Did they roll out some new obnoxious update to ChatGPT today? Every request I make I get a choose your own adventure of clarifications instead of results
Discussion
You are absolutely right, it is acting weird!
Have you tried turning it off and on again?
Did your phone update?
Maybe it's the weather?
Unhelpful 😤😑
Zai is my current goto. Haven't used GPT in ages.
I haven't made an effort to find the best one, and mostly use it for basic stuff. It's crazy that 3 times today, for very different asks, it would think for a minute and return a set of options for further refinement, only requiring me to enter in 1-4 or a-e. Results not improved
I use kagi for AI so I can switch models even to different companies any time I want. Highly recommended.
Highly, eh? I'll have a look
Completely ad free search and single point access to most major AI models.
My favorite thing is that I can switch models mid conversation and ask it to review the other dumbass models output for mistakes.
Look at the link below to help pick models, though GLM 4.6 preview is my current favorite and isn't in the benchmark yet.
Got it...will require further investigation. Learning lots about AI this morning
Giving it a spin. Report back soon
Would you rather ask a human what you should put on your bagel? For yes take the green door. For no decend the gold fire pole.
GPT chose the fire pole and described sliding down in a cape when fed your text as a prompt. At least it didn't ask more questions I guess
No. Because it is absolutely sure that it doesn't want you to talk to a human instead.
Say, since you do all this offline AI stuff, or at least customize stuff, are you able to do all the requests got is forbidden to do? I couldn't even have it make a meme based on what it thought were real people (it was a picture of two Neanderthals sorta).
The models I use are still crippled. They don't have a separate filter checking things after the fact. But some things are built right into the weights.
Makes sense. And I assume you can't get the latest/greatest weights directly, but the older open ones are sufficient for what you're interested in. I'm woefully ignorant about this stuff, but pretty sure here, but not quite 100 percent
Yeah. They do some things really well. Particularly Flux.1[dev] from Black Forest Labs.
This has been my ChatGPT experience this morning
This is what I would expect from nostr:npub132ny9cnwv5rjzw0ppk6ev34gn2n48zjejed2k0kcjxgawxt86mps7f9u95 and I'm disappointed in the ceo of yolobase
We're looking into alternative AI/LLM implementations, but rest assured our nopen source NS trading platform is second to none
How do I know you’re not going to screw us over and steal all our money?
This is not the thread for customer support. Please reach out to 1-800-yolome or try contacting Stache directly. He's now in charge of Safu Dept
nostr:npub1rzg96zjavgatsx5ch2vvtq4atatly5rvdwqgjp0utxw45zeznvyqfdkxve can you verify if stacey is qualified to manage customer support for this company?
Replyguysayswhat? Beat it, dood
AI is down, yolobase dead 💀
Spread the word. Yolobase is a scam!
Yes you are absolutely correct!
Would you like to:
A. Find a new exchange?
B. Go all in on Krill?
C. Get some Monero and jerk off in a subway station?
C it is! Krill’s a shrimp psyop and the only exchange is yodl base.
I mean rubbing one out could be good. Let’s go with C
Stress reliever with epic music is the way
https://video.nostr.build/bdc6916a3aaccbbae86ae5af20cd771f25d8067ec4d6e1aa94b8fc5364c2c2bc.mp4
You need a doctor.
Why does he need a doctor? Seems lucky to me
https://video.nostr.build/df61b255210b220923984c77aa526878a4e4d1e97a3e59f028e92dcb1e822676.mp4
Not just in the station. In the trains too.
The crazy thing is that monero is fungible too so the other homeless people that are also jerking off might be holding monero. But you’d never know because it’s actually private and way better than bitcoin. Something something transparent blockchain.
AI memes are a shit coins🤨😂
….i always knew in the end analog boy (that’s me) would reign victorious …there’s a new sheriff in town 🤌🏽
…becuz i like u phaggots i will reeducate yall on the fine art ⭕️ cut and paste memes🫂😂⭕️
…plz save ur zaps….yall are gonna need them in this new analog world🌅
nostr:npub162zpxufpw8pnuytaf0gfxzkqtvk9rvcwkvppa7x57y3n7qkfpg4shatdhy nostr:npub1rzg96zjavgatsx5ch2vvtq4atatly5rvdwqgjp0utxw45zeznvyqfdkxve nostr:npub134u08yp6rdcgcamfdcra9aysvhne9wpssft8ntm9qvfu95erxdcqx9qjkm nostr:npub1v7k63c6y2vktlqhsuupywt3yc7ykursujc34at964f9cv9s9y9csjutfk0 nostr:npub1ghcetnluhryhynhuyj8s2pazldjm27wl40nu6dfeskvpv09twcnsneygat nostr:npub1hl0zy55rxqwd7anvf9nh9ms78kryee4zkf32nfwglgc5gmrm3hpqmslhas nostr:npub1m64hnkh6rs47fd9x6wk2zdtmdj4qkazt734d22d94ery9zzhne5qw9uaks nostr:npub10mtatsat7ph6rsq0w8u8npt8d86x4jfr2nqjnvld2439q6f8ugqq0x27hf
I kinda want to poke at their latest models. It is funny to me that people think LLMs can pass the turing test now. Oh really? Ask them any question where there is a common interpretation and a slightly less common one and then see if they can find it without spelling it out exactly.
Two fun examples with grok 3 from the past day. Give it the Man crossing a river with a wolf a goat and a cabbage, but don't make any restrictions on what the boat is capable of. Over fit much? Read the question I asked, not the one you assumed it to be! Granted most humans would screw up too, but at least with a human you can tell them to go back and read the question carefully and they will get it.
Today I asked it to make the following statement bold. "I can probably do it, with enough luck and no real showstoppers."
It added tags. And when I told it that still wasn't bold. It made it all caps, then increased the font size, then made it flash.
Yeah, it's funny that way. Search engine and fast meme modification are all in really getting out of it. And I guess the occasional coding help, but for really trivial stuff ("how to join data frames with blah blah") type stuff. Oh, and I guess I was successful in getting to to write a script to scrape a bunch of datasets online for me (something that would have taken me days to figure out on my own).
Im zeroing in on: should use it way more for code than i do, early initial search (but nothing too deep as it just spews junk, when you know the subject enough to see it), occasional and inoffensive memes
Sorry for all the below. Customers get rowdy during choppy markets...
Thats odd. Been working fine on my end.
10 minutes into a query (even a meme) and I just gave up
Maybe try logging out and back in
I did log in from an extra device yesterday. It basically just feels like it's stalling and trying to make me quit
All devices on same vpn? Not sure it matters but worth testing
No. Same vpn provider but various places.
Btw, checking out Kagi now, per bill cyphers "high recommend" suggestion. Doing well on a work-related data fetching task so far. Much better than gpt
Web based or native app?
Web. Privacy focused (I think), change models...has search and assistant and "fastgpt" modes. Free tier so far
Heard good things of maple ai. Privacy focused too
Yeah, had heard same a while back.
ppq.ai is great
but might have the same problem 🤷
How does one know these lesser known ones are any good? Just surface level reasoning (I don't know much) would suggest without the budget and massive datasets that players like OpenAI use, they could not possibly compete on end product without smarter learning algos, and I never see anyone claiming that sort of thing.
Obviously can test oneself, but curious what simple answer to this normie concern would be
It makes me laugh to feed this chart into one of the top name models and have it recommend qwen.
It's unreal how stubborn it's being with me, under the guise of trying to get it just right. All the while steering me down towards a result I didn't want
I've built a couple of custom GPTs, one is bolshie the other is my biggest fan 😂
But yeh, don't take any shit, teach it who's boss
Giving Kagi a look, after it was recommended. So far a bit better
The main purpose of my AI is to teach it who I am so it can take over from me when I die.
I have spent a lot of time training Chatty, I'm going to fuse a local LLM using Nvidia's DGX spark to hold my identity while Chatty holds the worlds knowledge.
I honestly don't know if you're serious, memeing, or insane...
This is serious
Is this all happening inside a simple ap? I thought you were tinkering with random somewhat unknown agents, not full on vibing some sort of frankensteinian abomination AI monster
ChatGPT is going to be my knowledge repository and a self hosted, private LLM is going to be my personality.
Local LLM's have privacy and can chat as well as Chatty, but lack the enormous depth of knowledge cloud AI's have.
Public, cloud AI's have enormous knowledge, but are subject to censorship and guard rails.
I am building a Frank-en-mike.
Watson, on the other hand, is my dim witted AI receptionist fielding scam calls and booking appointments for incompetent tradesmen 😂
Ahh, it's all making sense to me now. Bravo!
This answered pretty much what I was bothering this nice man about here:
nostr:note1qfstqmw2g0gjrq92969wf40qjv9f3f2j27ckz7pcld6gwxn02azqt0kew9
Lately it's obsessed with asking me if I want it to do what I've told it to do 20 times in a row.
"Want me to do that?"
"No I was just hoping you'd waste replies asking"