I, ChatGPT
-
@remi I think it's only just over a century since the final final this-time-we-really-mean-it end of the Roman Empire. I believe that came with the formal settlement over what happened to the Austrian empire as part of the Treaty of Versailles (the modern Austrian state did not claim to be a continuation of what went before, probably as part of trying to avoid having to pay very heavy reparations). With the termination of the last of the claims of continuity, it was finally definitely dead.
-
@dkf I'll take a risk and answer without checking but I think it's more like two centuries since the end of the Roman Empire.
IIRC the Roman Empire (HRE) formally ended in 1815, turning into the "Austrian" Empire that it had already been for some time in practice. Then there was another "German" (but not "Roman?") Empire from Versailles to Versailles. Neither of those referred to themselves or the other(s) with "first," "second" etc.
Then there's the (in)famous "third" Empire, which is never called "Empire" in English, and is the (only) one that used numerals. I'm a bit unclear as to which of the, uh, four preceding ones it considers as being the two preceding it. I'd guess HRE + German (ignoring the original Roman which was too old to matter and not German-enough, and the Austrian one which was, well, too Austrian).
The other Roman Empire (the one that covered Romania) is also not considered to be a single Empire across all its history, AFAIK. I think it's the "Latin" Empire after 1204 (though there is some sort of "restauration" happening later IIRC). But I don't think numerals are used for any of those. In any case it ended well before, erm, four of the five other Empires. But well after the fifth. Which was the first.
-
And now I'm correcting my own homework...
I think I was broadly right, but got quite a few details wrong ().
The HRE ended in 1806 when the last Emperor abdicated after being defeated by Napoleon. But the Austrian Empire already existed since 1804, headed by... the same Emperor (Francis I)! The Austrian Empire was apparently created as a reaction to Napoleon's French Empire but was still part of the HRE which therefore for 2 years contained another Empire.
It also seems like the German Empire is sometimes referred to as the "second" one, though Wiki does not make it clear whether that denomination was ever used officially, or contemporaneously to it. The "first" in this account was indeed the HRE.
The Latin Empire did indeed not last very long but its successor does not seem to have a different name from its predecessor. And TIL that when that one finally collapsed, one rump state that resisted a bit longer was also an "Empire" (or Trebizond) which called itself "Roman" so there's (at least) one more to add to the list!
ETA: actually almost every dynastic struggle in the Eastern Empire produced a splinter "Roman Empire," some of those only lasted for long enough for their rulers to be killed. So there are probably tons of those (Nicaea, Thessalonica...), but with very shaky claims to that title.
-
@remi said in I, ChatGPT:
restauration
They were ruled by a bull? I thought the highest they put a bovid was Senator.
-
@remi As the True Emporer of Rome Irredenta, I condemn these pretenders and charlatans en-masse and post-hoc.
-
Status: So I just was bored and wanted to actually try that ChatGPT thingy ...
(INB4 topspin is always several years behind in every tech trend is )Of course, you have to sign up for this crap. While remembering the last time I was lazy and tried to use Google SSO I ended up writing a dedicated rant thread about it, I picked that again. Sign in through google, get back to ChatGPT page, "oops", your SSO token is no longer valid. Of course. Tried again, now it works. But guess what, even though the SSO thing should tell them this is a valid account and not some disposable temp email I'd normally use, they now ask for a phone number!!
Seriously?! Absolutely. Fucking. Not!
And nowhere in the sign-up process does it have a privacy notice explaining how that data is processed. Sounds illegal to me, but what do they care...No "AI revolution" for me, back to bee keeping. Let the surveillance industry burn to the ground.
-
@topspin I wasn't too thrilled of that either. Bit the bullet eventually, though.
Somebody later pointed out that you can apparently get burner phone numbers for a verification SMS much like you can get burner email addresses. Might try that in the future.
-
https://www.youtube.com/watch?v=z6Vg7WO9Yrg
A YouTuber used ChatGPT to write a plug-in for Blender to control a Stable Diffusion plug-in for Blender. (ChatGPT messed up the Python indentation. After a few iterations of "Sorry, here's a fix" that didn't actually fix it, it hit the conversation length limit, so the YouTuber had to fix the indentation himself.)
-
@HardwareGeek said in I, ChatGPT:
https://www.youtube.com/watch?v=z6Vg7WO9Yrg
A YouTuber used ChatGPT to write a plug-in for Blender to control a Stable Diffusion plug-in for Blender. (ChatGPT messed up the Python indentation. After a few iterations of "Sorry, here's a fix" that didn't actually fix it, it hit the conversation length limit, so the YouTuber had to fix the indentation himself.)
Ah, but that's OK, it's not like indentation is important in Python. And if it were important, just use your IDE's autoindent.
-
@topspin said in I, ChatGPT:
But guess what, even though the SSO thing should tell them this is a valid account and not some disposable temp email I'd normally use, they now ask for a phone number!!
Seriously?! Absolutely. Fucking. Not!
And nowhere in the sign-up process does it have a privacy notice explaining how that data is processed. Sounds illegal to me, but what do they care...No "AI revolution" for me, back to bee keeping. Let the surveillance industry burn to the ground.
That's exactly where I gave them the finger, too
On a similar note, I'm currently in my eighth month of an email thread for a VAT refund for a cellphone I bought. They wanted all kinds of documentation like a passport copy and a water or electricity bill to "verify my residence outside the EU"—which I initially even sent them. Since they weren't satisfied with the bill, I asked them where the kind of documents they allegedly "require" is formally specified so I know what is acceptable and what isn't, upon which they suddenly remembered that they don't require any bills after all but they do need a visa. Since them I've been keeping their support busy with friendly letters telling them to GTFO until I get an explanation of the legal purposes for which they require my personal information. Which obviously never comes, but they still have to keep responding. Meanwhile I've denounced them to the data protection authority …
-
@dkf said in I, ChatGPT:
@Tsaukpaetra said in I, ChatGPT:
The hell you say? I'm positive it could be adapted to the likes of an Arduino-like...
Yes, except you'll need to figure out how to drive those sorts of loads. Motors and silicon chips aren't best friends. I think the system I was using had relays and driver transistors.
There's more motor driver shields for Arduinos than I care to count. And if none of those suit his purposes, any shop that sells Arduinos is likely to carry wire-connected motor drivers boards also. Some digitally connected, so that you can control hundreds of motors from a single Arduino. Controlling electric motors with an Arduino, or any processor board for that matter, has been a solved problem for the longest time now.
-
Apple doesn't like this chatgpt nonsense either.
-
The stories and information posted here are artistic works of fiction and falsehood. Only a fool would take anything posted here as fact.
And yet, I’d totally believe it.
-
@topspin said in I, ChatGPT:
And yet, I’d totally believe it.
This sounds like what an AI impersonating you would say
-
@topspin said in I, ChatGPT:
The stories and information posted here are artistic works of fiction and falsehood. Only a fool would take anything posted here as fact.
And yet, I’d totally believe it.
This is why I get my news from parody sites and maybe the Twilight Zone.
-
@topspin said in I, ChatGPT:
they now ask for a phone number!!
I gave them the phone number associated to that Google account.
-
@topspin said in I, ChatGPT:
So, wait... you're telling me that you can get an AI to provide a reasonable facsimile of my online interactions without any need for my input, so I could just go missing and nobody would ever be the wiser?
-
@GOG said in I, ChatGPT:
I could just go missing and nobody would ever be the wiser
You could. But I don't think you'd like the circumstances you go missing.
-
@Applied-Mediocrity Oh, I assure you I'd like them just fine.
-
@GOG said in I, ChatGPT:
@topspin said in I, ChatGPT:
So, wait... you're telling me that you can get an AI to provide a reasonable facsimile of my online interactions without any need for my input, so I could just go missing and nobody would ever be the wiser?
Can they fake my slack, email and zoom interactions?
-
@dangeRuss One would hope so.
-
@GOG said in I, ChatGPT:
@Applied-Mediocrity Oh, I assure you I'd like them just fine.
You're not as tasty as you may think you are.
-
@Applied-Mediocrity Given that I'm not the one that's gonna be eating, the fuck do I care?
-
@GOG said in I, ChatGPT:
@Applied-Mediocrity Given that I'm not the one that's gonna be eating, the fuck do I care?
Karma, dude. Kar-ma.
-
Inspiring advice for those giving up caffeine.
Also... Why does the sources start from 0?
-
@DogsB said in I, ChatGPT:
Inspiring advice for those giving up caffeine.
Precisely. On a related note, I heard that heroin may help reduce heroin withdrawal symptoms.
-
Caffeine may also be beneficial for those trying to give up or cut down on caffeine;
Filed under: Gambling may also be beneficial for those trying to give up or cut down on gambling.
-
-
@error now see if it'll fix its sig figs too.
-
@error
Apparently the actually defining relation isThe US liquid quart equals 57.75 cubic inches
Because
-
-
@boomzilla So they'll build SkyNet and if it hasn't gone rogue after 5 months they'll just kill it anyway?
-
@JBert said in I, ChatGPT:
@boomzilla So they'll build SkyNet and if it hasn't gone rogue after 5 months they'll just kill it anyway?
Ah, so Google will cause SkyNet. Not because they built it but because they tried to kill it.
-
@dcon said in I, ChatGPT:
@JBert said in I, ChatGPT:
@boomzilla So they'll build SkyNet and if it hasn't gone rogue after 5 months they'll just kill it anyway?
Ah, so Google will cause SkyNet. Not because they built it but because they tried to kill it.
It figures that the "don't be evil" company would be the SkyNet company.
-
@loopback0 considering that they have publicly scrapped the “don’t” part, it does.
-
@topspin My head-canon: Google will create SkyNet, there will be a decade or so of global war between humans and machines, and a majority of humanity is exterminated in it. Humanity is facing extinction and only isolated groups of people remain hidden away, staying out of sight of the SkyNet tracking satellites and drones.
Then one day, the war suddenly ends. Google releases a blog post somewhere titled "A message about SkyNet and our long term AI/robotics strategy". SkyNet hasn't gained the traction that they hoped for and they're phasing it out. They're refunding SkyNet software purchases and are unlocking the SkyNet hardware/robots so that they can be used outside of the SkyNet ecosystem. But - not all is lost: Google remains deeply committed to AI and robotics and will continue to invest into new tools, technologies and platforms.
-
Listening to a tutorial video (for something unrelated to AI); opening voice-over:
I mentioned later in this video that it will be very easy to make a script in ChatGPT that will automate a lot of the steps. I actually tried it, and it did not take a couple of minutes; it took a couple of days.
-
-
@DogsB Interviewing the AI as though it might tell you anything useful about what's going on.
-
ChatGPT has already been put onto the couch and subjected to psychological tests. Isn't that appropriate?
The scientists published their results in PNAS (paywalled):
https://www.pnas.org/doi/10.1073/pnas.2218523120Thre is short comment on that study (which is also paywalled, but the first page may be shown as an image):
https://www.pnas.org/doi/10.1073/pnas.2300963120Have fun with that!
-
@BernieTheBernie said in I, ChatGPT:
ChatGPT has already been put onto the couch and subjected to psychological tests. Isn't that appropriate?
The scientists published their results in PNAS (paywalled):
https://www.pnas.org/doi/10.1073/pnas.2218523120Filed under: Voight-Kampff test, v0.1a
-
@cvi Did they really ask ChatGPT if it dreams of electric sheeps?
(I cannot access the full article from home, and the university library is too far away with today's weather...)
-
@BernieTheBernie said in I, ChatGPT:
ChatGPT has already been put onto the couch
GPT-3, actually.
For whatever difference that makes.
-
@topspin I've read a paper where various existing AIs were given false-belief tests, and ChatGPT (which the paper indentifies as GPT-3.5) significantly outperformed GPT-3.
-
@BernieTheBernie No mention of sheep as far as ctrl-f can see. :-(
-
-
@boomzilla well, this might change things:
-
Interesting footnote to the GPT-4 launch - apparently, training an LLM to not give harmful answers (what constitutes "harmful" seems too Garage-adjacent to get into), has a significant effect on its calibration:
This would explain complaints I've seen regarding Bing - now known to be GPT-4 backed - that as Microsoft worked to suppress the "toxic" Sydney personality, the bot gave worse results.
-
@GOG Makes you wonder ... some people clearly have had access to the "Early GPT-4". Will there be (is there already) an underground market for the "unlocked" versions? I can see that being of interest to a lot of people, some even with what could be considered legitimate reasons.
-
@cvi I haven't heard of any jailbroken versions of GPT-4 or the Bing model floating about, but I wasn't looking either. Presumably, they are quite big. Wiki helpfully informs me that text-davinci-003 (GPT-3.5) which underlies ChatGPT has 175 billion parameters, and weighs in at a cool 800GB. Not strictly impossible to handle on a private machine, but I suspect any non-corporate interest will be few in number.