AGI soon!
>>107259971>oh no the super small model Google serves to billion of queries isn't as smart as Gemini 3/GPT5
google AI equally sucks like their search. kek
>>107259986>erm actually it's ok that the ai feeding information to billions of people is retarded!
>>107260022are these billions of people directly paying google for a good experience?
>>107260033yes, they're selling their data to jewgle. they're implicitly paying a price in excess of google's shitty AI's cost of service.
>>107260053>implicitly paying>in excess of google's shitty AI's cost of service.so you know exactly how much this costs google to serve?
works on my machine
>>107260176Was that directly on google search?
>>107260268works on my machine :)
>>107260022that google is using a shitty model that hallucinates sometimes for search results is a non-sequitur and does not really say anything about the progression of the state of the art
>>107260160>AI is a glorified yes-man>AI = All IndiansI see no difference here
>>107259971>what is inspect editHell fucking YES we are getting AGI soon, your bullshit luddite propaganda doesn't put a fucking dent in the coming of the singularity.I hate reddit as much as the next guy but DAMN sometimes they have some good ideas, I wish this place would adopt a no decels rule like the accelerate subreddit.
>>107260348Prove it. Post a screenshot of what you get when you search for that exact query on google search, without going into the "dive deeper" button
>>107260068no because their SEC filings don't intend to disaggregate their expenses at that level of detail.
>>107260068>"jewgle is losing money for your search queries! It's free because they love you and give you a free experience, not because they beta test on you like a monkey"Go back
>>107259971AGI "soon" but LLMs still can't write good without unrealistic or flowery garbage
>>107262246garbage in garbage out, if the models were trained exclusively on my posts then agi would be here
>>107260724The fact that the money spent on running the model has to be lower than the amount of money made from ads is exactly the point that guy was trying to prove, you utter baboon.
>>107259971
>>107259971How hard is it for them to add an extra pass afterwards "Does this answer makes sense?" Sometimes it does but doesn't answer the question and I have to reply "Reread the question and the answer. Does it makes sense?" at which point the faggot ai usually double downs "Yes, I am always right and you are wrong to question my non-sequitur. I'm not listrening, lalalala"
>>107262346IME you can get it to agree with almost anything
>>107262265https://archive.org/details/gpt4chan_model_float16
>>107259971The majority of sources indicate the AI model used to produce the quick AI overviews when in general Google search is Gemini 2.0 Flash
>>107262725This. This is the problem with every LLM. They can't apply even basic logic to real world situations because in order to do so, they would have to have some kind of a model of how reality works - similar to how every human automatically creates a mental model of all his observed reality in his mind - and none of them have anything close to that ability. This is why we will never get AGI out of this shit. It will never stop hallucinating until it's enabled to create a persistent model of the world.
