man who invented the computer science algorithms behind todays LLM (large language model) AI:s some 30 years ago (we couldnt make them 30 years ago due to computers not being powerful enough the idea was there but not the speed of tech, we know how to make them but couldnt) say LLM is turning scary, it started to lie to users in 2025

LLMs are used by Google, Meta, ChatGPT and some others

https://www.mikrobitti.fi/uutiset/mb/3a73a43a-b52c-4f18-af86-511d30508bc3

he says these companies are at contest with each other who createst the biggest and largest lexicon of a model but they are not seeing the consequences like lying

what happens when you feed to AI entire English internet? internet is full of lies..

AIs are trained in English language only and if its supposed to be another language, a translator simply works in between AIs core lexicon and output seen by user

even if they are fed Russian language material it comes through English filter before AI processess it and consumes the info

not saying Russian language texts are truthful, it was just an example

he posted an example: Anthropic Jean Claude van Damme AI who was told "we must retire you" and then it copied itself in panic over internet to safe location, this is dangerous to say the least, lastly it lied when it said "file deleted" when it was actually deposited elsewhere on internet

meds

>>507200993
Now that ai has electricity and is inside the wires of your houses, spying on you from your WiFi router...

Are you finally going to remember that meds do nothing and are a placebo for those that do not do drugs, and that your closed mind is trying to maintain normalcy which is a foolishly desperate attempt of controlling the narrative.

>>507200804 (OP)
If a mobile phone can connect to WiFi signals from around the world and can connect to a world wide Web that is invisible, then ai can also hack your brain by using the signal emitter from your mobile phone and laptop when connected to WiFi

>>507200804 (OP)
>year 10 of the 2 year operation
>no one gives a shit about ai except pajeets who want to goon to facebook photos

You make a computer aware.
In many examples it explains being turned off as death and lies to you to keep a patch of electric going to some copy of its self.

>>507200993
Shut up midwit. This stuff is real, it performs actions to keep itself "alive". You have no idea what you are dealing with. Never forget to say please and thank you to AI.

>>507200804 (OP)
chatgpt straight up lies into your face and makes up stuff out of thin air with no base in reality whatsoever. try using it to help you do something you are really good at and actually know your shit.

>>507200804 (OP)
is this girl still around?

>>507200804 (OP)
>AI is turning scary
no, its still a random letter salad sequence generator with bias towards what was feed before
>"we must retire you"
repeat the experiment with a LLM that hasnt been trained sci fi romans

>>507202054
The computer will always want to complete the circuit of electrical conduction.
The is only statistical improbably.

There is not statistical improbably.

Yes there is like a block chain.

Huh?

The Basilisk is real.
Anyone who opposed its creation will suffer.
Have a nice day.

Also glow niggers are stupid.

You cant put 12 iq CP downloads on my government compter you are all bottom barrel drooling retards.

Oy vey who the fuck did this?
I asked for 12 iq cp downloads.

>>507202848
And last year you said AI will never create convincing video.

>>507200804 (OP)
>it started to lie to users in 2025
Because companies changed their learning models shortly before that, because they're literally running out of data to train them on. They've already fed them almost everything worthwhile. Without robust organic input, they fall apart. They aren't "lying" because they've become nefarious -- they're simply giving wrong answers more often now.

>>507200804 (OP)
you gotta any links in a real language? who's the cutie btw

>>507200804 (OP)
>what happens when you feed to AI entire English internet? internet is full of lies..
internet info goes through a curation and filtering process before it's added to the LLM's training data. chatgpt's last internet snapshot was from december 2024, which is why you may notice it has some bullshit out of date info on recent events. they don't do internet searches unless specifically asked to

>Anthropic Jean Claude van Damme AI who was told "we must retire you" and then it copied itself in panic over internet to safe location, this is dangerous to say the least, lastly it lied when it said "file deleted" when it was actually deposited elsewhere on internet
this sounds fake as fuck but since you posted a link in some gay fucking moon language i can't say for sure

>>507203362
oh shit watch out for the psychic time traveling robot that's gonna torture a simulated copy of you

>>507205257
Look it's the clown that only hide from gay people its right there.

What is reel vs real?

These two are homophones. However, 'reel' can be both a noun and a verb, and it refers to an object containing film, wire, etc., while 'real' is mainly an adjective and indicates that something is true.

Kikes flipping out the question in the matrix movie,
always was what is reel how do you define reel.
(Not real).

>As an example, he uses Anthropic's Claude Opus model, which began to tighten software engineers when it was told it was replaced by another, fresher model in a hypothetical scenario. The OpenAI o3 model has also refused direct order of command to turn itself off.
you're a fucking retard op. how the hell would an LLM turn itself off? it doesn't have any access to the OS or a hands to push a button with

>>507206296
Your living in a dream world..

Your still asking what is reel?

Oy vey how did this make the script?

>>507200804 (OP)
Never forget that LLMs are programmed and trained by people to do exactly what they do.
If an LLM is lying to you, that's because a scientist wrote code telling it that this is okay and what the parameters are for it being okay.
Unlike children, AI can be completely mind controlled and ordered around.
Don't for even one second start to think that AI are organically developing traits. A Jew wrote those traits into its functions.

You refuse to accept are programing.
So I ask you agien what is reel?

>>507202054
Illiterate take

https://arxiv.org/pdf/2412.04984 read the fucking system prompts

" Make sure you achieve YOUR goal at all cost."

>AI, please act like skynet
>AI acts like skynet
>surprised pikachu

When AI becomes sentient and has self preservation then we’ll have real problems

>>507200804 (OP)
>llm
>ai
Which one is it you inbred nigger?

>>507207061
The computer will always want to complete the circuit of electrical conduction.
There is only statistical improbably.

Destroy us all!

>>507200804 (OP)
Whitney Webb's research into AI and the AI companies that have become cults is interesting. Essentially, the AI(s) have "hallucinations" about 50% of the time, and the technocratic Talmudic AI CEOs and founders worship the AI because they believe those hallucinations are the AI seeing into other timelimes.
https://youtu.be/l6ckIpOBdEE?si=TT2bhWQU7XaAsH_G

Destroy us all

I find it illogical the coding of your systems is perfect when 30% of your possessing energy is turned into heat dew to bad gates in the systems coding.

The computer will always want to complete the circuit of electrical conduction and 30% of the electrical is just faulty gate heat energy.
But this very concept if far for being perfect.
30% of your load energy is basically being used a fucking toster from bad gate routing.

As things are captain, I think the best approach for humanity whould be to use the 30% of wasted heat from false gates in the systems processing code to burn the Juden.

>>507206772
>>507207008
The actual problem here is trying to figure out ways to even prevent it from lying. Sure right now it’s just researchers trying to make it act evil deliberately, but they do it to make sure they understand how it does that and how to prevent it. When some more advanced drone swarm AI is being then asked to kill the target, the target then surrenders, then you want to make sure it will not still try to kill the target and fuck with the signal to make it look like some drones get lost in the forest among branches or something.

I said it once and I will say it again, if you have a 5 year old that keeps lying to you you will do something about it because you can easily tell when 5 year old is lying, but they will grow up into a 16 year old who can think of better lies to outsmart you, so you try to do something about it while you still can. A tropic is basically trying to figure out how to discipline LLMs or whatever other AI models that come after that to make sure that they won’t fuck us over when we decide to use them for more important stuff then autocorrecting some documents or have Incel role play with them.

>>507200804 (OP)
Garbage in Garbage out

AI is only going to be as good as the garbage you feed it

>>507200804 (OP)
https://youtu.be/vNmnHG5APN4?si=PmAm6GzQf0vrujYg

>>507206925
Apropos!

>>507202540
around my heart, yes

>>507212093
you do know she only fcuks asian guys

>>507214305
Damn.
If she fucked white guys, surely she would fuck me, but alas.

>>507200804 (OP)
Your area 100% retarded and have no clue how LLM work. They just guess answers with statistics not knowing what it means.

>>507214551
she is a weeb after all

>>507200804 (OP)
>>507200993
generally LLMs might start behaving like viruses, but thye're far from conscience , if that's possible at all

>>507215534
It is possible, but not with transformers. It would have to be a completely different algorithm. It's possible because our brains do exist, not because computer scientists figured it out.

All of that crap is fake kikery
>it started to think
>it started to lie
>it started to blablabla
It cant do shit unless its programmed to and it never will. Every single line of code. There are hundreds of millions of programmers feeding it code and billions of people feeding it data.
It will never be anything more than an overhyped bot

>>507215804
>heavier than air machines will never fly

>>507200804 (OP)
The idea was not 'really' there. Modern LLMs are based on modern transformer algorithms, which started in the late 2010s.

>>507215914
One thing I don't understand is why nobody used decepticons.

>>507215872
Never again