If I remember correctly, a lot of these tests have a bullshit writing segment where they hire boomer women to read your 5 paragraph essays and score it subjectively. If anything this just proves those tests are garbage.
you're a retard, stop flooding the catalog. there's so many of these stupid gpt4 threads. why do you need a new thread? holy fuck your stupid BOT hype gays are dumb
>It's a bunch of 60 IQ retards making 4chan-quality spam posts about the world ending. They should all be banned.
They’d be banned if the mods were awake. These doomer threads are still up
it parses instructions better and writes better dialog
however, its reaaaally fucking slow, its obvious they're maxing out their resources right now trying to offer this much computation via a cloud service
It went from failing grades and 50th percentile to As and Bs and 90th percentile in every category. What are you talking about? This is like the difference between a black kid's and an Asian kid's college application.
>I guess to a highschooler a 4 on calc bc is impressive. GUESS BOT IS FULL OF HIGHSCHOOLERS LMAO
But seriously, there are better AIs that focus on math and programming. gpt-4 is a chatbot it's supposed to be good at language, but it's now so good that it's showing us that all our tests are just poor language games.
This is the alarm math educators have been sounding for the past hundred years. Creativity vs language games. Math needs the former, but we only test for the latter.
the version with th 50 page token limit seems interesting tbh.
you'd be able to throw full scientific papers at the thing. how well it would do is a different question altogether, but still.
that price is fine, it's the EXACT same price per token that GPT-3 davinci was for more than two and half years after release, from 2020 to near the end of 2022
(and there's been a lot of inflation since then so it's cheaper if inflation-adjusted)
no one here knows this though since you're bunch of newgays to AI, so you think that price is unprecedentedly high
it's actually cheaper than release gpt3 even if not inflation adjusted
since they didn't give you a discount on prompt tokens like this, it was just 6 cents per token for all tokens
it's actually cheaper than release gpt3 even if not inflation adjusted
since they didn't give you a discount on prompt tokens like this, it was just 6 cents per token for all tokens
You're right, but it does not change the relationship between the two. It's still ~15-60x more, but nowhere near that much better.
<<92100938 <<92100980
It's poor because it's not a logical follow up from the content it produced before it. It's an analogy that supposedly flows logically from the previous, but it doesn't. It's a punchline to a nonexistant setup. I'm not talking about the actual contents.
3 was amazing so if 4 is just a marginal upgrade we've got Chrome version 113.0.021.4 and Firefox version 104.0.434.04 retarded naming schemes for updates. Should have called it 3.1 at best.
Can't believe it doesn't do that good on AP Lang and AP Lit despite it being a language model.
Also it's Codeforces ranking lmao. I thought it was supposed to be good at coding
If I remember correctly, a lot of these tests have a bullshit writing segment where they hire boomer women to read your 5 paragraph essays and score it subjectively. If anything this just proves those tests are garbage.
Or the answers to the exam aren’t plainly stated in the training set.
But in general not surprising. The bigger a block of text a language model needs to generate the more incoherent it tends to become.
>yet another AI thread
There's like 50 of them now, STOP FUCKING FLOODING YOU moron
just close BOT for the day bro it's today's news.
you're a retard, stop flooding the catalog. there's so many of these stupid gpt4 threads. why do you need a new thread? holy fuck your stupid BOT hype gays are dumb
It's a bunch of 60 IQ retards making 4chan-quality spam posts about the world ending. They should all be banned.
>It's a bunch of 60 IQ retards making 4chan-quality spam posts about the world ending. They should all be banned.
They’d be banned if the mods were awake. These doomer threads are still up
>I am overhyped, I'm not much of a better coder than I was last year
it parses instructions better and writes better dialog
however, its reaaaally fucking slow, its obvious they're maxing out their resources right now trying to offer this much computation via a cloud service
I'm too stingy to pay the 20 bucks. Get me some funny steamed hams re-writes.
IT WAS THE BEST OF TIMES IT WAS THE BLURST OF TIMES
shrug, i've been subscribed for a while now
its a good tool for devs and i mean, $20 aint much for what they offer
>good tool for devs
this is a paid post everybody, everyone look at this paid post so you can identify more like them in the future.
>it's like shitsunami
3 lines in and it's already revealed how poor thw riting is.
That's almost completely in character for the show it's emulating.
Comparable to phase 5 mcu
>the AI designed on a gigantic corpus of english language text failed the AP English Language and Composition exam
>the only AP exams it doesn't pass
English the hardest subject
Writerchads dabbing
It went from failing grades and 50th percentile to As and Bs and 90th percentile in every category. What are you talking about? This is like the difference between a black kid's and an Asian kid's college application.
>I guess to a highschooler a 4 on calc bc is impressive. GUESS BOT IS FULL OF HIGHSCHOOLERS LMAO
But seriously, there are better AIs that focus on math and programming. gpt-4 is a chatbot it's supposed to be good at language, but it's now so good that it's showing us that all our tests are just poor language games.
This is the alarm math educators have been sounding for the past hundred years. Creativity vs language games. Math needs the former, but we only test for the latter.
it is better in nearly every category.
explain how that is "not different"
So wait for China's or Microsoft's AI's to be leaked or released publicly
the version with th 50 page token limit seems interesting tbh.
you'd be able to throw full scientific papers at the thing. how well it would do is a different question altogether, but still.
Most of the time spent for GPT-4 was asking diversity hires what made them upset.
https://openai.com/pricing
GPT-4 costs anywhere from 15-60x more compared to GTP-3.5. So exponential processing requirements for incremental gains.
that price is fine, it's the EXACT same price per token that GPT-3 davinci was for more than two and half years after release, from 2020 to near the end of 2022
(and there's been a lot of inflation since then so it's cheaper if inflation-adjusted)
no one here knows this though since you're bunch of newgays to AI, so you think that price is unprecedentedly high
it's actually cheaper than release gpt3 even if not inflation adjusted
since they didn't give you a discount on prompt tokens like this, it was just 6 cents per token for all tokens
ugh 6 cents per 1k tokens I meant
You're right, but it does not change the relationship between the two. It's still ~15-60x more, but nowhere near that much better.
yeah and no one used davinci until it's priced was slashed for a reason
<<92100938 <<92100980
It's poor because it's not a logical follow up from the content it produced before it. It's an analogy that supposedly flows logically from the previous, but it doesn't. It's a punchline to a nonexistant setup. I'm not talking about the actual contents.
3 was amazing so if 4 is just a marginal upgrade we've got Chrome version 113.0.021.4 and Firefox version 104.0.434.04 retarded naming schemes for updates. Should have called it 3.1 at best.
I was excited about the video feature but considering that's not available yet and it probably won't let me generate porn, I'm now underwhelmed
daily reminder