H7 vs taking things into our own hands

[dredknight]

Hey Q,

Some feedback to return about the 3.1AI mod.
If when you drop a new version of the AI (I mean the new version) is it possible to add the 8th skill option box to be also available for tweaking in the Multiplayer -> Local Area network game lobby.

We found out through tests that 8th skills option works in LAN but it has to be activated by all players before joining the lobby and starting the map.
This also should be done when loading a save game that was played with 8 skills.

This has to be done because the save file itself does not keep track of the number of skills (they are kept in the EXE file, that is been modified through the check box) but the data of the skills themselves.

Cheers

[inxan3]

Hi Q,

Just wondering if there is any update regarding the AI and when you expect it can be ready for release?

many thanks

[Quantomas]

Not that long any more.

It's probably time to explain what work is being done, that you can put the aforegoing statement in context.

My initial idea was to develop an AI to turn HoMM into a Mecca for TBS enthusiasts, simply by offering challenges that no other TBS can offer. In principle HoMM is ideally suited for this, because the gameplay is so varied that it can offer unlimited strategic challenges. What was missing is an AI that can truly handle all the complexity, while being equally fast and competent to match wits with the best human players. You can easily imagine what HoMM would be, had it an AI that actually could outwit you. You know how great the best HoMM maps are.

There were also a few people who saw this as a chance to create content, a grand campaign, that would deliver to the fans what Ubisoft failed to provide with H6 and H7. A lot of work went into this, but ultimately this work was halted.

It's still important to know this though, because the project went through different phases to become what it is now. Naturally, for all people involved the motivation was for the project to become economically viable.

After it was clear that we wouldn't create a HoMM spiritual successor, I probed various alternatives before I settled on what the project is now. The most important milestones are this:

What is intelligence? -- I took two steps back, first from application design, and second from understanding AI as a discipline that is primarily concerned with developing algorithms and implementing these. In truth intelligence is vast, it covers so many different aspects and you find it in so many places, that you have to ask this question in earnest: what is intelligence and how does it work? It's in humans and in life, but if you go deeper it is in many physical systems as well. Ultimately what you need is an understanding of concepts like recognition, differentiation, adaptation and choice.

AI - or more precisely a theory of intelligence -- Once you understand what it is about, you end up with the challenge to develop a model that describes all interactions in a system and how different actors arrive at choices that can be interpreted as their best options. This is fairly abstract.

Technical feasible AI -- This is technically a solution based on what you learned from the aforementioned tasks. A blueprint of a system that works on bits and yields algorithms and processes, binary logic. Yet many of its algorithms are essentially recursive and recursively interconnected. It's more like a system of differential equations, not so much building blocks that you can apply just so. But it's a solution and powerful if you understand it, good enough for HoMM for certain.

Physical recursions -- Completing the aforementioned tasks took many years. You would be forgiven if you, after all these years, would say let's complete the cycle and finally do what you came for in the first place. But, having looked at all these fundamentals and having noted how similar these are to fundamental physical concepts, I had to take a look and try to understand what it means for our understanding of the world. I am glad that I did and it emerged something greater than AI, but it took time.

The AI for HoMM -- The next logical step from the general AI model is to apply it to HoMM. You know dominion over castles, heroes collecting forces and traveling over the map. It's technically application design, not its implementation. It's understanding how the different elements of HoMM interact with each other and figuring out algorithms that work efficiently to arrive at the choices the factions and heroes make in the game. It's big and it's obvious why mostly all AI programmers, from CIv over Endless Legend to At the Gates but also H6 and H7, can do only so much. We have the chance to do better now, building on the research done earlier.

Implementation -- That is where it's at now. Techncially rewriting the AI for H5 by completing the cycle that began with the AI research.

It's not long now and hopefully we can begin writing a new chapter for HoMM once it is complete.

[Panda Tar]

Great news, Quantomas. Hope things keep going well henceforth!

[Pol]

Kinda, kinda, kinda a good news
Are you not afraid that some politican will install your AI, change the game, and the humanity will suffer? Naturally some retardant must be in use. We all know how it ended with Shodan Microsoft AI.

[Quantomas]

Thanks, Panda & Pol.

Actually a lot of work in the application design and implementation is done in order to not expose the know-how.

Everything you do has consequences. The bigger the impact the more you are required to be circumspect.

[inxan3]

Thanks for the update Q. Seems like its been quite some journey. Looking forward to see the result of your work.

[dark-whisperer]

What the hell has happened to Quantomas and this project?

[Quantomas]

Working as always.

[chaoslordd]

Hi Q! i have a rather stupid question, but will the new ai work on lan? (i love playing maps on coop with my friend, and having a the new ai to pose a real challenge would be really interesting).
thanks!

[Quantomas]

It is on my roadmap, but technically a lot of features are, possibly involving years of work.

In order to resolve this, I have a plan to gauge the interest there is for each feature that needs work. I will say more about it, once I am a bit further along on the roadmap and have a bit of time to set up my blog. You can't miss it, if you follow along, ideally at our Discord.

[chaoslordd]

awesome!, can i have a link to the discord? (didnt knew there was one)

[Quantomas]

The link is in the signature (the last line of a post usually), in the hope that bots don't find it there. Who knows?

[chaoslordd]

haha, im blind, i didnt saw it, thanks!.