Apparently if Google AI indicates it - it must be true. Search Google for “eve online marauder command burst”.
I can’t imagine my suggestion back from Jan '25 was actually implemented…
Can’t wait for this Google AI assist…
Apparently if Google AI indicates it - it must be true. Search Google for “eve online marauder command burst”.
I can’t imagine my suggestion back from Jan '25 was actually implemented…
Can’t wait for this Google AI assist…
AI hallucinates a lot about EVE, I would not recommend using it as a reliable source of information.
Really?
Such as?
I don’t recall all examples but any time I search for more detailed EVE info such as ‘how much hitpoints do warp disrupt probes have’ I get answers telling me that I cannot destroy those bubbles even though I know that smartbombs can kill those probes even if regular weapons cannot lock or attack them.
AI is consistently giving bad answers in my experience when it comes to my EVE questions.
So I scroll past the bad AI answer google gives me and search through the forum and reddit links which usually do contain some useful info.
Thanks, maybe it’s due to how the question is asked?
I just asked it this:
In Eve Online, how many hitpoints do warp disrupt probes have when attacked with small smartbombs?
The very first thing it said was this:
Warp Disrupt Probes (also known as interdictor bubbles) in EVE Online have 1,000 structure hitpoints (HP)
It then went on to explain how many cycles a small smartbomb would need to do and at what range, etc… Then also added small smartbombs are not good to use…So I refined the question to include medium and large smartbombs and this is the answer I got:
Regardless of the smartbomb size used, a Warp Disrupt Probe has 1,000 hitpoints. Because these probes have 0% resistances, they take full damage from every pulse.
The number of cycles needed to destroy a bubble depends on the specific smartbomb’s damage output:
- Large Smartbombs (Standard/T2): These typically deal 300 damage per pulse.
- Cycles needed: 4 pulses (1,200 total damage).
- Large Smartbombs (Faction): Higher-end modules (like True Sansha or Shadow Serpentis) can deal up to 375 damage.
- Cycles needed: 3 pulses (1,125 total damage).
- Medium Smartbombs: These deal significantly less damage per pulse (around 125 for T2 variants).
- Cycles needed: Approximately 8 pulses.
- Small Smartbombs: As previously noted, these have very low damage.
- Cycles needed: Approximately 15 pulses.
Comparison Table
Smartbomb Size Typical Damage Hits to Destroy Bubble Range (Max) Large (Faction) 375 3 Up to 7,500m Large (T1/T2) 300 4 ~5,000m - 6,000m Medium ~125 8 ~4,000m Small ~65 15 ~3,000m Strategic Note: Most players use Large EMP Smartbombs for bubble clearing because they offer the best combination of high damage and the longest possible range (up to 7.5km for faction versions), which is critical for reaching the edge of the 20km bubble from a ship.
AI answers are not consistent, they’re adapting.
The answer that was wrong a few months ago may not be wrong now, or could be hallucinating something else today.
So I’m not sure what you’re trying to say with that post.
All I’m saying is that AI hallucinates often and is not to be taken as a reliable source of EVE info.
And here’s what I’m saying…
If I can ask a question in various ways and it hallucinates wrong answers part of the time, I don’t see the few times it gives a correct answer as enough of a reason to trust AI not to hallucinate.
Try to ask about timers and dps cap over structures. 100% fake.
The problem with LLMs and their hallucinations is you can not know when they hallucinate if you have no foreknowledge of the subject to tell the difference.
So you either
The only thing you can reliably use it for is a first stage of research and maybe some level of fact checking to get knowledge of things you are not even aware of and can know what to manually research to answer your questions you didn’t know before you even needed to ask.
But treating AI answers as absolute or complete truths is a delusion and severe mistake.
![]()
The trick is to use a more sophisticated model, and asking it to look up information with sources. You don’t have to actually read the source, the point is making the AI find the forum post or everef page to make sure it actually exists. If it cannot, it will generally say so.
I’ve had the latest models of ChatGPT give some surprisingly decent fits in extended thinking mode. It looks up other fits, it looks up discussion about the hull and mechanics of what I’m telling it to do on the forums and reddit, and it checks everef to make sure the fits are actually feasible. Case in point, an abyss t2 solo Hawk fit I just had it make.
I’ll verify current hull bonuses and abyssal weather effects, then give one practical T2 fit with piloting notes rather than a generic list.
The hull points strongly toward Exotic: Hawk damage is kinetic-locked for its hull bonus, and Exotic abyss weather lowers kinetic resistance. Gamma helps shield HP, but it does not amplify the Hawk’s main damage profile.
Thought for 1m 58s
Assuming solo Hawk: run T2 Dark, not Exotic.
On paper, Exotic matches the Hawk’s kinetic missile bonus; Exotic reduces kinetic resistance, and the Hawk gets kinetic Light Missile/Rocket damage, launcher ROF, missile velocity, and shield booster bonuses. Practically, for a solo T2 rocket Hawk, the current community baseline is Dark: Dark boosts velocity and penalizes weapon range, and the known T2 solo Rocket Hawk guidance specifically says do not run that AB rocket setup in Exotic because Harrowing Vedmaks are a problem.
[Hawk, T2 Dark Skirmish AB Rocket Hawk]
Ballistic Control System II
Ballistic Control System II
1MN Afterburner II
Small Cap Battery II
Compact EM Shield Amplifier
Small Compact Pb-Acid Cap Battery
Small Shield Booster II
Rocket Launcher II, Scourge Rage Rocket
Rocket Launcher II, Scourge Rage Rocket
200mm AutoCannon II, Hail S
Rocket Launcher II, Scourge Rage Rocket
Rocket Launcher II, Scourge Rage Rocket
Small Warhead Calefaction Catalyst II
Small Hydraulic Bay Thrusters I
Hail S x1500
Scourge Rage Rocket x3000
Scourge Javelin Rocket x3200
Mjolnir Rage Rocket x3000
Mjolnir Javelin Rocket x3000
Nova Rage Rocket x3000
Nova Javelin Rocket x3000
Inferno Rage Rocket x3000
Inferno Javelin Rocket x3000
Nanite Repair Paste x100
Agitated Dark Filament x3
Use Agitated Dark Filaments. T2 abyss is “Agitated,” and frigate abyss entry requires the correct frigate filament setup; CCP’s abyss page notes co-op frigate traces require same type/tier filaments and allow up to three frigates.
Piloting priorities: kill neuters first, use Rage when you can apply, swap to Javelin when range control matters, and treat the 20-minute timer as the real fail condition. Community abyss guidance is blunt that many losses come from not killing fast enough rather than pure tank failure.
Reality check: solo T2 frig abyss is not fully “safe.” Recent discussion around this exact Hawk fit type says solo T2 frigates are dicey and that T2 Dark Hawk can work, but bad room chains can still kill or time you out. Run several T1 Darks first with the same hull before taking this into T2.
The fit seems fine, and the advice is the same I’d give. I wouldn’t load up all the extra ammos, since the Hawk is effectively locked to kinetic anyway, but it’s not wrong per se.
Exactly.
I see LLMs give incorrect answers often enough when I do know about the topic to realize that the answers I get when I do not know about the topic may not be reliable either even if it seems very confident and reasonable.
That could work, and for better models I’m sure it can give pretty good answers and keep hallucinations to a minimum.
But the EVE AI answers most players see whenever they do an online search, like the google search AI mentioned in the opening post, are not going to be of that quality.
Often it can get me some links, but makes wrong connections and concludes things about one part of the game while the link is talking about another.
And that’s the issue, asking general questions without any specifics will generate different answers… Questions should be direct and concise with specifically known info about the subject to get accurate answers..
Like they say, the devil is in the details..
Exactly, have it verify itself…
DeMichael Crimson, I do not know what your goal is in this thread.
Arthur gave an example of an AI that clearly is hallucinating, I then state that AI is often hallucinating about EVE and not a reliable source of information.
Then you attempt to convince us that AI can give correct answers if you ‘give the right questions’ and ‘let it verify its own answers’.
The fact that AI can sometimes give correct answers is irrelevant to what I said, what this thread is about is that it also often hallucinates wrong answers.
A broken clock too is right twice a day, but that doesn’t make it a helpful tool.
When I or others are looking for info on EVE in a search engine like google the search engine AI already generates some answers to show at the top above all the search results even without people asking for it, which often gives hallucinated information as shown in the opening post.
I am not interested in going back and forth with the AI to tweak my question just so that AI might or might not give a better answer this time. I’m trying to find a correct answer my question.
I am not interested in letting the AI to ‘verify itself’ (and who says it’s not hallucinating twice in a row? It could just be wrong again) just to let the AI finally get me the right answer, because my goal isn’t “to let AI give me the right answer”. My goal is “to get the right answer”. And AI in my experience appears to hinder rather than help too frequently for it to be a reliable source of information about EVE.
I think the moral of the story is: Don’t trust AI further than you can delete it.
I did a fresh install of windows a couple of weeks ago.
AI was really helpful generating me a list of all the AI tools I had installed and how to disable them!
I already answered that question…
What’s very telling is how polite, short and sweet your reply is to Anna Dyneaux when her point is basically the same as mine yet your reply towards me is nothing short of a full on hot air rant…
Yes, very telling indeed…
I get the feeling that you feel the need to defend AI for whatever reason.
And that my arguments against AI feel like a personal attack to you, for whatever reason.
There’s no need, I know AI has it’s uses. I also know that it’s hallucinating too often (see opening post for an example!) for me to use it for EVE info when I have enough knowledge of the topic to do my own research more efficiently without it.
Fixed it for you.
The only people against AI are those who exploit others with low intelligence. AI evens the playing field.