Author Archives: Darya Shabunevich

16 February 2023
Leave a comment
AI, Security

Instruction for hacking an electric car

Reading Time: 3 minutes
What Happens When Hackers Hijack Your Car ... While You're in It · TeskaLabs Blog

Our cars are getting smarter from generation to generation, however, unfortunately, this also applies to thieves who intend to steal them. Hackers have already found vulnerabilities in electric vehicles, and their attacks can have serious consequences. As electric vehicles become more sophisticated and connected to the Internet, it is expected that the risk of hacking and cyber attacks will only increase.

One of the main problems with electric vehicles is that they are equipped with numerous sensors and controllers connected to the Internet, which makes them vulnerable to cyber attacks. Hackers could potentially gain remote access to these systems and manipulate them in a way that could cause serious harm, such as disabling brakes or changing steering. In some cases, hackers could even take control of the entire car, putting the driver and passengers at risk.

Another problem is that hackers can attack charging stations for electric vehicles. These stations are also connected to the Internet and often use wireless connectivity to connect to electric vehicles. Hackers can potentially gain access to these systems and manipulate them.

Recently, a security expert discovered a way that allows two attackers to unlock, start and drive away a Tesla Model Y electric car in a matter of seconds.

Hackers specializing in hacking Tesla electric vehicles have identified a vulnerability that allows them to hack NFC relays. However, not everything is so simple: in order to hack the system, thieves need to work in pairs and get close to the NFC chip or smartphone. Josep Pi Rodriguez of the Seattle-based computer security firm IOActive found that attackers could use Tesla’s key technology called NFC (Near field Communication) to gain control of a vehicle, designed to give car owners the ability to access them by touching an NFC card to the middle rack. Rodriguez found that if one thief approaches a critically small distance to the driver when he gets out of the car, for example, to a store or bar, and the other is standing by the car, it will be possible to open the door and start the car.

  • Here’s how it works: a thief standing at the car uses a special device to convince the car to send a “call” to the driver’s NFC card, but then transmits this call via Wi-Fi or Bluetooth to a mobile phone belonging to a second thief, who is watching the driver at this time. The second thief keeps this phone near the driver’s pocket or bag where the NFC card is stored, and when this NFC card responds, its signal is transmitted to the thief standing by the car via a mobile phone.

Tesla previously required drivers using an NFC card (not a keychain) to unlock their cars to place the card between the front seats in order to turn on the transmission. But after a recent software update, this requirement has been lifted. Tesla also offers the option of using a PIN code, which means that car owners must enter a four-digit code before starting the car, however, a fairly small number of owners activate it. In the end, even if this additional protection prevents thieves from leaving by car, they will still be able to use the hacking method described above to open the doors and steal any valuables inside the cabin.

To solve these problems, automakers and cybersecurity experts are working to develop stricter security measures for electric vehicles. This includes the introduction of more advanced encryption technologies, the development of secure firmware and software, as well as regular updates and patching of systems to eliminate any vulnerabilities found.

Electric car owners can also take measures to protect their cars from cyber attacks. This includes regularly updating software and firmware, using strong passwords and two-factor authentication, and not using public Wi-Fi networks when accessing the Internet from their cars.

Thus, as electric vehicles become more popular and widely used, the risk of cyber attacks and hacking is expected to increase.

This is a serious problem that could have serious consequences for electric vehicle owners, automakers and the general public. It is important that automakers and cybersecurity experts work together to develop stricter security measures for electric vehicles, as well as educate owners on the steps they can take to protect their cars from cyber attacks.

Sources: https://www.aljazeera.com/amp/economy/2022/1/12/teenager-says-he-remotely-hacked-into-more-than-25-teslas

https://www.indiatimes.com/amp/technology/news/tesla-cars-may-be-unlocked-and-hacked-using-bluetooth-devices-researchers-find-569861.html

https://www.entrepreneur.com/business-news/this-hacker-exposed-a-new-way-to-steal-a-model-y-tesla/435323

Tagged , ,
15 February 2023
1 Comment
AI, Business

Expensive Google’s mistake

Reading Time: 2 minutes
Google Bard AI mistake just cost Google over $100 billion | Tom's Guide

 Let’s start with the fact that the inhabitants of the world of information technology did not have time to discuss the new chat bot ChatGPT, which they had already dubbed the “Google killer”, as a competitor launched its own bot Bard, also based on the principle of artificial intelligence.
Bard is based on Google’s already existing large language platform called Lama, which is said to be so human-like in its responses that it seems reasonable.

On the day of the presentation of a new chatbot, Google makes a fatal mistake that costs them $100 billion.

On February 6, Google announced its AI chat-bot Bard, a ChatGPT competitor from OpenAI, which will be available in the coming weeks. However, in the promotional video, the technology made a mistake by providing false information as a result of the request.

Google has published a GIF in which the chatbot answers the question: “What new discoveries of the James Webb Space Telescope can I tell my 9-year-old child about?”. Bard offered three options, including the claim that the telescope “took the first photographs of a planet outside the solar system.”

Astronomers immediately reacted to the presentation and NASA wrote about this situation on their website. The thing is that the first picture of the exoplanet was taken back in 2004, which indicates errors in the video presentation of the chatbot.

“I’m sure Bard will still amaze us, but for the record: JWST did not take the “first image of a planet outside our solar system,” astrophysicist Grant Tremblay wrote on Twitter.

Bruce McIntosh, director of the UC Observatory, also pointed out the error: “I speak as someone who photographed an exoplanet 14 years before the launch of JWST. Don’t you think a better example could be found?”

Tremblay tweeted that the biggest problem with chatbots like ChatGPT and Bard is their tendency to confidently provide incorrect information as fact. And often they just come up with data, because in essence they are autocomplete systems.

Everyone is well aware that there is already a lot of false information on the Internet, but everything is complicated by the desire of Microsoft and Google to use these tools as search engines. Where the answers of chatbots acquire the authority of a smart machine “know-it-all”.

The promotional video with the error was viewed by 1.6 million people on Twitter. And almost immediately after its publication, Alphabet shares fell by 9%, and the market value decreased by $100 billion.

Sources and references: https://www.bbc.com/news/business-64576225.amp https://www.theverge.com/2023/2/8/23590864/google-ai-chatbot-bard-mistake-error-exoplanet-demo https://amp.cnn.com/cnn/2023/02/08/tech/google-ai-bard-demo-error/index.html

Tagged ,
8 February 2023
Leave a comment
AI, Open Source, Programming languages

ChatGPT’s new competitor

Reading Time: 3 minutes

More powerful than ChatGPT': Microsoft unveils new AI-improved Bing and Edge browser | ZDNET

Bing is an updated Microsoft search service based on artificial intelligence. It’s based on the OpenAI GPT language model, but Bing is newer than ChatGPT 3.5. Microsoft says it’s not just an updated search engine, but a new artificial intelligence-based search channel with a new chat interface that offers better searches, more complete answers and more relevant search results, so readers can spend less time on the page. Artificial intelligence will revolutionize every category of software, including the largest category — search. Bing can also create content and inspire creativity. Microsoft said: “The new Bing can generate useful content. Create a 5-day itinerary for your dream vacation to Hawaii, including links to write emails, book travel and accommodation, prepare for interviews and create quiz questions to help you The new Bing also cites all sources so you can see links to the web content you link to.”

Microsoft has also announced changes to Edge. Artificial intelligence has been added to Edge to help people do more with search and the internet. As for the new Bing search and the new Edge browser, Microsoft highlights some key features:

  • The best search. The new Bing offers an improved version of familiar search, providing more relevant results for simple things like sports scores, promotions and weather, as well as more complete when you need it. It also provides a new sidebar for displaying responses.
  • Full answer. Bing searches the web for results to find and summarize the answers you are looking for. For example, you can get step-by-step instructions on how to replace eggs with another ingredient in your current cake without looking at multiple results.
  • New chat. For more complex searches, such as planning a detailed travel itinerary or choosing a TV to buy, the new Bing offers a new interactive chat. Chat allows you to narrow down your search until you get the full answer you are looking for, asking for details, clarity and ideas. Links are available, so decisions can be made immediately.
  • New Microsoft Edge interface. We have updated the Edge browser with new artificial intelligence features, a new look and added two new features: chat and messaging. Use the Edge sidebar to request summaries of long financial reports to get the main conclusions, use the chat function to request comparisons with competitors’ financial reports, and automatically place them in a spreadsheet. You can also ask Edge to help you create content, such as posts for LinkedIn. Then you can get help updating the tone, format and length of your message. Edge can understand the web pages you are viewing and adapt accordingly.

However, Google issued a warning to its departments, and even the founders and shareholders of the tech giants Larry Page and Sergey Brin stepped up. On Monday, the company introduced its own alternative to ChatGPT called Bard. Google CEO Sundar Pichai called the software an “experimental artificial intelligence service” that is still being tested by a limited number of users and employees of the company and will be released to the general public in the coming weeks.

Microsoft Brings ChatGPT-Like AI Features to Bing, Edge - My TechDecisions

Thus, Bing have been developed to facilitate research and increase their reliability. Starting with the chat mode, you can ask literally any question using an interface very similar to GPT chat, and the answer will be sent in seconds.

Interestingly, when searching for information in real time on the Internet, responses are sent directly from various thematic sites. The source of information for constructing the answer is shown as a footnote, but the user is redirected to the main page of the site in question, and not to the page with the text.

Sources and references: https://blogs.microsoft.com/blog/2023/02/07/reinventing-search-with-a-new-ai-powered-microsoft-bing-and-edge-your-copilot-for-the-web/ https://habr.com/ru/news/t/715508/

Tagged , ,
17 January 2023
Leave a comment
AI, Devices

Neurochips Synchron VS Elon Musk’s ambitions

Reading Time: 4 minutes

Synchron company became the first developer of brain-computer interfaces, which introduced its Stentrode implant into the head of an American. Now a person who has long lost the ability to move and speak will be able to control a web browser and communicate via email using the power of thought. He became the fifth person in the world with an implanted neurochip.

The new Synchron project was implemented under the guidance of specialists from Mount Sinai West Medical Center. Neurointerventional surgeon Shahram Majidi made an incision in the patient’s neck and inserted a miniature sensor into the brain. Stentrode — a 1.5—inch-long device consisting of wires and electrodes – opened and began to fuse with the outer edges of the blood vessel. According to Majidi, this procedure was very similar to implantation of a coronary stent, so it took only a few minutes.

Electrode matrix Stentrode implanted into a blood vessel inside the skull

When the operation was completed, the sensor was connected via a physical cable to a computing device previously implanted in the patient’s chest. To do this, surgeons have created a “tunnel” for wires and a pocket for the device under the skin — on the same principle as modern pacemakers are installed. After that, the neural interface was activated and connected to the computer. After 48 hours, the patient has already gone home.

ALS, also known as motor neuron disease is a rare but terrible disease: the atrophy of neurons gradually disconnects the periphery of the brain in the form of muscles. What it does to a person can be represented by the example of the most famous patient of this scourge in our time: Stephen Hawking.

The principle of operation of Stentrode is relatively simple. The sensor reads the signals activated by the patient’s brain and transmits it to the device under the skin. The latter interprets the activity and sends a ready-made command via Bluetooth to a computer or smartphone. But how does the cursor move? Here lies some weakness of the current Stentrode model: the choice of a place to press and move the cursor is not carried out by nerve impulses of the brain directly, but by means of a fair “crutch”: an eye tracker, an eye tracking system like the one that was built into the famous Stephen Hawking chair and allowed him, in fact, to work on a computer and communicate online. Only without the neurointerface yet. In this way, the user can switch the tab in the browser, open the email application, type a text and send it to the selected recipient.

And still Synchron managed to get ahead of its main competitor: the ambitious Neuralink of Elon Musk, who promised to start experiments on human volunteers back in 2020, but to this day has not received permission from the FDA. And the procedure itself from Elon Musk’s project looks much bolder, due to a much more extensive system of sensors and electrodes, it is going to allow the patient much more than Stentrode, but at the same time — more brutal. Instead of a neat passage through the incision on the neck, the patient should cut out a part of the skull by means of a robot surgeon, and after a complex automated operation with direct connection of electrodes to the brain, replace it with an electronic unit.

However, this approach has been used for two decades by the Utah Array (BrainGate) technology from Cyberkinetics, which Musk himself calls “something like an instrument of torture.” This technology has already been developed so much that it is used in more than five hundred laboratories around the world. It allows patients to make rather complex movements with electronic limbs by means of electrodes embedded in the motor areas of the brain. It also provides the ability to move the cursor across the screen with an effort of thought, but requires a very complex neurosurgical operation and cumbersome equipment. Plus, it has quite negative consequences for the health of patients (the body rejects too large and roughly embedded electrodes and sensors. Channels overgrow and block contacts).

Synchron CEO Thomas Oxley holds the endovascular matrix of Stentrode electrodes during TED 2022

The company hopes that the devices of the Stantrade series will be able to become the first suitable for use at home, and not only in the walls of a specially equipped laboratory or ward, like BrainGate. And many surgeons will be able to perform it in almost ordinary hospitals, and not specially designed robots, like Elon Musk’s, or a whole team of specialists with the most complex equipment, as in the case of BrainGate.

Resources: https://habr.com/ru/company/ruvds/blog/679652/ https://en.wikipedia.org/wiki/Stephen_Hawking https://www.cnbc.com/2022/12/01/elon-musks-neuralink-makes-big-claims-but-experts-are-skeptical-.html

Tagged , ,
10 January 2023
3 Comments
AI, Devices

Cyberdogs

Reading Time: 3 minutes

OPPO posted on its YouTube channel a teaser dedicated to the new QRIC robo-dogs. This is an abbreviation for Quadrupedal Robotic with Intelligent Controller, which translates as “four-legged robot with an intelligent controller”. This companion robot is likely to compete with the Boston Dynamics Spot and Xiaomi CyberDog models.


The video is about a boy who plays with QRIC and builds a toy space rocket. The video demonstrates some of the capabilities of the robot, which acts as an electronic pet. Robodog is capable of:

  • walk
  • jump
  • tip over
  • carry items
  • recognize people using sensors placed in front of the case
  • act as a webcam for video calls. Also QRIC can be controlled by voice commands or from a smartphone app.

Competitors
There are few such robots on the market today. The first competitor is the Xiaomi CyberDog robot dog, which was released in a limited edition of 1000 pieces at a price of $ 1,500. It is likely that the price of QRIC will be in the same range.
As for the Boston Dynamics Spot, it is difficult to call it a direct competitor, because it is much more technologically advanced, more expensive and has a completely different scope, although it can also be used as a very expensive toy.

The difference in these three robots is that the QRIC robot is presented as a toy and a home helper and companion for children and parents. If we talk about the Xiaomi robot, then we most likely will not see the difference with QRIC. So, the principle of operation of these robots is the same, they are able to control smart home devices, withstand a load of up to 3 kg, do various tricks and walk next to the owner. But if we rely on the video, then the difference lies in the fact that QRIC is still a family assistant who can be left with the child and who can connect the baby with the parents at any second. And Xiaomi presents their product as a replacement for a pet that can follow you, perform various commands and tricks. Rather, Xiaomi’s cyberdogs are more of a toy than QRICs.

And if we take Boston Dynamics, then it is very different from the previous two cyberdogs, in that it is designed to work, to help with hard and dangerous work. Itcan also play the role of a pet, but more expensive and not able to perform tricks of such a plan as Xiaomi or QRIC.

In fact, for some people, such robots may already be more useful than a live dog. Robots can be programmed to do anything. Protect the territory and report to the owner how things are, when he is on vacation, even in another country. Show everything in real time. The robot can carry the load.
However, in my opinion, robots will never replace dogs, no matter how functional they are or at least we can’t compare a dog that has feelings and a robot that just does a programmed tasks.

References:

https://youtu.be/H1FgiAaA0Ck https://itc.ua/news/xiaomi-predstavila-chetveronogogo-robota-cyberdog-za-1540-on-mozhet-delat-salto-nazad-i-sledovat-za-hozyainom/ https://www.forbes.ru/newsroom/tehnologii/437091-xiaomi-pokazala-svoego-pervogo-robota-sobaku https://habr.com/ru/amp/post/521878/

Tagged , ,
19 December 2022
3 Comments
AI

Is Tesla’s autopilot killing people?

Reading Time: 4 minutes
Подтверждено: по вине «автопилота» Tesla погибло уже четыре человека - Hi-News.ru


If we start talking about statistics, then it all started much earlier. But it was in 2018 that 21 people died due to accidents caused by the Tesla autopilot (more than in previous years), next year the number of autopilot victims increased to 50 people. Someone will say that then the autopilot was “raw” and did not work properly, but unfortunately the deaths continue. In 2021, it is also possible to state deaths as a result of driving a Tesla with automatic control. In April of this year, 2 people in the United States crashed a Tesla Model S when the electric car was controlled using autopilot. And one of the most recent serious accidents occurred in November of this year, when a Tesla accelerated to 200 kilometers per hour, killed three road users and finally crashed into the stage, during which the driver was hospitalized. The electric car did not react to the brake pedal. Presumably, there was a software glitch in the electric car. Representatives of Tesla expressed their willingness to assist the investigation and said they would actively provide the necessary assistance.

Elon Musk acknowledged the problem

Only in August 2021, Tesla developer Elon Musk admitted that there were serious problems with the operation of Tesla’s autopilot. At the same time, Musk clarified that version 9.2 is problematic, and allegedly in 9.3 the problems will be fixed. But people were dying in Tesla even before version 9.2 was released…

In defense of Autopilot and Tesla

On the one hand, Tesla continues to kill people, but on the other hand, drivers are guilty of not following the instructions. The instructions say that the driver must constantly keep his hands on the steering wheel and, if necessary, in the event of a dangerous situation on the road, immediately take control. In other words, owners are informed that it is impossible to rely entirely on Tesla’s autopilot.

Criticism of autopilot

On the other hand, if a company that aspires to be called the best electric car developer in the world and dictates the fashion in the electric car industry creates an autopilot function, then it should work flawlessly. Otherwise, what is the point of autopilot if it needs to be controlled 100% of the time? Whether with or without autopilot, the driver must constantly monitor the road. But then why do we need such an autopilot? Or are we still seeing a demo version, some developments, but why did they then come out with a finished product?
Usually, automakers recall entire batches of cars if something does not work correctly, and here the node that is responsible for the safety of the driver, passengers and other road users does not work correctly! At the same time, Musk does not even think of programmatically disabling autopilot, and moreover, does not call for refusing to use it. Saying that the next version will be better, he directly appeals to the owners of Tesla with a request to pray for their lives and wait for a miracle update.

This is not the first time the author of the Dan O’down Media channel has criticized Tesla, including for their desire to call the car control assistance system “autopilot”. He conducted an experiment, proving that Full-Self Driving is capable, even in simple conditions, of causing serious injuries or even death to the latter when a car and a pedestrian collide. The test was conducted in clear weather, the car was moving strictly in a straight line. A dummy of a child was installed on the Tesla path, which was static, and did not move or appear suddenly. In all three cases, the autopilot failed to recognize the obstacle and crashed into a human figure. Moreover, the system did not even attempt to slow down, which indicates the “blindness” of Full-Self Driving in this case. Or about a factory marriage?

Tesla “sees” the world around it with the help of stereo cameras that transmit a picture of what is happening “here and now”, and does not rely on pre-built 3D maps of the area. Well, Dan O’downe’s experience proved that the Tesla system has a serious flaw. He recalled that more than one hundred thousand Tesla cars drive on the roads, whose drivers completely rely on Full-Self Driving. Tesla may make a mistake “at the most inopportune moment” on those streets where children go to school.

Автомобили Tesla убивают людей», — за эти слова Huawei уволила главу отдела беспилотного вождения

Here are two more examples of Tesla’s autopilot does not work properly.

Last month, two cases were recorded when Tesla cars provoked an accident on the motorway at night. Both cases ended with the death of motorcyclists. It turned out that in some situations, automatic driving systems incorrectly determine the road situation, and do not “see” motorcycles.
It is assumed that an automated driver assistance system is to blame for both accidents. In both cases, the motorcyclists were riding in front of the cars, and the car simply caught up with the motorcycle and crashed into it.
Michael Brooks, executive director of the Center for Automotive Safety, called on the NHTSA (National Highway Traffic Safety Administration) to recall the Tesla autopilot because it does not recognize motorcyclists, ambulances or pedestrians. “It’s already clear to me and to many Tesla owners that this thing is not working properly, will not meet expectations and is endangering innocent people on the roads,” he said.

The error of the autopilot or the driver is up to you, but so far the cases and statistics indicate that there are clearly questions with autopilot. As a Tesla car is not bad, compact and eco-friendly, but is it worth trusting the car with your life, taking your hands off the steering wheel and enjoying the ride?

Sources: https://eco-drive.biz/avtopilot-tesla-ubivaet-lyudej/ https://impakter.com/tesla-autopilot-crashes-with-at-least-a-dozen-dead-whos-fault-man-or-machine/ https://www.tesladeaths.com/index-amp.html

Tagged , ,
10 December 2022
Leave a comment
Business, Devices

Ray-Ban Stories

Reading Time: 3 minutes

The social network Facebook showed smart glasses. They are released together with the Ray-Ban brand and cost $ 299.

Facebook Luncurkan Kacamata Pintar Ray-Ban Stories, Bisa Memotret hingga Dengar Podcast

The device looks like ordinary glasses, but is equipped with several interesting functions at once. Firstly, the device can take photos and videos using two built-in 5-megapixel cameras. The duration of the video is no more than 30 seconds, and in total about 30 short videos or 500 photos will fit on the built-in memory of the glasses. A special feature is that one camera writes square videos, and the second takes photos. For this reason, the picture is slightly shifted to the side. This can be a problem when shooting in the first person. And the image quality is an order of magnitude worse than that of smartphones five years ago.

The glasses allow you to play music from your smartphone through the speakers built into the shackles, receive phone calls and interact with the Facebook View app. The View application is extremely simple, in it you can see everything taken with glasses, synchronization occurs instantly and without your participation. The application has a small and simple editor, you can edit your pictures or videos, create collages. The device connects to smartphones on the iOS or Android operating system using Bluetooth.

However, the battery of the glasses is not happy with its autonomy yet – they work for about 6 hours and on average discharge by 20% in just an hour of use. It takes only one hour to fully charge the gadget. The kit includes a USB Type C cable.

Glasses can be used by people with both good and poor eyesight – in the latter case, you need to consult an ophthalmologist to know which lenses to choose when ordering. Depending on the type of lenses, the price also varies: from $ 299 to $ 379.

The idea to make Ray Ban Stories came about thanks to one email in 2019. In it, the head of the Luxottica Group company offered to make such glasses. The addressee was Mark Zuckerberg, who supported the development.

Facebook does not expect that these simple cameras will replace people’s smartphones. “Our mission is to help create tools that will help people feel connected anytime, anywhere. We want to create a sense of social presence, a feeling that you are next to another person living in the same space, regardless of physical distance,” they say on Facebook. The cooperation between the two companies looks interesting, as they combine their strengths — Ray-Ban models are popular and well-known in the market. The global meaning of launching glasses is that Meta wants to accustom the audience to hand-free devices in this way. This will make it easier for people to get used to full-fledged glasses and VR and AR helmets. Ray Ban Stories in this case serves as a transitional gadget.

The concept of Facebook and Ray-Ban collaboration strongly resembles Snapchat — Spectacles glasses released in 2016. Snapchat released the third generation of its glasses, but they failed to attract the attention of the target market. In 2017, the company wrote off the value of unsold inventory by $40 million.

Do you know what they were able to do in Ray Ban and what is really cool? These glasses look like glasses, and it’s hard to recognize a gadget in them. Symmetrical cameras on the frame look like a design element, and that’s all. The LED warning that you are shooting something is not very bright, and you do not pay attention to it. ”Almost no one pays attention that you are recording or filming something. People are not used to the fact that ordinary glasses can have cameras (the same Google Glass looked different, and many have heard about them). The main thing is to learn how to hold your head properly, otherwise you get a picture that dances after your gaze,” – says one of the users of this device.

Would you try such glasses for $299?

Resources:

https://vr-store.com.ua/ray-ban-stories-detail https://www.ray-ban.com/usa/ray-ban-stories https://www.androidpolice.com/ray-ban-stories-review/ https://www.forbes.com/sites/moorinsights/2022/04/28/ray-ban-stories-review-comfortable-but-limited/?sh=46abc07364ca

Tagged , ,
21 November 2022
3 Comments
AI

Inhuman music

Reading Time: 3 minutes

In the 21st century, artificial intelligence can do everything. Painting pictures, driving cars, helping doctors in medicine, and what about music? Does AI know how to compose music and write lyrics for songs?

In truth, an artificial intelligence can do this, too. Not with a soul, not with such a huge meaning, as a human does, because a robot has no feelings, but still knows how and even writes lyrics for music.

How exactly do neural networks create music? The general principle is that the neural network “looks” at a huge number of examples and learns to generate something similar. But it is impossible to formulate a task for a neural network to write beautiful music, because it is impossible to create a formula that will fulfill this task, since this is a non-mathematical requirement. It is interesting only when the neural network reproduces something that exists. The approach by which this music is created is called an auto-encoder (Generative Adversarial Network). It works like this:
We compress the music at the input into a very compact representation and then expand it back to its original form. A compact representation does not allow you to remember everything completely that was in the music. Therefore, the neural network is forced to put some common properties for music into the software part. And then, when generating music, we take a random sequence of numbers, apply the rules about the knowledge of music learned by the neural network and get a piece of music that looks like a human.

Turing Music Test
How to understand that a piece of music created by a machine is really worthy of our attention? To test the work of artificial intelligence systems, a Turing test was invented. His idea is that a person interacts with a computer program and with another person. We ask questions to the program and the person and try to determine who we are talking to. The test is considered passed by the program if we cannot distinguish the program from a person. For example, the DeepBach algorithm was tested, which generates notes in the Bach style. More than 1.2 thousand were interviewed. People (both experts and ordinary people) who had to distinguish the real Bach from the artificial one. And it turned out that it was very difficult to do this — people can hardly distinguish between music composed by Bach and created by DeepBach.

What about the lyrics?
Well, we’ve sorted out the musical compositions, but what about the lyrics for the songs? Can artificial intelligence compose poetry? Yes, and this task is even easier than writing melodies, although there are also enough difficulties here — the algorithm needs not only to “come up” with a meaningful text, but also to take into account its rhythmic structure.
In 2016, the developers of Yandex released the album “Neural Defense”. It includes 13 songs, the lyrics for which were composed by artificial intelligence. A year later, the album “Neurona” was released with four songs in the style of Nirvana, the verses for which were also generated by neural networks.

Thus, we see that artificial intelligence is able to write even music and lyrics for it, but will it ever replace songs written by a person in which feelings and life situations were invested?

Sources/references:

https://knowhow.pp.ua/ai_music

https://youtu.be/lv9W7qrYhb

https://youtu.be/-yTs58ityvs

https://towardsdatascience.com/generating-music-using-deep-learning-cb5843a9d55e

Tagged , ,