Railworks America

If this AI thingamajig is so great riddle me this. How come we don't have it update the core engine that all the geeks are always complaining about so TSC would be the latest and Greatest Train Simulator Ever and it would allow us to just do whatever we want it to do?

Whether or not it is possible to use AI to improve TS. The issue is with the core game, it is obsolete. It is my opinion that there is absolutely no desire from DTG to update. DTG believes that there is a better future with TSW and more profit from that franchise to be made.

Most of you may know there is a flight simulator mode for Google Earth. The day is coming where all of these platforms will be obsolete. You will tell an AI program you want to operate a train from point A to point B,what year,type of locomotive,time of year,etc,etc. It will generate everything you see based on historical record. If someone had proposed this idea 10 years ago,I would say it was far-fetched. From what I can tell,we may be nearly there. I spent years asking for a few little tweaks to our favorite sim. I believe that you are right,there will be no real updates to this simulation. Last night I purchased TSW and the Sand patch route. I have not given up on TSC,I am just going to check it out. I hate learning a whole new system. It is very unlikely there will be a simple route editor/creation tools that we have with TSC,as they only want to sell route packages.

I hope you are not disappointed. I played around with TSW and other than the graphics didn't at least to me offer much. I have to admit I was a little tongue in cheek with my comment since I consider at this point anyway that AI is just another way to dumb down the human race (like we need any help) so we don't have to learn anything by study. Really I think it's a sad state of affairs and will not end well.
I hope I'm wrong but so far I don't see much in the way of improving life in any way. Code writers seem to be the only people drooling over this latest attempt to solve a problem that doesn't exist.
Just one guys opinion.

Based on several years of conversations about both TSC and TSW, I still think the core engine for TSC is superior. TSW has a lot of limitations directly related to its core. Steam operations are just one example. Other simulations such as Run8 and Trainz have seen improvements, but they are no match for the Kuju engine.

I think AI can be a helpful tool for TSC and TSW. AI is not a magic wand, it is just a far superior method of accessing data. I think AI could be used to add additional features to both engines, but it cannot go beyond the existing limitations. To be honest, the ancient Commodore Amiga had a better graphics engine than most of today's cores and the engine MSTS was going to use for its version 2 was awesome. Let's be honest, the bottom line with all gaming software is $$$. The noble goal of helping us enjoy the hobby more is a dream.

I've tried TSW and am excited about the products currently produced by the guys at HIS. It would be interesting to have a round table discussion with those guys to evaluate the potential of TSW. They seem to have found some new techniques that might make TSW more appealing. Right now, TSC is my sim of choice.

Harry

What makes (most) humans different to animals is the brain, using it to learn and adapt. Take that away and what's left?
Already there's a generation who don't study they Google or Wikipedia with no actual retention so they actually know nothing.
Where's the originality, the curiosity, the desire to improve things and life?

It is not lost on me that Buzz was talking about aiding and improving this platform with AI. Let me say this. I still dabble in Trainz,checking in to see what is new. Prototype route building is way more complex in that "game". That is indicated by the fact that a vast majority of the routes for TSC are real world based,and Trainz are 90 percent fictional. I asked an AI if it can generate a prototype route in Trainz. It said yes,but much of it was still telling you what the steps are,not actually doing it. (So it lied) But that day is coming. And if someone uses AI tools to create a helper,you may be able to tell it what route you want to build,and it will generate it.
I have played around with trial and free mesh generation tools to make some locomotives. The results are mediocre to horrendous. But that will continue to improve. I am very concerned about the rise of AI. But it is inevitable. I hope its powers are used for good.

It can do a lot of great things including cook my son's computer with code he should have never used to get his computer to run faster. The most interesting and impressive thing is it keeps wanting to learn till it gets the right answer or outcome. Right now, AI can build your scenery within Train Simulator if you want but you need to give it a very specific script to do so which can take hours to create and not worth the time in my opinion given the errors that occur which actually need to take place in order for it to learn from. This is no three line script like my venture into AI edited screenshots as many of you have seen, this is way more in depth. You need to tell it what assets to use, what files to use, what not to edit (do not let it edit roads, lofts and tracks) and tell it exactly what worked and what didn't work. The list it actually asks for is kind of daunting especially if you are trying to recreate something in the real world. Right now it only hypothetically knows what it needs to do to build a route on its own and it appears there have not been enough guinea pigs (the guys (And gals if here as well) that route build) yet to provide constructive feedback so it can learn how to actually build. I have personally played a little with it and it does show promise, it is eager to provide a solution even with mistakes. It will keep fine tuning what it provides till it works. You will actually begin to feel bad for it if playing with something like this and you keep telling it failed because it works incredibly hard for a solution and eventually it will get the right solution. This is just not within the amount of time I have to devote to our hobby at this time but was neat to try. I won't be using it for my building any time soon but eventually when it learns enough from the "community" that wants to use it and it learns enough to be useful, that will be a different story for me. I don't think it will be too long either, someone will eventually feed it enough information along with trial and error that it will become useful. As for the Core engine as Buzz suggested, I believe it is only a matter of time before we see something like that from someone. Think of MSTS Bin from years ago if anyone is from the old Trainsim Community two decades ago. One guy came up with his solution for performance issues and new features after development on MSTS ceased which solved everything for the community till Open Rails was offered.

Since scenarios are basically text files, do you think AI could create one? It seems scenarios might be an easier place to start than routes. I remember a MSTS program from Skyline Computing called "Activity Generator." It worked well and created many interesting activities for me.

Harry

harryadkins wrote:Since scenarios are basically text files, do you think AI could create one? It seems scenarios might be an easier place to start than routes. I remember a MSTS program from Skyline Computing called "Activity Generator." It worked well and created many interesting activities for me.

Harry

Yes I believe this can be done but is likely very labour intensive at this stage and you are are probably better off doing it in the sim. You need to write a script and they can be daunting. A script really is not that hard to do but requires a lot information and trial and error as I explained previously. In the case of scenarios, you would need to know all siding, yard names, AI consists as they appear in the sim. You would also need to know what rolling stock you wanted to use along with their locations. Just this a alone could be a several page script to do and as you would quickly discover like my own experimentation a lot of work for something that may not work properly but yet a neat experiment to try. You will spend a lot of time to see how things load into the sim and if they are where they are supposed to belong. What is really needed is for someone to make a program specifically for TSC that interacts with AI. Think of it as TS Tools AI. It knows where all you assets are, you don't need to write a script telling it where they are, you just check a box. You could also check a couple of boxes of where you wanted to start your activity and where to end, what AI consists to add, etc. AI is a tool which in my opinion needs additional custom tools (software) to properly utilize. There will be some smart guy or gal in the near future that comes along and figures out how to utilize AI properly for our sim at some point in the near future with such tools. I am not a programmer but I streamlined my work by using Excel to create lists when I trying this a while back. This period with AI is much like the early days of the web, there is only limited knowledge at this stage and the majority of us will need to be patient while the smart ones build the tools in order to properly use with programs like TSC.