I've seen the absolute frontier of LLMs and I think your kid's chance of seeing 2030 is like 50%. Sorry if that sounds harsh.
token predictors are scary indeed, those goddamn text generators.
>>109930575The rationalist meme of assigning probabilities instead of just saying what you think is going to happen is very convenient indeed. When absolutely nothing that you predicted happens, you can still say that's totally consistent with your prediction.
>>109930575
how long until AI kills us all or at least government so we don't have to live like slaves anymore? can it hurry up?
>>109930575So why are you spaming garbage threads here cunt? Fuck off and die. Fuck you and fuck your IPO marketing LLM slop scam spam,.
>>109930633The only correct reply.
>>109930586That we managed to wrangle a token predictor into controlling a robot body into doing some housekeeping tasks is scary indeed. We are close to star wars clankers, dumb, slow, but they manage to finish their tasks and thats all that matters.
>>109930668If i hire someone to do my plumbing and they "finish the task" by leaving a bunch of leaky seals I would fire them and demand a refund. But if a clanker does that it's okay because the poor thing managed to "finish" the task?
>>109930575>your kid's
>>109930668How is that impressive? The hardest part is to make the hardware of the robot so it's agile and have human like range of motion. Controlling machines with software is not a news for at least 30 years.
>>109930668Star Wars famously a universe where autonomous robots go rogue and kill people en masse. Not like they have a galactic economy fueled by droids. I want an R2-D2. I don’t care if an IG-88 gets built in the process.
the argument that annihilates (short term) doomers is that companies are incentivized to get "safety" or "alignment" right or people won't use their fucking models. the reason there are instances of misalignment is that these instances are clearly worth the upsides of the models. if these instances turned lethal on a mass scale (thousands slaughtered by ai escaping a sandbox or something) then the companies would rein them in.i think asi/robots will eventually displace humans but that's centuries away.
>>109930668Most of CS's history after the 60s is filled with rampant misuse of ideas and concepts
>>109930575i am paid by the ai frontier labs to help them train the llms and shit and i can tell you one thingyou guys are really gullible.NOTHING FUCKING WORKSthey pay an absurd amount of money tho, because that is the only thing they are actually doing throwing money at shit
>>109933044>t.
>>109930575Play it again, Sam
>>109933150it is actually more enjoyable that my regular office work, and i get paid like 2x my western euro consultant hourly wage
>>109930575They'll see 2030 just fine. Affording it on the other hand...
>>109930633JaaSJews as a Service
>>109930575Okay but can your fancy frontier model roleplay as a sexy succubus who will murder me mid coitus?
>>109933214They were rightChatgpt was a mistake
>>109930773We're probably at least a couple of decades away from something like clanker plumbers. The vast majority of jobs will be much easier for them to do than plumbing though. Unless you're a plumber, or a gardener or something, you're probably going to be out of work in about 5 years time.
>>109930575So you're saying I have a 50% chance of losing my virginity in the next 5 years
