I like it 
Itâs not about he bot name itâs about what it is. Any form of automated game play will be frowned upon because we want to interact with player. We hate bots for what they are, not because they are called bots. You can call your bots whatever the hell you want, if it looks like a bot, smalls like a bot and quack like a bot, itâs a bot.
i didnât read through all comments. itâs way cooler now, btw. 
your idea reminded me of something my brain once came up with, and i wonder if you mean the same thing.
Basically, have software record actions and then have that software play an advanced npc?
1 Like
IâLL try to explain OPs idea a little better as itâs not about the bot either.
What we are talking about is AI. Artificial Intelligence. The key word being intelligence.
Intelligence through machine learning.
What science already has proven, with Googles AlphaGo program, is that you via âdeep learningâ can learn a program how to play a game in a humanoid way, through experience, intuition and the understanding of possible outcomes of its own actions.
Sure, you can easily make a bot that can mine asteroids or do a mission, but that will only be a task specific algorithms.
The true question is, can you actually make a bot that can play Eve Online via an learning data representations algorithm, that is without programming it to do so?
1 Like
I would like to thank you from the depths of my heart for this explanation of what I meant⌠so I will, thank you, from the depths of my heart 
Far better worded than anything I was likely to come up with at the moment.
1 Like
Very much yes but it wouldnât be about making NPCâs, the point would be the experiment.
I was only envisioning a handful of them in game & the only reason to hand out a few licences to players would be to broaden the team working on it as it were so different visions & approaches might be applied to the task simultaneously.
So effectively getting any interested geeks in-game to do most of the work rather than CCPâs own programmers, who should be concentrating on making the game better instead of this 
Edit:
If you havenât already have a glance at what bbb2020 has said, as heâs got what Iâm talking about & has given a somewhat better explanation than I did.
i think this can be done by CCP already. They have tons of behaviours saved from players. they could feed an AI with the millions of ways players have reacted in given situations and then just let it loose. An AI only âtrainedâ by a single player wouldnât receive enough input. it would be as limited as the player is.
AlphaGo was fed with a metric duckton of data, trained by high high-grade players, and then let loose on itself so it learns whatever else there is to learn. a single player, or even a handful concentrating on one thing only, wouldnât provide enough for an AI to learn.
/me sceptically looks at his coffee and the temperature.
1 Like
So the core idea may need some tweaking perhaps, I wonât deny it may be a bad one as it stands, just not for being a âbotâ in the fashion you originally assumed 
which of course was my bad for the choice of word 
1 Like
With the current state of machine learning, we are still a long way off from a neural network or similar, being able to play EVE in a human-like manner.
DeepMind AI is a phenomenal piece of engineering and has even taught itself to walk now:
However, there is a long way between learning how to play Go (which has only a very simple set of rules, although great depth in strategy and positions) and learning to play EVE (which has not only complex rules, but also great depth in strategy and gameplay).
In the near term, you wouldnât get an AI playing EVE any better than DeepMind currently walks.
It could be improved if CCP created an API to provide data to consume, but for the most part, an AI taught to play EVE, would be like DeepMind, requiring a way to see the game.
There is a great series (currently still being added to) on YouTube involving a neural network being taught to play GTA V; and thatâs not on a server involving other players:
If instead, it was a chatbot, well I for one would welcome a Tay equivalent in New Eden. Good taste might not.
2 Likes
I donât care if it learn or not, I want players and not line of codes to play the game.
Thatâs not much of an argument though. Except for dedicated PvP players, the vast majority of EVE players interact with lines of code a LOT more than other players. Mining, ratting, running missions, exploration all are activities where one interacts with server controlled entities. It is absolutely nothing new.
Better PvE is something many an EVE player has been looking for. In fact, it is something that CCP is working on as well (see NPC mining ops and the new shipyard, one may even include the sleeper AI in there as well).
What is suggested here is that instead of letting CCP do all the AI design for these new NPCs, a limited form could be made available to the player base as well to open up the experimentation. This would enable so many more possibilities for the AI to try because players will have different goals and skill levels when creating these AIs. It can become PvP by AI proxy.
An easy limitation for the AIs would be that since they are AI, CONCORD does not retaliate for them, or does so MUCH slower than for capsuleers since it is a much lower priority for them. Sure, you could work on making your own AI-drone miner, but then it could get ganked by pretty much anyone, so we might actually see roams in belts to kill AI-drones (especially by people like you who hate the idea) and people might want to defend their AI-drones too. CODE might instead invest heavily in developing a miner ganking AI-drone to carry their fight (pack behaviour, a defined set of acceptable targets and an easy search pattern: check asteroid belts for single targets). Other people could decide that theyâre fed up with CODE and develop CODE hunting AI-drones to interfere with them (actually a similar AI as the miner-ganker, except it looks for catalysts from specific corps/alliances).
It could easily be limited in power in such a way that it is always more worth it to play your own character than to âworkâ on your AI (by giving them an even more limited set of skills than alphas). Many people chose their EVE-activity not strictly for isk/hr but also for personal interest. Developing an AI for EVE would make a really interesting pet project for many in this nerdy community and could actually be useful for the game by doing research for NPCs. Since we already have drones in the setting, it is also incredibly easy to integrate story-wise.
1 Like
All of this is already done by players so I see absolutely no need for CCP to develop a base AI + added code modules for it to learn and then balance everything in the game under the assumption that everyone will use one of those at all time.
The assumption that âeveryone will use one of those at all timeâ is rather irrelevant. Even flying spaceships isnât a feature that everyone uses at all time (there are lots of ship spinners and market/industry players that stay docked the majority of the time), so full time use by everyone certainly isnât a requirement for new features. What % of the player base will fight a shipyard? Yet it was implemented.
You are right though, of course players already do all that. But thatâs not really important either.
The questions are:
- Would the game benefit from more possible activities and more happening in game?
Iâd say more content is better. By launching a bunch of AI-drones, space will be less empty, more might happen. - Would NPC AI benefit from it?
Presumably, the more experimentation the better - Would it excite some players to work on AI?
Most likely!
And the ultimate question:
- Should CCP sink the time to wrap its own AI experimentation tools into a package usable by players?
It very much depends on how much effort that is, which only they know, and how much everyone involved values the idea, which is why itâs good that it was suggested and we can discuss it. Chances are, however, that they have other priorities. So you donât have to worry, and hopefully, maybe, this discussion can spark some kind of idea within CCP for something cool, new and different.
1 Like
Iâm thinking of it in terms of the Microsoft twitter bot that ended up becoming hateful and racist after âlearningâ from a bunch of twitter trolls. lol What would happen to an AI trained in the basics of EvE and then sent off to fend for itself in the game?
Would it survive, would it thrive, or would it kill itself by corrupting the server hard drives?
We all know AI is being advanced every day while we sit here, I think this would be one good way to see what weâre actually asking for before we find out we donât want it after all.
2 Likes
So was I 
The big difference that might modify that outcome a bit is the twitter bot was just a standard Chinese room with parrot features chatbot.
So nothing more than a mirror or echo chamber of those it interacts with which isnât even trying to be AI, itâs about as much like real AI as those old 1800âs automatons where like real robots.
While what I was thinking of was a results & reward driven neural net.
Which isnât a lot better but itâs a few steps further in the right direction.
The fun for me with something like that would be trying to devise different good / bad settings to try & develop a white hat or a black hat, a risk averse station spinner or a kamikaze combat junkie etc.
No. Tay was not a standard chat bot (eg. nothing at all like a suped up version of Discobot here).
Tay was based off a Microsoft AI bot developed in China and was neural network based. Hilariously though, the learning wasnât reinforcement learning (ie. it wasnât trained on existing data as a way to learn appropriate responses), but was specifically designed to apply user input as learning data.
Microsoft have now replaced Tay with Zo:
1 Like
Damn! really? well color me surprised.
I just assumed based on the outcome, I should have researched it 
Well that was dumb of them 
Well you do want ongoing learning & development (if I understand you correctly?) but presumably they didnât include a selection of words & sentences in response to what it said that might indicate an unfavorable response as part of the cumulative reward set?
Well, in all fairness to them. The original AI bot they created in China had more than 10 billion interactions without incident. Not unreasonable of the engineers to think that was a pretty good result, but I guess the way we interact online in the rest of the World is substantially different to the way people interact online in China.
Perhaps.
But anyone whoâs spent five minutes anywhere like Reddit or Yahoo Answers would have known that was going to need 24/7 monitoring for the first few months.
Plus weekly if not daily adjustments to the reward set until they got it right as the trolls found new ways to screw with it they hadnât thought of.
And I find it really hard to believe anyone working for companies like Google or Microsoft have been so sheltered from the internet that they didnât know all that, so I still think it was pretty dumb of them.
Besides, itâs more amusing that way too
MicroSoft is currently the 3rd largest company in the World by market cap, with over 120,000 employees. Of course they hire some very sheltered people. They also hire a â â â â tonne of very smart people too. Hard to think of too many other companies that have impacted modern culture more.
So irrespective of what you think of Microsoft, or Google (or Alphabet in a broader sense), or Apple, or any of the others among the worlds largest companies, none of us would have the quality of life we have without them.
1 Like