Ian Bell's Straight4 Studios Launches Website, Showcases Key Staff

straight4.jpg
It seems like Straight4 Studios is going all in. Now they have released a website with a blog. Furthermore, they promised a monthly newsletter! What's more interesting, however, is the information that has already been published on the blog.

The first blog post on Straight4's website features two different interviews with key members of the development team for the upcoming title GTR Revival.

Before we dive into the post, the key takeaway is that it seems like the original team is really back together. Henrik Roos, one of the key people of SimBin back in the day, is the first person featured in this post.

The Key Personnel Introduced​

As mentioned before, Henrik Roos is introduced as the Straight4 handling consultant.

If you don't know Henrik Roos, he has been a racing driver, especially known for early 2000s FIA GT racing.

His role at Straight4 includes translating how the cars feel IRL to the simulation of GTR Revival.

Furthermore, Straight4's Physics Director AJ Weber gets introduced. He "has gone on to work on many bestselling AAA-sim racers over the years". According to mobygames.com, he has been working on the first two Project Cars games in the physics department as well as Fast & Furious: Crossroads.

Physics Philosophies​

What follows is an intricate introduction into the philosophy behind the direction the physics department is supposed to go.

The post goes into detail on how realism is supposed to be portrayed in GTR Revival. Furthermore, tyre models and the fact that the physics engine is built from the ground up get featured.

You can read the full blog post here on Straight4's official website.

What do you think of this approach? Is Straight4 getting you interested in the ongoing development of GTR Revival? Let us know in the comments down below!
About author
Julian Strasser
Motorsports and Maker-stuff enthusiast. Part time jack-of-all-trades. Owner of tracc.eu, a sim racing-related service provider and its racing community.

Comments

Premium
Having played around with OpenAI Playground and ChatGPT extensively (and documented the weird and wacky yet admittedly impressive results) I can say what I'm hearing from those videos is, yes, prompted by 'events' (eg. car 3 hit wall at t3, car 32 passed car 5 at time xxx, etc.) but the AI isn't necessarily being told what to say.

AI like ChatGPT can make up stuff based on the context of racing, race craft, the personalities of who was involved, the type of car, how bad the impact was, etc. Some things are implicit and don't need to be told (racing, race craft, braking, speed, impact -- they're part of the AI's knowledgebase), while the others likely would (name of drivers, car number, the event that actually happened, etc.).

It wouldn't need to sound pre-scripted at all. And the actual exact response need never repeat, ever, although the AI's idioms and phrase formation habits may repeat.

The impressive thing about OpenAI/ChatGPT is that it's doing things dynamically that in the past we've only tried to do prescriptively.

I have a whole tweet thread just playing around with what OpenAI Playground (OpenAI owns ChatGPT) and ChatGPT can do.

Twitter Thread of OpenAI/ChatGPT Interactions

So, I'm not surprised at all. ChatGPT AI tech is not perfect yet but the leap ahead it made in human-like AI is surprising and significant.

1676396390291.png


1676396198478.png

1676396464676.png
 
Last edited:
Premium
Specific to racing broadcast color commentary -->

Note how it makes up a few likely but unstated details like Radillon being one of the fastest and most famous turns, medical teams attending the driver, prayers from fans, safety improvements to tracks, and a race interruption.

It's all things it has been trained on gargantuan amounts of data to realize are related to racing crashes.

Impressively, it gets it realistically in order.

1676403595569.png

1676403614616.png
 
Last edited:
Premium
Sure, to German and Italian

Give it a try yourself it's free


Perhaps even more impressively I think when I'm asking it to translate it's not translating from the original it's translating from the translation. Usually things get lost in translation when you do it that way but it seems to work out here judging by Google Translate of the results.

1676404729683.png


1676404778965.png
 
Last edited:
Premium
The elephant in the room is that this is all happening on massive scale in the cloud

You could remote API into ChatGPT servers to accomplish this in a video game

But I don't yet see how such wide-ranging abilities could be shrunk down to a local solution that wouldn't require internet access

I remember when Project Cars 2 lost their real-time weather API access and it was never brought back. Major feature gone just like that.
 
Premium
But what if the database of informations that the AI should use to create the phrases and dialogues is given in game (I mean, as a file to analyze)? No need to go online if the APi is working in game.

I suspect even with gigabytes of trained data the variety of speech and knowledge base to draw peripheral information from would be very apparently limited to human listeners

Maybe they could go with a dual approach, like Microsoft Flight Simulator did with satelite data, and you could have an offline mode that would be somewhat less impressive but still fit the bill.

Still, who knows how cpu heavy the AI algorithm itself would be to process those gigabytes of data. Maybe it needs distributed cpu power in the cloud to do it.
 
Last edited:
I have a good example how
Perhaps even more impressively I think when I'm asking it to translate it's not translating from the original it's translating from the translation. Usually things get lost in translation when you do it that way but it seems to work out here judging by Google Translate of the results.
Hehe I have a good example of what can be lost in translation by newest "advanced" AI translation - when you translate from english to danish and back to english:

Eng: Oh god there is nothing left.

Dan: Oh gud der er ingenting til venstre.

Eng: Oh god there is nothing to the left.
:roflmao:
 
Premium
I have a good example how

Hehe I have a good example of what can be lost in translation by newest "advanced" AI translation - when you translate from english to danish and back to english:

Eng: Oh god there is nothing left.

Dan: Oh gud der er ingenting til venstre.

Eng: Oh god there is nothing to the left.
:roflmao:

ChatGPT: 1

Others: 0

:)

1676413883821.png
 
We aren't in the future yet. In this case there are lines of codes that trigger a reaction when something happens. Just like in Pacman.
You give the AI a database of phrases and when it's triggered, it reads these with a synthetized voice generated on the fly, choosing the line to read according to the parameters predefined by the developer.
There is the possibility of the adaptive learning by the AI, but it happen always according to predefined rules, that set the patterns that the AI has to follow doing its things.
So when the AI listens to the phrase "I'm oversteering in turn 1", it picks the possible answer to the problem of oversteer within a bunch of phrases, choosing randomly, so the next time that will be triggered by the word "oversteer" it will pick a different one and won't sound scripted. It could be working this way.
How it recognizes the word that triggers it? In the same way that a program like Dragon Naturally Speaking recognizes your words and train itself to identify your tone of voice and pronunciation. It's nothing new: the good part is that it sounds pretty much human and that's of course is cool.

So, the times where a Tesla can start itself go for a ride by night, because it feels melancholic are still far. A monkey could do that, but not the AI. :)
I asked for them to clarify what the video is showing, not for a condescending reply about how it could be done.
 
Premium
Looks like a few of the posters here think that their 50/60 quid entitles them to pick the Straight 4 Games studio's staff... talk about entitled
If you don't like it don't buy it.
But seriously guys, with all of your knowledge and presumably skills I wonder why you don't simply develop the game you want and choose your very own staff... that'll teach Mr Ian 'smartypants' Bell for not doing what you want.
 

Latest News

Article information

Author
Julian Strasser
Article read time
2 min read
Views
21,692
Comments
140
Last update

What is the reason for your passion for sim racing?

  • Watching real motorsport

    Votes: 174 66.9%
  • Physics and mechanics

    Votes: 114 43.8%
  • Competition and adrenaline

    Votes: 121 46.5%
  • Practice for real racing

    Votes: 45 17.3%
  • Community and simracers

    Votes: 71 27.3%
Back
Top