Connect with us
Build in 60 seconds Thats how long it took GPT 4 Build in 60 seconds Thats how long it took GPT 4

Gaming

Build in 60 seconds: That’s how long it took GPT-4 to generate a Pong game in JavaScript

Published

on

Neural networks continue to amaze with their capabilities. Developer Pietro Scirano reveals that GPT-4 was able to create JavaScript versions of popular games Pong, Asteroids, Breakout And Pac man. At the same time, it took about a minute to generate each.

Build in 60 seconds: That's how long it took GPT-4 to generate a Pong game in JavaScript

Source: Getty Images

The games were created according to a short TOR, and the specifics were not prescribed in the tasks. First, Shirano asked GPT-4 what the neural network thought was the best programming language for “creating a Pong-like game.” In response, the system recommended using JavaScript, and then generated a playable version itself.

In addition, the neural network has added a scoring system to Breakout. As the developer himself stated, the GPT-3.5 version could not create Pong.

It took 90 seconds to create Pac-Man, this game was “ordered” by the user under the nickname Shota. After generation, he only slightly changed the colors in the code and corrected the position of the mouth of the hero of the game. At the same time, the user stated that he did not understand the code.

Earlier it was reported that GPT-4 was able to create a JavaScript game where you have to “rob cows”.

Click to comment

Leave a Reply

Your email address will not be published.

Gaming

Nvidia unveiled a monster GPU with 208 billion transistors, but not for gamers. This is a Blackwell GPU for AI

Published

on

Nvidia unveiled a monster GPU with 208 billion transistors but

As expected the other day, today at the GTC 2024 event Nvidia presented its new GPU – the first of the Blackwell generation.

Nvidia unveiled a monster GPU with 208 billion transistors, but not for gamers.  This is a Blackwell GPU for AI

This GPU will power next-generation AI accelerators. The announcement is quite extensive, although Nvidia has not yet disclosed all the information. In particular, there is not even a clearly stated name. The company mainly talks about the Blackwell GPU, and on the website you can find different options, but they apparently do not refer to the graphics core, but to ready-made solutions, which Nvidia sometimes also calls GPU, although this is a full-fledged accelerator.

Nvidia unveiled a monster GPU with 208 billion transistors, but not for gamers.  This is a Blackwell GPU for AI

Let's start with the fact that the Blackwell GPU is a giant graphics processor. It contains 208 billion transistors. For comparison, the GH100 has 80 billion transistors. The second feature is that the Blackwell GPU is Nvidia's first multi-chip GPU. It consists of two identical crystals, and even surrounded by memory chips. The crystals are connected to each other by an NV-High Bandwidth (NV-HBI) bus with a throughput of 10 TB/s. The number of CUDA cores or any other computing units has not yet been disclosed, but given the giant leap in the number of transistors, it is logical to expect a huge increase in the number of computing units.

Around the GPU are eight stacks of HBM3e memory with a total capacity of 192 GB with an 8192-bit bus and a bandwidth of an incredible 8 TB/s.

All this is produced using a 4 nm process technology, not 3 nm as expected, but this is an updated TSMC 4NP process technology.

Nvidia unveiled a monster GPU with 208 billion transistors, but not for gamers.  This is a Blackwell GPU for AI

Nvidia itself, comparing the Blackwell GPU with the GH100, says that the new solution is 2.5-5 times faster than the old one, depending on the mode. Interestingly, performance in FP32 mode is not specified at all, although due to the emphasis on AI calculations this may not be so important.

In fact, the parameters of the new GPU, among other things, clearly show where Nvidia is now putting all its efforts. Looking at the history of the company's gaming GPUs, such a giant leap in a single generation seems to have never happened. Given the hype around AI and Nvidia's position in this market, there is nothing surprising in such a bet – this market makes a lot of money for the company.

Nvidia also immediately talked about the GB200 Grace Blackwell Superchip accelerator, which will contain two new generation GPUs and a processor with 72 Arm Neoverse V2 cores. The performance of such a monster in FP4 mode will be an incredible 40 PFLOPS.

Continue Reading

Gaming

LinkedIn will add games to boost engagement among billions of users

Published

on

LinkedIn will add games to boost engagement among billions of

LinkedIn is adding an unexpected feature to its platform: games. The social network that helps you find and make business contacts is looking to add puzzles to its platform as a method to increase user engagement.

Inspired by the viral success of Wordle and other puzzle games, LinkedIn is developing its own puzzle games called Queens, Inference and Crossclimb. According to a LinkedIn spokesperson, these casual games will help “fun, deepen relationships, and spark conversations” among users.

LinkedIn will add games to boost engagement among billions of users

Picture of DALL-E

The idea involves integrating these puzzles into the LinkedIn interface. Non-gaming platforms are increasingly recognizing the ability of casual games to increase user engagement.

LinkedIn has more than 1 billion registered users (as of November 2023), representing 150 business sectors in 200 countries.

There is no official launch date yet.

Continue Reading

Gaming

Nvidia DLSS beats AMD FSR in modern Full HD games. New comparison shows a noticeable difference

Published

on

Nvidia DLSS beats AMD FSR in modern Full HD games

The authors of the Hardware Unboxed channel compared Nvidia DLSS and AMD FSR scaling technologies in modern games and Full HD resolution, where upscalers perform worse than in higher resolutions.

Nvidia DLSS beats AMD FSR in modern Full HD games.  New comparison shows a noticeable difference

Technologies were compared in games such as Cyberpunk 2077, Ratchet & Clank Rift Apart, Dead Space, Star Wars Jedi Survivor, Starfield and several others. But all the games were quite fresh.

Nvidia DLSS beats AMD FSR in modern Full HD games.  New comparison shows a noticeable difference

The comparison showed that DLSS is still better than FSR. Not always and not everywhere. For example, in Cyberpunk 2077, the asphalt texture is much better rendered when FSR is activated, but then FSR shows a much more obvious ghosting effect. In general, when FSR is activated, various image artifacts appear more often and more strongly, including flickering, ghosting, ghosting, and so on. The authors of the video say that FSR lags significantly behind DLSS and native rendering in terms of image quality in most games tested. Let us repeat, the conclusion is valid specifically for Full HD resolution.

To be fair, it is worth saying that sometimes the very artifacts that the authors of the video talk about are not easy to detect on the video, even if the video is zoomed in and slowed down, and even after the author explains what exactly to look for. And during a dynamic game, without looking at the picture in this mode, many players do not notice any problems – many such comments can be found on various platforms in the relevant topics.

Continue Reading

Most Popular