, and this is really interesting. And of course, shortly after that, the entire state has gone into shelter-in-place. And also more shortly after that,

I Purell to all of these microphones. As you should have, very wisely. So...

I guess we are all adjusting and trying to return back to our somewhat normal activities while still being sheltering at home. Somehow I managed to pass my PhD quals during this time. Oh yeah, you I presume had to do it all over video calls and whatnot? It was all remote over Zoom, yes. I would hope that did not make it harder.

I hope not. It allowed me to have a little bit of fun with the backgrounds. Okay, that's good. And all right, so that's enough about our daily lives. Let's go on with last week's news about AI. And the first thing we're going to get into is a bit more discussion of the technical details of how AI is being used to combat this crisis. So the first article here is from IEEE Spectrum.

And it is titled Five Companies Using AI to Fight Coronavirus. And one of the companies is DeerGen from South Korea. And DeerGen scientists published a preprint paper, a paper that has not yet been peer reviewed by other scientists, and that's what preprint means, with the results from a deep learning based model called MTDTI. And this model uses simplified chemical sequences rather than 2D or 3D molecular structures to

to predict how strongly a molecule will bind to a target protein.

So the model predicted that of available FDA-approved antiviral drugs, actually the HIV medication adizanavir is the most likely to bind and block a prominent protein on the outside of SARS-CoV-2, the virus that causes COVID-19. I don't know much about chemistry, so I'm just going to assume that means that it should help. I think that's saying that we suspect this HIV medication will actually help.

against COVID-19. We suspect, of course, this paper is a preprint so that we can get these papers out more quickly, but of course it's not peer reviewed. And yeah, but it's research that's being pumped out as fast as possible into open source just so that people can leverage it globally. Yeah. And this is something to note, of course, is that in general, all of these companies and AI in general

can be used to predict which drugs might help, might be useful. And then it is, of course, important to do actual clinical trials with all the necessary precautions to make sure that is actually true. So it's good to have some leads to go on, but this isn't immediately a cure. And another company in Hong Kong called Insilico Medicine, instead of seeking to repurpose available drugs, so

Instead of finding drugs that already exist, the team used an AI-based drug discovery platform to generate tens of thousands of novel molecules that have the potential to bind to that specific protein that causes COVID-19 and block the virus's ability to replicate. And so they were able to generate these different candidates, and then they had a deep learning filtering system to narrow down that list.

And this group published the original 100 molecules and planned to follow up by making and testing seven of them. However, and very, very unfortunately, this group was interrupted by the pandemic with over 20 of their contract chemists quarantined in Wuhan in the very beginning. Yeah, but they've still been working on it. So this article states that they have since synthesized two of the seven molecules.

and with a pharmaceutical partner

uh plan to still put them into test over the next few weeks and they're also in a process of licensing their platform to several large pharmaceutical companies and next we have sri biosciences and the second company ictos which are working in collaboration so ictos has a deep learning model that does this prediction part of designing novel molecules

Well, this SRI company has a synthetic chemistry platform that can figure out ways to make that molecules. And then the two together work to be able to test new drug-like molecules in just a few weeks.

And yet another company, and as you can probably tell, there are a bunch of companies working in this space, which is fantastic. But Nevelin AI, a British company, published two articles in February identifying approved drugs that might block the viral replication process online.

of the protein that causes COVID-19. So they used a large repository of medical information, trained on that data using machine learning and identified six compounds that effectively block a cellular pathway that

appears to allow the virus into cells to make more virus particles. So of these six candidates, they found that a once-daily pill that's been approved to treat rheumatoid arthritis actually looks to be the best candidate, both in safety and efficacy, against the COVID-19 protein. And Benevolent has reached out to manufacturers of this drug about testing it as a potential treatment, which is exciting.

Yeah, so we discussed this a little bit last week of how there are many different companies working to use AI for drug discovery. And so this is kind of going into more detail on how varied they are, how they are in multiple parts of the world, how they are pursuing different techniques and have different focus.

And as we discussed last week, this is sort of a real kind of moment for them to prove themselves possibly and show that these techniques are very useful because they are somewhat young as a category of businesses.

And so it'll be interesting to see if this actually does help us significantly. I'm also struck by the different things that people do find or the different candidates that the various companies do find. And I realize that we need...

a lot of diverse possible candidates to be surfaced right now since I've been told that it actually will take about 18 months for everything to go through clinical trial with humans and also just be approved and go through that entire approval process. So it takes a while. And so one question that we might ask now is, should AI help make life or death decisions in the coronavirus fight? Yeah.

Yeah, so this was an article in the South China Morning Post, which has many articles covering events and how AI is being used in China.

And it makes the point that Chinese researchers say we have developed an AI tool that can assist doctors in analyzing blood samples to predict survival rates. And so now you have this AI telling you how likely a person is to survive. And the question is, should that prediction be used by the doctor to decide who gets treatment, who gets priority, which is basically a life or death decision. So basically it's...

should this AI be able to make decisions on how likely, given how likely someone might die based on various sources of information that might be biased, should they be able to make that decision? Say like this person has a higher risk of death and therefore

should get treated first or should get treated last because there's a smaller just chance of survival in general. And so this is a persistent question in the medical space. And I think, I mean, we can basically agree that AI should not be making any decisions related to life or death, but it can inform those decisions for actual medical professionals and staff.

So given the prediction of AI as to survival rates, that can be an important thing to consider when a doctor decides on a course of treatment or who should get priority on resources that are already kind of being stretched. Yes. And I think that especially when, let's say, doctors making decisions might be very fatigued, sleep deprived and everything, stressed.

It may be significantly more helpful to have a tool like this to support decision-making at times. And of course, a tool is going to be inevitably biased. Doctors will also be inevitably biased. I've been told that various hospitals, including those in the U.S., have started to prioritize certain individuals over others. For example, using BDL,

even as a way of distinguishing whether someone gets care or not in terms of getting a ventilator. So it's happening in terms of just prioritizing people in treatment, both in China and in the U.S. Yeah, and so I guess a more nuanced take on this question is to what extent should AI influence the decisions of the staff in general or the doctors involved?

And that really depends, I think, as we can say as developers of AI on how reliable the algorithm or model is. So in this case, it was developed by Chinese researchers from the Huazhong University of Science and Technology and the Tongji Hospital in Wuhan.

And the claim is that the AI system has achieved 90% accuracy on the fatality and survival rates of more than 400 patients based on blood samples.

collected in this Wuhan hospital. So that's a promising sign that it is an indication that this is a useful tool probably, but at the same time, it is quite new. So I would say medical professionals should probably be careful about trusting it too much.

I suppose one question to ask is, what is the current method without such a tool, right? How are people prioritizing different patients right now? And where is that flawed? And where does that succeed? And how can we keep the aspects that are very good? And how do we try to mitigate some of the flaws and shortcomings, perhaps with an AI tool?

So moving on from that, we also have an article on infrared AI cameras at polling places could spot voters with a fever to detect potential coronavirus carriers.

So Athena Security, which was launched in 2018, recently pivoted from using thermal imaging and computer vision to detect guns concealed under clothing. Now to detect coronavirus, to do fever detection for coronavirus for public health purposes.

and security. And so the co-founder and CEO, Lisa Falzone, said the platform combines infrared cameras and an algorithm that analyzes body temperature to detect people who have a temperature higher than 100 degrees. Yeah, so the claim is that different companies can use the system to screen employees and basically prevent them from coming to work sick.

And this would be at a relatively low cost of something like $7,000 for the system plus some monthly subscription cost.

And they do note that the system blurs the faces of people in the camera footage, especially for gun detection. And so it is kind of supposed to not display the person's race or ethnicity. I will say that this puts certain people at a disadvantage. For example, folks who are perhaps deaf,

disabled, have any type of disability, it's quite easy to spot any sort of disfigurement in these types of cameras, even if you blur that type of information away. Even if you just use depth sensors, for example, we could still re-identify people from this data. So I think this is to help perhaps sell the system, but realistically, this type of private information could be recovered.

Yeah, so it's kind of a trade-off as usual in terms of privacy versus how safe everyone can be made. Maybe in this time, the trade-off is worth it to record people with fevers, but of course, we'll have to be careful not to overstep. There is kind of an interesting connection to research being done at Stanford. So in my lab, actually, at the Stanford Vision Learning Lab,

One of the groups is focused on healthcare, and they have worked for a few years on using computer vision and depth sensing, and in some cases, thermal imagery, to be able to detect if staff in hospitals are cleaning their hands regularly as they move about. So if they're moving up to little hand dispensers and actually...

getting rid of germs as they should be. So there's some precedent for this sort of thing. And their claim was also that if you use depth images, you could anonymize and not show people's faces while still detecting if they are following the necessary routine. Surprise, surprise. I was on that project. Oh, okay.

You can actually recover identity from it. Yeah. Which is not too surprising. I mean, it's not as easy as if you had RGB imagery, but still, if you have enough data and you have a good enough sensor, it helps. But I think it does help with the creepiness factor, right? Yes. I think it helps with selling a product. I think it helps with preventing humans from detecting anything. But it...

the neural network could probably figure out the pattern quite easily. Okay, and then a related article we have here is how China is using AI and big data to fight the coronavirus. I think we might have already mentioned it in a previous episode, but it's related here because one of the things it mentions is that it has already deployed such cameras, such thermal cameras to be able to

screen people for fever as they go into this railway station. So there is precedent for this. And again, it seems like at this point, especially in a place like a railway where you have a really large number of people, you cannot check each person individually. It might be reasonable.

But then there's a question of are these systems going to stay in our airports or train systems after this crisis passes? Well, certainly after 9-11, we changed the entire security TSA process in airports, right, to board an airplane. So the question definitely persists whether this type of technology will simply remain. Mm-hmm.

And an additional interesting tidbit of what the Chinese police now have are these AI helmets for temperature screening. In fact, these infrared cameras are attached to these helmets that enable police to measure temperature from up to five meters away of people around them. And they can scan QR codes and use facial recognition to

according to the company that makes them. And so this is the Kuangchi Technology Company that's based in Shenzhen that introduces police smart helmets. Yes. So again, this is from the South China Morning Post. And the subtitle to this article is, this RoboCop style headgear can screen out potential coronavirus carriers.

I think we were looking at the image of topic and agree. It's not really that RoboCop E it's more like a helmet with a camera and fancy glasses attached to it. The news might be a little sensationalist. You know, a little bit, it doesn't make you want to click and read and look at the actual headgear. Um,

But it is interesting how having built out a system for facial recognition and thermal imagery, now it can be made to be this mobile, this local, such that a police officer can have it and see it immediately in front of their eyes as they look around. It does seem a bit science fiction-y. This definitely reminds me of...

the speeding guns that police officers use on the highway. It's kind of like supercharging the police with some kind of technology. And in this case, it's an infrared camera that makes it much easier for them to track people in crowds.

Yeah, I think one thing to note maybe about all these thermal systems is right now when we are in the Bay Area and our orders are basically to stay at home as much as possible, maybe they do seem a bit like overkill. But I guess the hope is we can actually go back to going outside, you know, within a few months or something. But yeah,

People are also predicting that coronavirus will not go away just within a few months. So we'll still need to be careful to wash our hands. We'll still need to be tested a lot. This will last a while. So possibly when we do start to reduce these extreme measures of just staying home,

These sorts of fever detection systems can be really useful for airports or railway systems in helping us kind of avoid another really large outbreak. Another thing to note is that Tylenol and Advil are...

pretty much everyday drugs that we have in our households that we take even here in the U.S. But in China, it's actually much rarer to use these types of drugs. And they are essentially fever reducers. So these infrared cameras are actually significantly more effective outside of places that typically use fever reducers or that people would use fever reducers. So that is also something to note.

Yes. And so on the topic of ways in which the world is changing and might, uh, persistent being changed, um, for quite a while, our next article here is from tech crunch and discusses how the company diligent robotics has raised an additional 10 million, uh, in funding to develop a robot to assist nurses in hospitals. So they, uh,

have this interesting statistic that 28% of a nurse's time is wasted on low-skilled tasks like fencing medical tools. So that obviously sounds like a lot of time that could be spent doing more useful things. And the idea for this company is to have these robots be able to take over a lot of that low-skill labor and work with nurses to take care of easy stuff so that the nurses can use their expertise more effectively.

This robot is called Moxie, and Moxie is about the size of a human. But it's designed to look like an 80s movie robot. So it's not trying to cover that uncanny valley cyborg weirdness. The founders of this robotics company have thought a lot about human-centered design and thought a lot about human-robot interaction and how to make this a very usable robot.

Yeah, I think it evokes a little bit Eve from Wall-E, if I remember correctly. Yeah, with an arm. With an arm and not floating. Of course, it has a little wheeled base, but it's a lot of white plastic. It has a screen that displays kind of emojis. So it's meant to be very friendly and to really work with people, which is something that has been challenging for robotics so far. Right.

Right, exactly. And diligent robotics is not the only player. They've been competing with companies like Atheon's TUG Tugbot for pulling laundry and pharmacy carts, and other players in the hospital tech space also include Xenix's machine that disinfects rooms with light, and surgical robotics like those from Johnson & Johnson's Oris and Intuitive Surgical.

Yeah, something interesting about it is that it was founded actually by a robotics professor, Andrea Tomaz, who has done quite a bit of research in human-robot interaction and having humans teach robots skills. And so they are really building on that expertise in developing this robot.

And they've been at it for a few years now. They are at the point of having 18 people. But it seems now with our current moment, the need for such robots is becoming ever more clear. And so hopefully they are able to continue their work and actually make this robot be effective at what they're making it for.

I think the demand for such a robot and in general robotics right now is ever increasing and higher than ever. This is the time essentially perhaps for robotics to be not only just useful and slightly better, but significantly better and significantly more helpful where it's dangerous for humans to be working or it's much more useful for a human to be doing, like they said, complex and compassionate work.

where robots could perhaps work on some rote tasks that would help significantly in getting nurses to work on those other types of compassionate work. Yeah, and one other thing we can mention about this topic of robots working with humans is actually it speaks to how AI or some problems in AI are still very hard. So we've seen a lot of progress over the last decade,

And, you know, translations, speech recognition are getting a lot better. But that's in part because you can tackle them with this approach of you get a lot of data, you get some powerful computers, and you throw some algorithms at the data and it just sort of does its thing. But in this case, when you're working in robotics, you have actual physical systems.

And when you're working with humans, you need to actually test and deploy these systems with humans. And you can't just have data and work at it on your computer in your office. So even though there's been a lot of hype about AI and how it's going to take away everyone's job, the fact that we don't have such robots in hospitals assisting nurses already is a real indicator that this is really challenging. And we're not about to be able to do it efficiently.

in a short time just because we've made progress on other areas of AI.

So along the same vein of how COVID-19 has really been pushing robotics and enabling robotics to become ever important, it's also been changing various other industries. For example, the beauty industry now has a much larger demand for both AI and AR, augmented reality tools, for their customers to try on because they're

they no longer have in-store testing of beauty products. And so this is pretty interesting as companies like L'Oreal and Ulta are starting to really ramp up their AR try-on tools. And so this will really be an actual push for AR technology to grow during this time. Yeah, so what this looks like is basically one of these face filters you have on Snapchat or Instagram or anywhere else.

But instead of just making some funny kind of overlay, it actually shows you what certain types of makeup like lipstick would look to help you make a decision on whether you want to try this product out. So yeah, that's just a kind of fun thing. And it does show how AI...

is becoming more prevalent in a lot of small ways and can be used in a flexible way. But let's get a little more serious again because it is a time of crisis. So, of course, we are all trying to figure out how to best combat the situation and make sure we are able to overcome it.

So the next set of articles we're going to discuss are really about how the AI research community as a whole has been talking and trying to come together to do that. So the first one we are going to discuss is from Wired and is titled Global Officials Call for Free Access to COVID-19 Research. And it says that on Friday the 13th,

science advisors from the U S and 11 other countries called on scientific publishers to make all coronavirus and COVID-19 related research freely available. So this is really interesting because, uh,

In computer science and AI research, we are very accustomed to making things open source and freely available. But this is not the case for a lot of scientific research. So this is actually an extremely big deal for research in general to make this open source and free. And so this is very exciting as a global effort of all of us collaborating and putting out what knowledge we do have. Yep. And...

And this also discusses how the call goes beyond just saying that paper should be made free. It also says that data and information should be made available in both human and machine readable formats so that us people in AI can use that data and try to help in whatever ways we can.

Yes, and I think this is an interesting point where sometimes the data, especially now with the COVID-19 crisis,

There is still a lot of private data, right? There's still PHI, there's private health information. So we do need to be very careful about that. But it is this trade-off between privacy and openness of this data to essentially help the world. And I think some countries have been hesitant in releasing that data since outside of their countries to be used on

models or to be used by other countries necessarily, not to just safeguard or just hog their own models, but actually around maintaining the privacy of their data and of their patients. So I think there is this trade-off that is occurring and, of course, some tension within the research community about it.

Yeah, so we do need to be mindful about how we go about it. At the same time, there are clear benefits to moving fast and being able to share things quickly. So in particular, this article cites one example of the open science project Nextstrain, which analyzes genomic data shared on this platform Nextstrain.

So the next train was able to confirm that COVID-19 was spreading in the Seattle area by looking at data uploaded to this service. So there are clear benefits, there are clear drawbacks as usual. Hopefully we are all able to kind of come together and agree on what makes sense and how we can be effective, but also considerate.

So the next strain, co-founder Trevor Bedford, who is a researcher at Fred Hutchinson Cancer Research Center, actually said that the outbreak of COVID-19 in Italy might be linked to a case in Munich where public health officials actually believed the virus had been contained and other people

have criticized Bedford's analysis and he soon apologized. So obviously there are downsides to this, but as long as there's open discussion, I believe that this could help move things forward. Yeah, so that's a case of maybe he should have been a little more considerate

But at least it was well-intentioned. Right. To be fair, there is a lot of papers on this topic already on bioRxiv and all sorts of different platforms right now. So it's hard to keep up on every single little one. Okay. Next we have from TechCrunch, an article about how the White House has launched the COVID-19 data hub and issued a call to AI researchers.

So, last Monday at the White House, research leaders across tech and academia and the government announced an open dataset full of science literature on COVID-19. The COVID-19 open research dataset. So, the U.S. CTO, Michael Kratios, called the new dataset, quote, "...the most extensive collection of machine-readable coronavirus literature to date."

And he also described it as a call to action for AI researchers who can employ machine learning techniques to surface unique insights in the large body of data that they have open sourced. Yeah. And to come up with guidance for researchers looking through his data, the National Academies of Science, Engineering and Medicine collaborated with the World Health Organization to come up with high priority questions.

about the coronavirus related to genetics, incubation, treatments, symptoms, and prevention? So the database brings together nearly 30,000 scientific articles about the virus. And it also brings up related viruses in the broader coronavirus group. And about half these articles make the full text available online.

critically, the database will include pre-publication research from resources like MedArchive and BioArchive that we mentioned before, and open access archives for pre-print health sciences and biology research. So note that these are pre-prints, that these are not peer-reviewed, and these are just to make sure that we can put scientific research out there as fast as possible. But of course, keeping note that these are not peer-reviewed. Yeah, it's actually really interesting. You can look up

with statistics of how much research and how many papers have been released about COVID-19 and the coronavirus. And it has been exploding in a somewhat exponential fashion, as you might expect.

So to be able to keep up with all these results, to be able to come through them, to get insights across multiple papers is quite challenging. So I think the idea here is that by combining them into one data set, by combining multiple data sources, we should hopefully be able to...

do that more easily. One thing that I think is essential now is for researchers not to overstate their claims.

One, because it's not peer reviewed work that is going to be out there and potentially affecting lots of people and lives on the line. And two, because overstating right now is really, really unhelpful. Overstating can lead to debts. And I think it should be more about limitations first, almost like you put out something and you say, this is what

my research, this is where my research starts and where it ends and where I need help. Like this is where it needs to be built off of and be made stronger. These are the holes and these are, these are basically what we need to be considering because I think I've seen some papers come out with bold claims and then I see, uh,

tweets by medical experts saying, you know, the way they analyze this was totally wrong. They made this huge claim about being able to detect COVID-19, but the way they divided the data set up was not fair at all.

it doesn't necessarily mean that they can actually detect this better than a doctor. And so I think right now, those types of bold claims are even sharper. And it matters more to be a little bit more modest as you will be affecting lives directly, not just trying to get tenure. Yeah.

Yeah, I think if nothing else this time, hopefully reminds us all to, let's say, think a little less about ourselves, think a little more about the collective good and really trying to help with whatever ways we can. And so hopefully most people keep that in mind and they don't try to rush anything out and do try to produce work that will be as beneficial as possible.

Cool. Well, this was not the only call to action. IEEE actually gave a call to action for the robotics community specifically. And so Professor Guangzhong Yang, who's the founding dean of the Institute for Medical Robotics at the Shanghai Jiao Tong University, said he was, quote, impressed by the different robotic systems being deployed as part of the COVID-19 response. There are robots everywhere.

checking patients for fever, robots disinfecting hospitals, and robots delivering medicine and food, all of which we mentioned in our previous episode. But he also thinks that robotics can do even more. And for example, robots can help with minimizing human-to-human contact, as we mentioned with Moxie, and also as a frontline tool to help contain the outbreak.

Yeah, he makes this point that we've had challenges to develop robots for crisis response. So for instance, robots are already used for things like bomb defusal, where you can't have a person go into the, let's say, bomb area to try and defuse it.

there have been actual disasters like the nuclear meltdown in Japan where people tailor-operated robots to try and go into the zones that are radioactive and people could not go there. So his point here, I think, is that we should now consider doing more of these challenges and more development specifically for this scenario or similar ones to be more prepared for the future. Yes, and coming a bit more local,

Stanford right now is hosting a virtual conference to focus on COVID-19 and artificial intelligence. This is hosted and sponsored by the HAI, Stanford Institute for Human-Centered Artificial Intelligence. And this will focus on the impact of

COVID-19 on society and how AI can be leveraged to increase the understanding of the virus and its spread. So what's cool about this conference is that it'll be live streamed and that Stanford has completely pivoted from a general conference to focusing on coronavirus very quickly. And this will be available to the public.

Yeah, I think just a couple of weeks ago, it was still planned to be in person and to have a range of topics related to HAI and the people working on it. But now it's completely readjusted. And I don't know about you, I do hope to be able to tune in and listen to the talks to kind of get an idea of what people are working on. Yeah, definitely. I would love to hear. I mean, this is going to be a cohesive conversation.

lots of different ideas jumping around kind of conference about this very intersection. So I'm very excited to hear this and see this and have it be available to everyone. And on the topic of moving the conference from being in person to a virtual one, that's been happening kind of across the field of AI. So we had a conference called iClear, which was scheduled to take place in Ethiopia.

And it has now been announced that it will be online. And this has also happened to another conference called ICML, which was to be in Vienna. But again, it's going to be entirely virtual because of the COVID-19 crisis. Another computer vision conference, Computer Vision and Pattern Recognition, or CVPR, that was set to take place in Seattle in June, is also thinking about making this virtual.

Yeah, I don't think anything has been announced, but it is one of the huge ones with something like, I don't know, 10,000 attendees probably. And so given how bad things are, it seems quite likely it will either be pushed back or will go virtual. Yes. And I am co-organizing one of the workshops at iClear this year. And we were really bummed out to see that it

It was made virtual, but we were also kind of excited in the sense that this would be a chance for us to test drive what a virtual conference might look like, what the pros and cons would be, and really think through, you know, and live through, go through, experience what that might look like.

And so it's been a whirlwind a bit in terms of pivoting, but it's been kind of exciting to see this huge conference be fully virtual and available to a lot more people.

Yeah, and I think we discussed a bit last week as well that this does have some nice benefits in terms of making it more accessible, making it more affordable, making it have less climate side effects in terms of people not flying in. So this is kind of a good forcing function to make us all try online things more, try

virtual interaction more so that we don't always fly when we can stay home or so that we can have affordable options that are remote. Right, right. I think one little unfortunate tidbit. I do, first of all, love how this helps with climate, but iClear was said to be in Addis Ababa in Ethiopia, and this would be the first international AI research conference held in Africa.

So that would have been huge. It would have been able to bring together folks in Africa who otherwise had trouble attending other AI conferences. And so it would have been accessible to them and really just hosted in their area and helped with growing the AI communities there.

So that is one little unfortunate bit. Of course, now, perhaps in a way it's, quote unquote, more accessible, but something that we've been discussing through within our workshop committees are the fact that now time zones exist. You know, our original time zones exist. So people who will convene are probably going to be from the same place as opposed to really be mixing across the globe.

Yeah, there are certainly drawbacks. And hopefully we'll consider just holding it in Addis Ababa next year or the year after. Well, regardless, thank you so much for listening to this week's episode of Skynet Today's Let's Talk AI podcast.

You can find the articles we discussed here today and subscribe to our weekly newsletter with similar ones at skynettoday.com. Please subscribe to us wherever you get your podcasts. And don't forget to leave us a rating if you liked the show. And be sure to tune in next week.