Not sure if this is anything.. but..

[TheWarmaster]

I just googled what the next Warhammer+ models would be and AI told me this for years 5.. Not sure if anything has been announced but I thought it was worth sharing.. anybody got any thoughts? 

[Grotsmasha]

IIRC new Sub year starts in September, and they usually announce the new models in August, and to my best recallection, yrs 1-4 didn't leak.

[Mmmmm Napalm]

heed not the whispers of the abominable intelligence 

[Lord Marshal]

Hadn't AoS already gotten a W+ Sorcerer model the other year?

 

AI's prolly just scraping the internet and cobbling an answer together.

[irlLordy]

AI will hallucinate and make up stuff and present it in very confident ways. Unless theres a link to an actual warcom article, don't trust it, its just repeating back speculation at best, or outright making up stuff.

[Crimson Longinus]

I'd trust tea leaves over this. 

[Orange Knight]

Google AI recently told me there are no dual type Psychic/Grass Pokemon. I can personally name 5.

It's not worth paying attention to

[Cactus]

9 minutes ago, Orange Knight said:

Google AI recently told me there are no dual type Psychic/Grass Pokemon. I can personally name 5.

It's not worth paying attention to

If there's anything that's widely catalogued in tables that are easy for a computer to ingest, it's Pokémon. Getting that wrong is shocking (but not surprising).

 

Generative AI makes guesses about what words go together like predictive text does. I'm not surprised to see it predict an AoS sorcerer, I think there have been a couple in the W+ miniatures already, but it's still an uninformed guess.

[Orange Knight]

2 minutes ago, Cactus said:

If there's anything that's widely catalogued in tables that are easy for a computer to ingest, it's Pokémon. Getting that wrong is shocking (but not surprising).

 

 

This is exactly why I bring it up. It's honestly shocking.

[Casual Heresy]

AI is as useful as a chocolate tea pot. You'd be better off doing you own research and making your own deductions than listening to the half baked, sycophantic raving of an Abominable Intelligence.

 

Detestation of the unclean machine minds aside, I believe that @Grotsmasha is correct that the last four years Warhammer+ models haven't really leaked, with on some rumours popping up from the usual vendors either right before release or not at all.

[Matcap86]

39 minutes ago, Orange Knight said:

Google AI recently told me there are no dual type Psychic/Grass Pokemon. I can personally name 5.

It's not worth paying attention to

Google AI got the definition of consonants and vowels wrong the other day. Since then I don't google stuff anymore. Especially since they're now promoting AI generated stuff over human made stuff. 

[grailkeeper]

A quick google throws up nothing for either name for me. 

[grailkeeper]

This is a hallucination, but I suppose there's a small chance that someone in GW uploaded the real details into an AI at some point, and the info is being spat back out at us now. If that's true I'd bet they're having a really bad day at work right now. 

[Antarius]

2 hours ago, irlLordy said:

AI will hallucinate and make up stuff and present it in very confident ways.

Indeed. My rule of thumb for people who put a lot of stock in AI's use as a research tool is "ask AI about something you know a lot about yourself and see what it answers, then ask yourself if you really want to trust it about something you don't know anything about". 

[HeadlessCross]

This post is proof that the Industrial Revolution and its consequences have been disastrous for the human race. 

[Hfran Morkai]

I still remember someone asking Grok to analyse an X-Ray where it was quite clearly severely fractured, coming from someone who is a total layman at viewing and analysing X-Rays.

 

It said "minor anomalies".

[The Praetorian of Inwit]

More importantly what about the shareholder value?

[Evil Eye]

6 hours ago, HeadlessCross said:

This post is proof that the Industrial Revolution and its consequences have been disastrous for the human race. 

Aww, I knew that already!

[Laurence]

11 hours ago, TheWarmaster said:

I just googled what the next Warhammer+ models would be and AI told me this for years 5.. Not sure if anything has been announced but I thought it was worth sharing.. anybody got any thoughts? 

 

This is all wrong as Year 5 is the current year of Warhammer Plus, and the Year 5 miniatures are Infinity's Lament and The Summons.

[LSM]

12 hours ago, irlLordy said:

AI will hallucinate and make up stuff and present it in very confident ways.

 

A year-and-a-bit-ago, someone on reddit posed a question, and I did my best to answer it; with the caveat that I was doing so based on some half-remembered thing I read thirty years ago as a kid. 

 

A bit-ago, someone on reddit posed the same question, and I was typing out the same answer when I thought: "hold on, I've got nothing better to do. I bet I can track down an actual proper source."

 

Whilst attempting to google various things, hoping to turn up where I had read it (and maybe find a scanned pdf), I noticed that (at least) the AI seemed to be in agreement with me; confidently stating the same as my vague recollection from a year previous.

 

Eventually, I saw that the AI gave a source for its statement, and I thought: "Aha! Maybe it will be able to tell me where this came from!" And I clicked through, only to find...

 

The AI's source was my reddit post from the year previous, presented with none of my caveats. 

 

//

 

Generative-AI is a fantastic tool, but it should be treated like paying a random unskilled worker (who lives millions of times faster than you) to help you out. They're very fast, but it's a coinflip whether or not they actually "get it".

 

Edited Wednesday at 11:55 PM by LSM

[Magos Takatus]

Yeah, if AI didn't seem to love sourcing so much from Reddit. Don't get me wrong, there are undoubtedly a lot of very clever people on Reddit but even then, it's not a verified source I'd rely on. LSM's  anecdote is a perfect example. The answer is probably correct, but becoming your own source has to feel weird, and the fact that it didn't pick up on the uncertainty is troubling.

[Antarius]

Plus, it's really eager to please, so even without realising it, you might still end up slanting its answers quite a lot by the way you phrase them.

They've tightened this up somewhat, at least on the very obvious high-profile stuff - for example, I can no longer get co-pilot to agree that the Earth is flat, at least not in the first three responses - but that seems like something they'd have put a bit of work into actively preventing, so I would bet that you can still nudge it to give wrong answers on a lot of other things. And of course, it simply has no conception of whether an answer is right or wrong, which is the real crux of the problem.

[Magos Takatus]

I've also found that depending on the sources it picks from there can be a lot of bias even when objective truth isn't the issue. Without going into specifics as it's out of the board's scope but I experimented on the definition of certain words, many of which it considered slurs, yet if I searched for words used as insults by another opposing group of people it did not consider them slurs, which to me indicates there was a definite political bias in the sources it was pulling from.

 

Even if the quality of AI improves, I think being wary of all the answers it provides is good practice.

[Evil Eye]

To be absolutely fair, judging AI as a whole by Google Gemini is like judging all cars by the Reliant Robin. Other AI can be a little unreliable depending on the subject but Gemini suggests using glue on pizza and that Google has two Ps in it.

 

I have found ChatGPT useful for technical troubleshooting, especially with 3D printing, but I still take its advice with a grain of salt and if I think it's getting confused I will double check with it, which usually works and gets it to see its error. You definitely have to know how to work with AI, I find.

 

That said clearly the best AI is Gork. Isn't that right Gork?

 

"Yeh"

[Magos Takatus]

The one I tried using was the Brave Browser AI rather than Google's version in this case but it was just another point in the graph of my scepticism toward the answers AI give me.