They claim it is AGI. So... what's your veredict, /g/?
>>109730196i think they deserve $1 trillion more dollars
>>109730196idgaf what they claim at this point.Sam will probably meet Luigi soon.
>>109730207Why not 10?
you can only use it on pro reasoning in chat which is 50 messages a day on the bigly planguess they got tired of people making bridge mcps for unlimited tokens
>>109730196AGI is here
>>109730298Very anti semetic.
>>109730196The current models are smarter and more generally intelligent than jeets and women. If (You) can't already see this then the model is already more intelligent than (You).
>>109730327That's not a high bar.
>>109730298imagine unironically believing that muse 1.3 is better than even gpt5.5 medium
>>10973034Agreed, but if we as a society consider women "generally intelligent" then I think it's fair to grant that title to the LLMs at this point.
>>109730196verdict for what? the model is not released and all we have is a blogpost about benchmaxxing
>>109730450Its literally out for plus users. Nigger.
>>109730196What jew is in charge of this one?
>>109730196>They claim it is AGIscam altman has claimed agi every year since like 2023
>>109730298https://x.com/maxbittker/status/2095990485177229777
>>109730196I'm a Grok guy
>>109730479not for me
>>109730196>first prompt>ask him to do one task>do something else and tell me it failed>took 16 min, consumed 4% of my x5 weekly limit>ask him why he didn't do the task I asked>tell me that he read what I wrote in another chat and thought I wanted to do this instead of the prompt I askedhere your AGI bro
>>109730298they already fixed it chud
>>109730479>>109730609not for me either.......very sad
>>109730298>>109730657>it's ACK-GIlmao media took meme-agi-3 benchmark too literally. It was just a robotic test in a virtual game nothing special.
>>109730594>>109730657where does qwen3 stand in comparison?
>>109730698
>>109730594Where is the human control?
>>109730745he asked about qwen3 not queer3
>>109730676They specifically created a harness to do better in arc-agi benchmark.Without that harness they get around 60%.
>>109730828Exhibit A.
>>109730196Are there any actual AGI tests?
>>109730828>>109730868It's literally the complete oppositeThe custom harness is ARC-AGI's which removes reasoning and autocompactionTheir 100% score was achieved by just using their standard API harness which uses reasoning and autocompaction same as everywhere else
>>109730882i think this just scored 97% on some math benchmark and its scary good at code, dont know about agi but looks good on theory
Does anyone have it in Codex but not ChatGPT?
>>109730196Verdict: LOL
>>109730887bro, you fell for the marketing meme.They just engineered a model that will pass this test for pure marketing scam. That's about it. There is nothing AGI-ish about it. It's just manipulation of your already turd filled brain.AGI is not possible with current ML architectures. It's not happening.
>>109730972https://arcprize.org/blog/astra
>>109730868"heh it only scored 63% while blindfolded and with its hands tied, nevermind that the previous generation of models scored like 5% before"
>>109730990>>109730972
>>109730887And that makes it incomparable to previous test results.
>>109730972The "Provider Adapter" harness is literally the same harness that is used on every request to Astra right now. There is nothing engineered specifically towards ARC about it.Plus they couldn't even really benchmaxx it in training because it's a private set.It also blew every model out of the water playing Runescape.
>>109731093>It also blew every model out of the water playing Runescape.That can absolutely be engineered for, though.Also, no human control on that.
>>109731068It's comparable to Sol which got 8% on ARC harness and 30% on standard harness
>>109731129>and 30% on standard harnessEstimated.
>>109730196Still haven't seen consensus on what defines AGI lol
>>109730671it's codex only for plus
>>109730196Probably not AGI, but Astra is helping me build a fuckbot so that's still pretty cool.https://litter.catbox.moe/3o4o6knfja1tdr9w.png
>>109731392>Wombforce 9000sounds intimidating
>>109730196>what's your veredict, /g/?snailcats obliterated
>>109730643it might as well bebut all that architecture and spending can only hold enough attention for a single user at the timebiology: 1silicon: 0
>>109731346If CEOs aren't replacing their executives with AI, then it's not AGI yet.
>>109731426The executives on up could have been replaced two years ago.But this was never about increasing efficiency or maximizing company profit (not having to may any execs would save more than cutting workers), it's about reinstitutionalizing slavery.
>>109730196has it solved Collatz yet?
