AI generated voice of Lady Gaga covers “video games” by Lana Del Rey
In recent months AI became dominant in many markets. Everybody use chat GPT for writing essays, song lyrics or even a screenplays. AI generated images flooded the internet, many artists use them to create video clips for they song. In call centres many human assistances got replaced by chatbots. It found it’s place even on legal market where first AI robots stats helping lawyers it their jobs in court. In recent days videos of AI covers started to appear on YouTube.
AI model created by gagaswings on YouTube
Using machine learning and huge assets of voice data of a person we can now create fully artificial asset of their speech. This technology is in an early stage for now however it is not disappointing. Cover of “Video games” was created in a program called diff-svc. For now it has closed access, just like DALL-E in it’s beta but will probably open to the public very soon. For AI to understand tones and assign it to given voice it need’s many types of audio such as singing acapella and speaking in a different ranges so it could generate the voice of a person as resembling as possible. Next step after gathering all the data is running a test consisting of at least 50 thousands steps so the AI voice could be realistic. After that process we input an acapella version of the song we want to cover so it could learn how to intonate the words with generated voice. Based on those two processes AI uses generated voice to sing chosen song.
The idea of artificially generated voice is quite new and for now it’s being used in non-harmful way, however this poses a lot of ethical and legal threats. We can do anything with the generated voice, and it don’t have to be the right, proper things. When it comes to famous people such as Lady Gaga there is also a problem of copyrights and intellectual property of their voices. With all those processes powered by AI we slowly enter an era where technology will be undistinguishable from reality.
Earlier this year Epic Games released new computer graphic engine that revolutionized industry. Two weeks ago new update appeared and now scenes created with Unreal Engine 5.1 are indistinguishable compared to real life footage.
When it comes to visual effects in movies, with the right budget and time, filmmakers can make everything look realistic. Movie shots are scripted, the camera moves in a certain direction and nothing will change that. The computer-generated images (CGI) that appear on the screen, which we see when watching a movie, are pre-rendered. In games where the player is responsible for moving the camera, the frames on the screen must be rendered in real time. In order for the human eye to see smoothly we need at least sixty of them for every second of the game. Generating this amount of data in a short period of time is very time consuming, and yet many players can’t afford to admire the beauty of the shots because they have to keep the graphics to a minimum to make the game work on their computers. This is where Unreal Engine 5 comes in.
One of the biggest problems in CGI is lightning. It is very easy to identify bad light in a scene. However, it is very easy to visualize thanks to global illumination. This is the way light bounces off an object and lights up another. This is a difficult process to do, especially in real time. So far, graphic designers have used light baking which pre-renderes lights map. This method works very well, but if there is even a slight change in the positioning of objects, the whole process has to be again done from the scratch. This is very time-consuming. The lighting system in Unreal Engine 5 called Lumen solves this problem. It renders light in real time regarding moving objects. It focuses only on the multi-reflection global illumination of the main light source and. That’s why it looks so realistic. This process improves the workflow and speed. Are very valuable considering short project deadlines.
Lighting was difficult for the developers, but from the player’s perspective, the most challenging part is rendering the map elements. This is all due to particles called polygons. The more detailed an element is, the more of them it assembles. A single CGI location can contain thousands or even millions of them. These are difficult to render on the average home computer, which can cause the game to crash or freeze frames, and the game will not be playable. A simple way to solve this problem is to lower the level of detail – the number of polygons for a scene, but then the game looks very flat and is not pleasing to the eye. A new Unreal Engine 5 option called Nanite dynamically deforms the environment by lowering the total number of polygons on an object. It changes the number of them depending on how far away the object is, the closer the item is to the camera, the more polys it consists of.
These two elements are revolutionary for the CGI industry. The new 5.1 update patches some of the problems that the engine has had so far. Worth mentioning are fixes for global illumination. Transparent objects resonate light, which didn’t work very well in the previous version. Now the reflection of glass is more realistic, as is water. It is no longer milky white, but actually shines through and reflects objects around it depending on the camera position, rather than being just a blurry patch of light.
Currently Epic Games is working on improving Nanite technology. As for now it can only be used for static objects but they are trying to implement it on a moving characters too.
The last thing is MetaHumans. It is a creator of realistic humans, which is similar to creating characters in The Sims series. It is an amazing tool for creating NPCs in games, but also figures in an animated movie. Another phenomenon is the animations and movements of the created character. These are no longer hours of making, but several commands. The computer itself calculates how a process should look like and performs it itself. This reduces the human work almost to zero. As a result, the time needed for a given project is much shorter – creating a movie will no longer take years as it used to.
In 2017, Rouge One: A Star Wars Story was the first film to use an on-set game engine – Unreal Engine 4, followed by Disney’s The Mandalorian series and HBO’s Westworld. Now, with the latest version 5.1, we can expect even more collaboration in the movie field in the future. As for the game industry, many developers have announced that they will be working on the Unreal Engine rather than its competitors. We can expect games such as the new Tomb Rider, The Witcher and Redfall. So far, we can experience UE5 in action by playing The Matrix Awakens and the latest season of Fortnite, or by watching a number of short films created by various small developers. One of them is The Eye: Calanthek created by ASC, which I highly recommend watching:
Semiconductors – chips are an integral part of every electronic device. That’s why it’s so important for companies to ensure that their product is equipped with the latest version.
For this reason, many organizations are outdoing themselves in creating better and better chips. For many years, the largest semiconductor manufacturer has been the Taiwanese company TSMC. It supplies more than half of the world chip market. It it’s responsible for producing most of the chips for iPhones and Macs that Apple is so proud of. Few people have heard of it, but it is one of the largest companies in the world. In the field of pocket devices, supercomputers and automobiles, TSMC is a nearly a monopoly with more than 90% of the global chip supply. Another 8% of companies are supplied by Samsung chips. That leaves just 2% for dozens of other producers.
But that doesn’t stop other organizations from becoming competitors in the production of new and better semiconductors. The American company NVIDIA recently took away TSMC’s long-occupied first place on the list of most profitable companies. The competition is growing and more rivals are joining the race for the first 2 nm process. One of the biggest contenders is Samsung. As of today, only Samsung and TSMC produce and sell 3 nm chips. Both companies expect to begin mass production of 2nm chips by 2025.
Now Rapidus is joining the game – the newly formed company is funded by a number of organizations such as Toyota, Sony and Kioxia. However, the biggest donor is the Japanese government itself since company’s headquarters will be in Japan. It will allocate about $500 million to help the new company develop. After a stagnant engineering market and a shortage of skilled workers, Japan is looking for prospects in semiconductor manufacturing. The country is also hoping to forge closer ties with the US, as Rapidus will work with IBM to develop next-generation, sub-2-nanometer chips. This move puts Japan in the team of United States in their technological war with China. Results of this colaboration will be visible in 2027, two years after estimated mass production of TSMC’s and Samsung’s samiconductors. In technological field it’s a lifetime. Probably Rapidus will not be a succes at first, but it may have a potential in future projects and development.
There is still a long way to go before mass production of 2-nm semiconductors. However, a bright future full of potential lies ahead. Battery life of smartphones equipped with such a chip will quadruple. Laptop functions will speed up dramatically. The same goes for object detection in autonomous vehicles, which will reduce possibility of accidents on the roads. On top of that, the carbon footprint will be much smaller. It is believed that 2nm could be useful in areas such as 5G internet, big data centres and even quantum computing. One thing is certain, whoever develops a good-quality 2-nanometer semiconductor first will gain great strength in the technology field.
Netflix is one of the oldest streaming platform on the market. It attracted its customers with unlimited content whenever and wherever without the adds like regular television have. High quality content and wide range of titles available made company unbeatably the most known for a very long time. Corporation announced that by November they will add fourth, cheaper subscription option containing advertisements. Lack of those was the reason why people signed up in the first place. How does the future looks like for Netflix?
Photo by David Balev on Unsplash
Even though Disney+ made its entrance to the market in 2019 other streaming platforms maintained their clients as thousands of people started their journey with broadcasting services during lockdown. Now when everything is coming back to normal they don’t have time to sit on a couch and watch movies all day, so they cancel many subscriptions and hold on to only few, the most entertaining ones. Statistic shows that Netflix is not one of them as they’re loosing more and more customers. It reached its peak in 2021 but in this year number of clients decreases. Why?
Mostly because Netflix’s content is simply no longer good. For past few years platform is mostly based on its original content, which, in many cases is very predictable and similar to each other. Even if something is more interesting it does not get renewed after one or two seasons. They don’t provide as much new movies, and the ones that already were on platform disappear in order to appear on other broadcasting services. Not to mention the fact, that Netflix is the most expensive streaming platform among its competition. The same old titles, disappointing new ones and high price makes obvious choice of which subscription to cancel.
Yet company does not give up. Reed Hastings revealed, that new created payment plan will cost only $6.99 but it will contain around five minutes of adds per one hour of watching. In that case scenario Netflix will come close to regular TV program. That solution will not attract a lot of new customers. More likely people that already pay monthly subscription will just change plan for cheaper. Part of them might’ve wanted to cancel payment completely, but some might have not even planned this as just an opportunity to save money appeared. However it is expected for Netflix to earn about $3 billion in 2026 just for advertising. That does not look like drop of incomes at all.
Even though forecasting for Netflix looks bright many people are bored of its content. So if in the future company will base on money from adds and not the customers public opinion of the platform might get lower and lower. It will lead to more people cancelling subscriptions. Unless Are we going to witness yet another great company’s fall? Because who will pay for advertising their product, if nobody will watch it.
DALL-E – machine learning model was created at the beginning of 2021 by artificial intelligence research laboratory OpenAI. Its purpose is to generates artificial images based on given text.
What started as small project quickly gain lot of interest that website couldn’t deal with. Authors made limit of one thousand users that had permission to use DALL-E for a week and a waitlist for other interested. It ended up with thousands of people waiting up to two months to get only few days to experiment with AI. In April 2022 DALL-E 2 got introduced yet still it had restricted access and waitlist. Finally on 28 September 2022 got opened to the public. Now 1.5 million people use it every day.
How does it work? You can create any image with just a text or import a photo to base on. The more descriptive words will you use the more accurate graphics will be. It’s important to specify what kind of image you have in mind: real photography, 3D design or maybe a pixel art. The only thing that limits the user is their imagination.
Image created with DALL-E and phrase “pixel art of a rabbit running through underwater forest”
When it comes to uploaded photo it shortens time spend on postproduction. Professional photographers don’t have to waste hours on cutting out unwanted guest on client’s wedding or overflowed trash bin in the background. Not only AI generated images can reduce time spend in front of a computer, it can also save time while traveling. It’s hard to get a perfect photo in a place full of tourists. Now you can get rid of all those people using DALL-E 2 and enjoy your holiday without getting angry at someone who stepped into your photo frame.
My holiday photo edited using DALL-E
Such powerful tool opens new possibilities but also a threats. Not only someone can create a completely new identity on social media and fool all of their followers or make a funny image of their friend. With help of AI generated images leaders can create perfect propaganda, especially in areas with restricted access to information. They could present their opponents as they please so the people would be in their favour. That’s quite abstract thinking but very much possible. However such technology would be mostly used to create pornography of any kind.
Creators of DALL-E foresaw that kind of situation and banned some words so people wouldn’t create inappropriate photographs, yet they are not the only one in possession of such technology e.g. StabillityAI – their rival on the market. Even though these companies prepared their models not to create violent or sexual content it will definitely happened in the future weather we like it or not. That is why we must prepare to eliminate spreading that kind of images online by group of moderators on website or hopefully in the near future AI systems.
But there is other, significant result of inventing AI generated images. Whole art industry will change drastically. From now on creating graphic is not a complexed process anymore. With AI pictures can be created within seconds. Every single day DALL-E 2 users make two million illustrations. OpenAI allows its customers to monetize work created with their AI model. It will result in huge NFT increase on the market. A lot of so called artists will steal generated images and sell them as created fully by themselves. It will be difficult to identify who puts hours of work into their projects and who only seek oppurtunity for money. However it opens a new path for artists. Just like film became 10th muse, AI illustration is a new way to express ourselves.
DALL-E version of “impression, painting of a flower fields in france”
