Blog Post
Most Interesting Artificial Intelligence and Machine Learning Projects in 2020
Software

Most Interesting Artificial Intelligence and Machine Learning Projects in 2020

While we’re still not at the point where we can develop artificial intelligence that’s on par with human intelligence, we’re definitely getting closer to that goal with each passing year. In the meantime, less advanced forms of AI are already used around the world by self-driving cars, virtual assistants, social networks, and even streaming services.

Companies like Tesla, Facebook, Google, and Amazon are constantly working to improve their automated computer algorithms by feeding them tons of data that they can use to make decisions and adapt to current situations without the need for a human operator. This process is known as machine learning and is considered a subset of artificial intelligence. Many people think that only the world’s biggest companies incorporate artificial intelligence into their products and services but that’s not actually the case. In fact, you likely interact with plenty of AIs on a daily basis but you just don’t know it.

With that in mind, we figured it would be interesting to put together a list of projects that most of you are familiar with but might not necessarily be aware that they incorporate machine learning. Some of these projects are fairly obvious but there are also a few rather obscure ones on this list, so join us as we take a look at the most interesting artificial intelligence and machine learning projects you can interact within 2020.

1. Google Search

We’re starting off with a very obvious project that everybody knows and loves, except perhaps the folks running Bing. We’re talking of course about Google. Although Google has developed countless projects over the years, many of which incorporate machine learning in one form or another, the company’s most important contribution is undoubtedly its massively popular search engine. Nowadays it’s pretty difficult to imagine a world without Google but that wasn’t always the case. The search engine is only 23-years-old at this point and the initial version was way less advanced than the Google we know today.

So how exactly does modern Google use machine learning? Well, there are a couple of concrete examples, one of the most obvious of which is the predictive search feature. Type in one or two words in the search bar and Google will automatically try to guess what you were planning to write next. Except it’s not technically guessing. Google’s artificial intelligence continuously gathers data about your search queries and browsing habits in an attempt to learn as much as it can about you. If you have been using Google for a while now, you’ve probably noticed that the search engine is getting better over time at predicting your next move, so to speak. That’s machine learning in action.

The fact that Google’s predictive algorithm is so successful can be attributed in no small part to its large user base. Machine learning algorithms need data in order to improve faster and Google has no shortage of it. As a result, the search engine can analyze and compile the search histories of millions upon millions of users to increase the accuracy of its predictions. When you’re looking up something you’ve never searched before, the algorithm relies only partially on your own data and partially on the search histories of other users interested in the same (or a similar) topic that matches the keywords you’re typing in.

2. Advertising and Product Recommendations

We’re lumping these two together because they work on a very similar principle. Everybody knows that digital advertising is a very common practice these days, with most websites, including massive platforms like Facebook and YouTube, relying on ads for revenue. What you may have noticed, especially in recent years, is that websites are getting better and better at serving you ads that that are often based exactly on products or services you’re already interested in.

This is once again thanks to algorithms like the ones used by Google and other search engines, which track your browsing history and feed the data to advertising networks that do their best to serve you relevant ads. This process isn’t perfect and there are a variety of other factors involved here, so there will be times when you search Google for “toothpaste” and get an ad for “wedding gowns” when you visit the next website. It’s important to note, however, that not all platforms rely on third-party data from search engines when serving ads.

Product recommendations are another good example of artificial intelligence that a lot of people interact with on a daily basis. Amazon, in particular, is renowned for its ability to recommend products that are very similar or at least somewhat related to the ones you’ve been browsing. The algorithm is so good that you don’t even have to be a regular Amazon shopper to receive relevant recommendations. It’s enough to visit the platform a couple of times and look at just a few products for the AI to start getting a picture of your interests. Among other things, Amazon also uses machine learning to recommend products that are often purchased together by other users.

3. Netflix Recommendations

We can’t talk about machine learning and recommendations without mentioning Netflix. A lot of the time you’re going to visit the platform because you want to binge-watch a specific show. However, it’s actually a lot more common for users to watch content recommended by the platform itself when they visit. Netflix starts by recommending what’s popular in your country and around the world if you’re new to the platform but as it’s often the case with these types of algorithms, you’ll get increasingly better personalized recommendations the more you use the service. It’s estimated that as much as 80% of the content consumed on Netflix is driven by these recommendations.

Netflix uses machine learning to put together recommendations based on a large variety of factors, including genres, actors, and time periods. The algorithm is very good at staying up to date with your current preferences so if you decide to watch South Park all of a sudden, you’ll quickly notice that you’ll start receiving more recommendations for animated series, even if you mostly watched documentaries or movies up to that point. Netflix also tracks your preferences based on the time of day. If you’re in the habit of watching shows in the morning and movies at night you’ll receive recommendations for content that’s similar to the one you usually consume during those time frames.

4. Personal Assistants

Pretty much everyone relies on their smartphone these days, whether it’s for calls, instant messaging, navigation, quickly looking up stuff on the internet, and so much more. But one of the most useful features of modern smartphones is undoubtedly the personal assistant. Siri and Google Now may have started out as a mere novelty back in the day, but over time the AI-powered assistants have become virtually indispensable for a lot of people. Between Siri, Cortana, Alexa, Bixby and the new (ish) and improved Google Assistant, it’s safe to say that everyone now has access to some sort of artificial intelligence right at their fingertips.

But while most of these personal assistants have been designed specifically for smartphones and home computers, some are already transitioning from the digital realm to the physical one. Probably the most well-known example at the moment is Alexa, a virtual assistant that works hand-in-hand with the Amazon Echo to bring artificial intelligence to your living room.

Alexa’s capabilities have increased tremendously since launch, from being able to perform around 1,000 functions back in 2016 to more than 90,000 in 2020. Some of the most popular ones include ordering items online, playing music, setting reminders, and answering questions, to name just a few examples. In addition, Alexa also comes with a home automation feature that allows it to interact with a wide variety of other smart devices.

5. Home Automation

Speaking of home automation, Amazon has some pretty stiff competition in this area from rival tech giant Google. Back in 2016, the company announced a competitor to Amazon Echo known as Google Home. As you’ve probably already guessed, this particular virtual assistant has similar capabilities to Echo but features better integration with other Google products and services like Nest, Google Play Music, the aforementioned Google Assistant, and of course YouTube.

While Facebook has yet to officially join the race to build the ultimate smart home assistant, CEO Mark Zuckerberg took it upon himself to build an AI that could automate his own home. Zuckerberg took inspiration from Iron Man’s Jarvis when designing the artificial intelligence and was able to connect it to various home systems like cameras, doors, lights, thermostat, and even his toaster. Among other things, Jarvis is capable of language processing, speech recognition, and face recognition. Interestingly enough, this also happened back in 2016, which seems to have been a milestone year for the development of AI-powered home automation.

6. Face ID

Going back to the topic of smartphones, another useful feature you may have on your mobile device is face ID. Now, you’ll need a relatively new smartphone in order to take advantage of this feature but it’s something that most people will be able to use in the years to come. The iPhone X and later models are prime examples of this AI-powered technology in action and was primarily meant as a form of advanced face recognition that would allow users to unlock their phones by simply looking at them. However, Face ID can also be used in other situations, such as making secure payments or in conjunction with apps like 1Password, Mint, and E-Trade.

Similar forms of biometrics have existed in the past but they weren’t very effective as they relied on a 2D picture of the user’s face. This meant devices could often be unlocked by tricking the camera with a simple picture of the user. Modern Face ID uses a depth sensor, dot projector and other hardware to capture a 3D image of the user instead of a flat one. The neat thing is that the system can adapt to changes and is able to recognize users even when they’re wearing glasses, changing their hairstyles, during bad lighting, and more. According to Apple, Face ID can adapt in this way because it’s using machine learning that allows the system to evolve over time.

7. Autonomous Vehicles

Thanks to the efforts of Tesla and other smart car manufacturers we are now one step closer to a future where people will be able to rely on artificial intelligence to drive their vehicles for them. A lot of people love driving so it’s unlikely that it will be replaced completely by autonomous cars but we’re already at the point where certain cars can take passengers from A to B with little to no input from a human user. Similarly, companies like Walmart and Amazon are now investing heavily into fully autonomous drones that can deliver products to your doorstep a lot faster than regular vehicles.

Tesla is by far the most well-known example of a car manufacturer that uses machine learning to enhance its already impressive electric vehicles. The interesting thing about Tesla cars is that they are all connected to a network and able to share information with each other. Vehicles that are part of the network learn not just from other cars but also from their human drivers. If you come across a situation where you need to perform an unusual maneuver in traffic, your Tesla will record the process and share it across the network so that all other cars can learn how to pull off the maneuver.

We can’t talk about autonomous vehicles without also mentioning commercial airliners. These were some of the first vehicles to incorporate artificial intelligence and have been using technologies like autopilot for over a century now. Sure, the autopilot of planes flying during the 1910s were based on gyroscopes rather than artificial intelligence but it’s still impressive that this technology has been around in some form or another for so long. Nowadays, autopilots are so reliable that human pilots only spend about 7 minutes on average manually flying a passenger airliner. Moreover, companies like Boeing have been working since as early as 2017 on commercial aircrafts that use machine learning to fly themselves without any sort of assistance from human pilots.

8. Commuting

It’s no secret that services like Uber and Lyft are slowly replacing traditional taxis and making life easier for both drivers and commuters. These types of services use artificial intelligence to determine wait times, plan the shortest route to your destination, estimate prices, and more. An Uber lead engineer revealed in an interview a few years back that the company uses machine learning to improve most of its services, including UberX, UberPOOL, Uber Eats, and Uber Maps. Among other things, the engineer noted that Uber would probably not be able to function as a company if it wasn’t using machine learning.

Uber drivers are certainly not the only ones who rely on artificial intelligence these days. Anybody who has ever used Google Maps can attest to its efficiency and adaptability. Not only is the service capable of showing you the fastest route to your destination based on the current state of traffic but it can also update the ETA in real time and suggest different routes in case the traffic ahead suddenly slows down for whatever reason. Google Maps is generally very accurate with its predictions because there are so many people using it and constantly feeding it new data. The service has become even better in recent years after Google’s acquisition of Waze, which made it easier for Maps to incorporate user-reported traffic changes.

9. Email Spam Filters and Categorization

Remember the Nigerian prince emails we all used to get back in the day? Whatever happened to those? You’re likely still receiving advance-fee scam emails but you probably miss most, if not all of them these days unless you check your spam folder. That’s because modern spam filters are very effective at weeding out fraudulent emails so that you never have to see them. While internet-savvy users learned to ignore such emails even back in the day, more than a handful of people have fallen for email scams over the years. So in addition to keeping annoying emails out of the way, spam filters have undoubtedly also prevented certain people from making a fool out of themselves by sending money to some con artist claiming to be a Nigerian prince.

Spam filters have existed almost as long as email clients themselves but they were initially pretty bad at their jobs. Old filters merely flagged as spam emails that contained certain keywords often associated with scams, such as “Nigerian prince”, or emails that came from untrustworthy addresses and domain names. But this meant that new phishing emails always went through until the spam filters were updated to actively look for them. Luckily, modern spam filters are a lot more advanced as they use machine learning to decide what constitutes a spam email without necessarily having to look for certain keywords, though this does remain an important component of the overall process.

Another nice feature offered by Gmail and other similar services is smart email categorization, which is also partially powered by machine learning. The system isn’t always perfect but you’ll find that regular, promotional and social emails are correctly sorted in your mailbox more often than not. You’ll also notice that if you manually move emails between folders the systems learns from it and tries to adapt to your sorting preferences in the future.

Moving forward, companies like Google promise that email clients will get even smarter, to the point where your inbox will be able to automatically reply to emails for you in a believable manner. Granted, smart reply has already existed for a number of years but has only been capable of creating fairly generic messages so far. That won’t be the case in the near future, though. Thanks to machine learning, email clients are constantly learning how to compose complex and personalized responses from human users. It’s only a matter of time before your inbox will be able to send email replies that are indistinguishable from your own with the help of artificial intelligence.

10. Social Networking

It’s pretty safe to say that social networks wouldn’t be the same without artificial intelligence. Everything from Instagram’s auto-suggest emojis to Snapchat’s facial filters uses machine learning in some form or another. But one of the most well-known features that falls into this category has to be Facebook’s facial recognition software. All you have to do in order to see the technology in action is upload a picture of you and some of your friends. The system will automatically attempt to identify everyone in the photo and will even suggest you tag the people it manages to recognize. Over the years Facebook acquired a number of startup companies that work with facial recognition software in order to improve its own system. A few notable examples include Face.com, Faciometrics, and Masquerade.

In addition to facial recognition, Facebook is also using AI to personalize every user’s newsfeed. The algorithm does its best to prioritize posts that it thinks would interest you most based on your browsing history and interactions. The same goes for page suggestions, friend suggestions, and of course, ads. As mentioned earlier, targeted advertising and artificial intelligence work hand-in-hand. This makes complete sense as you are much more likely to click an ad if it’s relevant to your interests, which means companies like Facebook are constantly pushing their machine learning software to give you personalized advertising.

Since we’re talking about Facebook we also have to mention DeepText, an AI-powered text understanding engine. The goal here is to develop a system that can understand conversions nearly as well as a human. In fact, even better in some ways because DeepText can understand more than 20 languages and can process thousands of posts per second. This feature of Facebook is a bit more subtle than, let’s say, facial recognition but you may have seen it in action if you’re using Messenger or if you’re creating a post where you’re trying to sell a product or a service. In essence, the aim of DeepText is to point you towards already existing tools that can make certain activities, such as selling items or looking for a taxi, a lot easier for the user.

Although we mainly focused on Facebook in this section, it’s far from the only social network that uses machine learning. Another neat application of AI that has been gaining a lot of popularity in recent years is Snapchat’s Lenses. Just in case you’re not already familiar with them, Lenses are animated filters that add a wide variety of digital effects to your face. The filters do a good job at sticking to your face even when you’re moving and they’re advanced enough that they can distinguish between the faces of various animals. If you ever wanted to see what your dog would look like with a moustache or a wig all you have to do is open Snapchat and check out the Lenses feature.

11. Spotify’s Discover Weekly

Spotify is by no means the only streaming service that offers music recommendations but it seems to be better than most at it. The Discover Weekly feature, in particular, is remarkably good at finding music that you are actually likely to enjoy. And I’m not just talking about songs from artist you’ve listened to in the past on Spotify either. In fact, the goal of the system is to primarily introduce you to music you’re likely not familiar with, but is right up your alley. The system isn’t that impressive when you first sign up for Spotify but keep adding songs to your Favorites and create a few playlists and you’ll soon notice that the AI learns over time to pinpoint your specific taste in music.

In addition to Discover Weekly, Spotify also presents you with daily personalized playlists and new releases you may have missed. Services like Google Music try to do something similar but Spotify’s system is probably the best right now because it has the most amount of data to work with. This wasn’t always the case but ever since Spotify was made available in more and more countries, its user base has increased exponentially. Similar to other streaming services, Spotify compares your music preferences to those of other users when making recommendations. As even more people sign up for Spotify, the recommendations you’ll get are likely to become even more accurate.

12. Video Games

Video games have been using artificial intelligence pretty much since their inception and are one of the best examples of how machine learning has evolved over time. A very important milestone in AI research came in 1997 when IBM’s Deep Blue computer defeated Garry Kasparov, the reigning chess champion at the time, at his favorite game. While Kasparov managed to take the win during their first encounter, Deep Blue came out on top a few months later when the two had a rematch. Deep Blue was considered very advanced for its time but it can’t hold a candle to the types of AI you’ll come across while playing modern video games.

Skip forward a few decades and you may be shocked to find that artificial intelligence is already defeating professional eSports players at virtually every game. And we’re talking about human players going up against regular bots either. A couple of years back. Elon Musk’s AI company revealed a neural network designed to go head-to-head with the best Dota 2 players in the world. Dubbed OpenAI Five, the machine learning project was only capable of playing 1v1 matches during its first public appearance in 2017. Just one year later the AI advanced to the point where it was able to take on and defeat teams of some of the best Dota 2 players in the world in 5v5 matches. OpenAI Five was open to the public for a brief period in 2019 when it was challenged by players all over the world in online matches. The artificial intelligence managed to win over 90% of the nearly 43,000 games it played as part of the event.

Another famous example of artificial intelligence surpassing pro gamers is DeepMind’s (owned by Google) project AlphaStar. The goal of AlphaStar was similar to that of OpenAI Five, however, this particular AI arguably had an even bigger challenge up ahead, as it was designed to go up against professional Starcraft 2 players. SC2 is a notoriously punishing real-time strategy game where, until recently, competitive high level play was too much to handle even for an AI. That all changed in late 2019 when AlphaStar became the first AI to reach the rank of Grandmaster, the highest possible rank in Stacraft 2. AlphaStar is currently only able to win around 99.8% of games so there are still a few human players that can best it, but that will probably change in the near future.

While artificial intelligence in becoming better and better at multiplayer games, AI-controlled enemies you can now face in single-player are very impressive in their own right. The Nemesis system found in the Middle-Earth series is particularly noteworthy and innovative. The system allows enemies to remember the interactions you had with them in the game and adapt to your playstyle in order to become better prepared for your next encounter. And since all enemies are procedurally-generated, you can expect to face entirely new baddies every time you play the game. Moreover, enemies have their own hierarchy and squabble amongst themselves for supremacy regardless of your involvement. This allows them to evolve and gain new abilities as the game progresses.

Another noteworthy example of how artificial intelligence helped revolutionize gaming can be found in 2005’s F.E.A.R. This was one of the first video games to feature a system capable of generating context-sensitive behaviors. In other words, the AI was able to adapt to its surroundings and would try to use the environment to its advantage whenever possible. Among other things, enemies in F.E.A.R. can take cover behind tables and other objects, use flanking maneuvers, suppressive fire, use grenades to flush out hiding players, and more.

What set F.E.A.R. apart is the fact that its enemies were remarkably good at all of these actions. A similar system can also be seen in the S.T.A.L.K.E.R. series where opponents are capable of advanced tactics in combat, including out-flanking players, healing wounded allies, switching weapons mid-combat to best suit the current situation, and more.

13. Banking and Finance

Money makes the world go round so it’s no surprise that artificial intelligence has slowly but surely made its way into the banking and finance sector as well. In fact, this has been the case for quite some time now and you may have already noticed. If you’ve ever made a transaction that your bank thought was suspicious, chances are you were notified by it via text or email immediately after. Banks use many automated systems that are not necessarily AI-based, however, they do need machine learning software to figure when someone is trying to commit fraud. These systems tend to only warn customers in the case of mildly suspicious transactions but they can also outright block transfers if necessary.

Perhaps somewhat ironically, trying to contact your bank because of a blocked transaction will often lead you to interact with another automated system. If you’re contacting customer support via live chat these days you’re pretty likely to run into an AI-powered chat bot. Current bots are more than capable of helping you with many tasks but banks will eventually switch you to a human representative if the bot isn’t able to help fix your issue by itself. Needless to say, chat bots aren’t used solely by banks and you’re likely to come across them when interacting with a large variety of other services.

14. Surveillance Systems

Nobody likes the idea that governments and corporations might be monitoring our lives on a daily basis but there’s no denying that surveillance does play a big role in today’s society. We could spend all day debating whether that’s right or wrong but what we really want to talk about is how and why surveillance systems are mainly operated by artificial intelligence these days. As with everything else on this list, using AI for surveillance simply makes things easier and more efficient. Spending all day looking at security cameras may sound like an easy enough job, but that’s not really the case when you have to monitor several different feeds at the same time. However, that stops being an issue if you automate the system.

Another important reason for why institutions may want to rely more on AI than human operators is because automated systems can take advantage of technologies like facial recognition. Needless to say, being able to identify someone in real time amongst a crowd of people can come in handy in a wide variety of situations. And, realistically speaking, that’s not something that a human can achieve. At least not consistently. For now, artificial intelligence is working hand in hand with humans because it’s not always reliable but itself, but that will probably change sooner than we may think.

15. Plagiarism Checkers and Essay Grading

Back in the day students had to visit their local library and gather sources the old fashioned way in preparation for writing a paper or an essay. Thanks to the internet, that’s usually no longer the case because you can find pretty much all the sources you will ever need online. But while the internet makes life easier for students, there’s also a lot of temptation to take the easy way out and simply plagiarize existing content. That puts teachers in a difficult position as checking essays for plagiarism one paragraph at a time is a very time-consuming process. Luckily, teachers now have tools at their disposal known as plagiarism checkers that can do most of the work for them.

Most programs like this detect plagiarism by simply checking the text against a massive database of articles, books and essays, so there’s no real artificial intelligence involved. However, there are some tools that have started to use machine learning in recent years in order to detect plagiarism even without access to a database. The algorithm does require an initial sizeable sample of texts during the learning process but can then identify plagiarism with a fairly high degree of accuracy without one. While they’re not 100% reliable just yet, AI-powered plagiarism checkers are constantly improving and will eventually be able to tell if the author plagiarized text from other languages or even text that isn’t available online.

Robo-graders are another powerful tool that can make teachers’ jobs at lot easier. Grading the essays of dozens or even hundreds of students takes a very long time, which is why researchers came up with scoring engines that use artificial intelligence to grade essays. Probably the most well-known example of such a scoring engine is the Educational Testing Service’s (ETS) e-Rater. ETS is a nonprofit educational testing and assessment organization that has been using e-Rater for many years now as part of its online writing evaluation service known as Criterion.

The Criterion system uses both e-Raters and human scorers in order to achieve maximum accuracy. The e-Rater is capable of analyzing and providing feedback in key areas like grammar, mechanics, style, usage, and organization & development. The AI and the human scorer grade each essay separately and then the results are compared to make sure that they are in close alignment with each other. In case of a noticeable discrepancy between the two scores, an additional human reader is called in to settle the dispute. The ETS says that e-Rater isn’t meant to replace teachers but rather to save time and provide additional feedback to students. That said, it’s easy to envision a future where artificial intelligence will be able to reliably grade essays without any human involvement.

Final Thoughts

The idea that artificial intelligence could become a big part of our day-to-day lives may seem pretty scary until you stop to consider that this has already been happening for quite some time now. Granted, the AI-powered systems of today aren’t as advanced as those found in certain SF movies, but some of them are likely the precursors of more sophisticated networks we’re bound to see in the years to come. In fact, people like Bill Gates, Stephen Hawking and Elon Musk have expressed their concerns about where AI research may lead us on multiple occasions over the years. So far, their concerns have remained mostly unheeded and it’s easy to see why.

For better or worse, artificial intelligence has changed our society in significant ways and has simplified a lot of activities. Judging by Sophia and the bots developed by Boston Dynamics, it’s pretty safe to say that advanced AI-powered robots aren’t exactly right around the corner. However, in 2020 artificial intelligence is certainly everywhere around us in digital form, whether it’s helping you drive your car, unlock your phone, discover new music or entertaining you while you’re playing your favorite video games.

Related posts

Leave a Reply

Required fields are marked *

Copyright © 2020 Blackdown.org. All rights reserved.