Stefano Casillo moves on to new adventures from Kunos Simulazioni

To add to what others already have answered, the AI are more lively and race each other instead of driving in a train. From what I understand he is trying to make the AI make slight mistakes as well, to make them feel more like human drivers.

 

My 2c.

rather than "traction mistakes" which are quite strange to see and to believe in cars with TC***, i'd like to see way more "braking mistakes" (again, not like blocking tires, not believable in cars with ABS) for example wrong brake points with cars going out the optimal line.
Or, in the best scenario, also have cars having some oversteer or going long or 180°/360° in turn entries (too much speed + wrong brakes applications).

This kind of errors are way more common to see in human races, both virtual and real.

***sadly, this could be right in AC1

 

After making the AI machine learned on all tracks they could become human by adding random delays in their reactions to simulate tiredness, distractions etc.

 

Stefano is trying to improve several areas of AI. So far he has solved the problem that made it difficult for IA controlled cars to overtake both the player and other IA drivers. Now IA is no longer hesitant and fearful as in the public version, but if it is in favorable conditions it overtakes you without hesitation. This has been a great step forward so far!
A dangerous behavior that happens at the start has also been solved. In the last streams he is trying to improve it in the accident situation, i.e. to prevent a car involved in an accident/spin from creating a mess by entering the track in a reckless way, and at the same time to allow the IA that arrives not to get stuck behind a stopped car. As he improves this aspect, Stefano can allow the AI to make driving errors, such as mistaking the braking point, spinning, etc.. In all this he is trying to parameterize some values to allow the player to faithfully choose the level of aggressiveness of the AI. I'm not sure but it should also improve AI in rainy conditions, making it more "human".
If he still has some time left once the main problems are solved, he can explore defensive maneuvers by AI, but this is not the priority for now.

 

It's funny how for example in music production there was a big push to make drums "perfect" as in perfect on the tempo or the beat in 2000s, they did make it and then everyone realized how it's just too perfect. It made music sound like a machine...so now all the new plugins have the "humanize" slider where you can select how much imperfection you want to bring back in to the beats.

In a similar way to what they are trying to do to AI here, music producers are bringing back human errors and inability to play 100% on the tempo...which is what actually makes music organic and lifelike.

 

Stefano Casillo fixes virtual reality please
you are our only hope

 

He hates dealing with VR implementation, if i remember correctly from AC days

 

I doubt they can do much more on the VR front. This engine is so heavy on the performance front that very few if any users can run triples at 144hz, much less VR with the settings turned up in VR. It's the engine itself as they rely heavily on TAA which isn't great for VR.

 

I have a friend that with the rtx 3090 and the hp reverb in epic everything but 60 fps

but that looks great

 

There doesn’t appear to be THAT much difference between a good overclocked 3080 and a 3090, I do wonder what the video memory usage of ACC with a reverb is though.

Any chance you could find out from him? Also what pixel density does he run?

 

ale robles

I can tell you

 

I'm just catching up here, has any of Stefano's new Ai work made its way into ACC yet?

 

Not yet. And we won't see new strims on ACC in next few weeks - Stefano is going to work on sensitive data, so he won't be able to share this work with us.

 

Really really excited to see today's stream about Hydrofoil.
Best news so far, the new graphic engine is awesome.
I'm happy even if it means a delay of some month to the open demo, the final result must always be the primary goal, and judging the very first results on screen without any kind of work and optimization by Stefano, the choice to change the engine is absolutely right.
Never thought to see this small sail project to reach the same level (imho even better) than ACC.
Tbh, since the beginning my biggest personal worry about Hydrofoil was all about graphic side, now we know that this could be a real milestone.
Congrats Stefano, really, good luck for this project.

Rivedere Luna Rossa 2000 regatare contro Black Magic sarà ancora più bello.

 

My shows in vrfps 11.6 g vram. But i dont think it uses that mutch. I am forcing 45 fps all at epic. Ss 150 and all looks great. Can crank it up more only 50-60 % use of gpu. With the reverb g1. I really have tried hard testing 45/90 fps. With the 3090 really dont feel or see any difference

 

Cheers, 4K and 150 ss, not impossible that it is using that much. Which wouldn’t be ideal for the 10GB 3080

 

Hydrofoil Generation is visually stunning now that Stefano changed to Unigine.
Even if it's not my type of game (boats ), considering it is from Stefano and looking as good as it is I will definitely be a costumer.

On another note, Stefano stopped streaming ACC AI development for the time being, but it seems his work is already available for the first tests in ACC, so that is an indication that the improvements are flowing nicely.

 

we must do a poll ---> What is Stefanos secret stuff?

a) He improves AI for AC
b) He implements AI for ACC - Nords
c) He helps Kevin with the thunderstorm simulator

 

Out of curiosity, how do you know, did I miss it?