Anonymous at Tue, 5 Mar 2024 23:46:20 UTC No. 16058987

GET FUCKED NUMBERCRUNCHERS

Anonymous at Tue, 5 Mar 2024 23:51:31 UTC No. 16058991

https://twitter.com/alexalbert__/status/1764722513014329620

Anonymous at Tue, 5 Mar 2024 23:59:22 UTC No. 16059000

Clown 3

Anonymous at Wed, 6 Mar 2024 00:20:21 UTC No. 16059020

>>16058986
>most of yuzu (a project that has been around for years) was written by Claude 3 (an AI that has been out for a few days)
huh?

Anonymous at Wed, 6 Mar 2024 00:33:27 UTC No. 16059033

Good, can't wait to get unemployment checks

Anonymous at Wed, 6 Mar 2024 01:11:14 UTC No. 16059068

I just tried to get it to write my graduate complex analysis homework and it failed, so I guess it can only replace people much less intelligent than mathematicians like computer science majors and physicists.

Anonymous at Wed, 6 Mar 2024 02:07:53 UTC No. 16059127

>>16058986
This will only get more painful the more time passes. It wont only be for physicists or mathematicians or artists or writers or anyone who uses intelligence as a marker for value. The market will move to unique skillsets rather than intelligence. And even that wont last because once AI is smart enough and mobile enough, it will generate its own unique skill sets that humans cannot match.

Anonymous at Wed, 6 Mar 2024 02:27:53 UTC No. 16059141

>>16059068
What textbook do you use for your complex analysis course?

Anonymous at Wed, 6 Mar 2024 02:29:55 UTC No. 16059144

>>16058986
>Claude was able to write FPGA code that compiled first time and ran Tears of the Kingdom
absolute bullshit

Anonymous at Wed, 6 Mar 2024 02:33:19 UTC No. 16059149

>>16059144
Yeah but you can generate almost unbounded hype through only fooling the bottom fraction of humans

Anonymous at Wed, 6 Mar 2024 02:33:24 UTC No. 16059150

>>16059020
This. What fucking psyops is this?

Anonymous at Wed, 6 Mar 2024 02:37:11 UTC No. 16059155

>>16058986
the verge tweet in the image doesn't exist

Anonymous at Wed, 6 Mar 2024 02:39:29 UTC No. 16059156

From poe.com

Essentially, AI models cannot form original research ideas, this part of research is what humans can perform but AI models cannot. Is this assumption correct?

"Yes, your assumption is correct. As of now, AI models like me are not capable of forming original research ideas autonomously. Generating truly novel and creative research ideas is a complex cognitive process that involves a deep understanding of the domain, creativity, critical thinking, and the ability to make connections between seemingly unrelated concepts. These aspects of research are still primarily performed by human researchers."

"While AI models can assist in research by analyzing data, identifying patterns, suggesting potential directions, and helping with computational aspects, they rely on pre-existing knowledge and patterns in the data they were trained on. They do not have the ability to generate completely original ideas or hypotheses without human input. researchers play a vital role in the scientific community by formulating research questions, designing experiments, developing theories, and pushing the boundaries of knowledge. AI models can support and enhance these efforts, but they are not currently capable of replacing the creative and innovative thinking that humans bring to the research process."

You're fine as long as a huge portion of your job has you deal with people in person or if your job requires original thought.

Anonymous at Wed, 6 Mar 2024 02:45:09 UTC No. 16059158

>>16058986
How much is Anthropic paying you to advertise? How good can their AI be if they need to bullshit you and lie about its capabilities and history (e.g., claiming that an AI that has only been available for days and in development for ~2ish years created Yuzu, a switch emulator that has been available since 2018) in advertising?

Anonymous at Wed, 6 Mar 2024 16:31:07 UTC No. 16060032

>>16059158
I've actually tried out Claude and it's better than OpenAI's ChatGPT... well at least the free versions. I'm not paying for that shit lmao.
The way I'm measuring it how well it can do my maths homework... both suck but ChatGPT is basically wrong 100% of the time.

Anonymous at Wed, 6 Mar 2024 18:52:01 UTC No. 16060193

>How certain are you, that you're really outside the box right now

Cult of Passion at Wed, 6 Mar 2024 19:15:19 UTC No. 16060225

>>16060193
https://youtu.be/OXSkfTedVb0

Deep [burp].

Anonymous at Wed, 6 Mar 2024 19:24:33 UTC No. 16060231

>>16058986
Claude 3 Opus cure the cancer.

Anonymous at Wed, 6 Mar 2024 19:35:45 UTC No. 16060237

stringsisters..

Cult of Passion at Wed, 6 Mar 2024 19:57:25 UTC No. 16060267

>>16060237
Ugh...his entire twitter feed is surface level Israel/US political stuff. Lame.

Anonymous at Wed, 6 Mar 2024 20:43:03 UTC No. 16060332

I'll humor you. I just paid 20 dollars for this new claude AI just to easily prove you wrong.

I asked the AI to solve a simple geometric proof (geometric mean theorem). Not only did it only approach it in an extremely convoluted method, but it also failed at the most basic of algebra, as I've make sure to emphasize in the picture attached.

>"Just did this quantum algorithm1!!!"
I don't know or care who Guillaume Verdon is, but he is clearly an idiot brainlet with worthless credentials.

>"Claude is one of the only people(lol...) ever to have understood the final paper of my quantum physics PhD!"
So his PhD had less rigor than secondary school proof based geometry. Not surprising as most "physics" majors are simply physics popularizers that get destroyed by the most basic mathematical applications that aren't simply plug and chug.

>"Claude 3 gets ~50% accuracy on GPQA!"

He's quoting marketing material from a company eager to break in and compete in this market.

I have yet to see any AI actually capable of doing real, basic mathematics, much less anything beyond.

Anonymous at Wed, 6 Mar 2024 22:34:44 UTC No. 16060460

>>16060332
>I'll humor you. I just paid 20 dollars
AHAHAHAHAHAH
MY JOB IS DONE
I WIN YOU LOSE

Anonymous at Wed, 6 Mar 2024 23:36:56 UTC No. 16060545

Claude is where GPT was on first release. Its really bad.

Anonymous at Wed, 6 Mar 2024 23:39:26 UTC No. 16060551

>>16060231
It's over AI fucking sucks

Anonymous at Thu, 7 Mar 2024 00:00:06 UTC No. 16060573

>>16059020
>>16059150
Niggers really can't read
>closed beta version

Anonymous at Thu, 7 Mar 2024 03:52:35 UTC No. 16060827

>>16058986
The GPQA (who’s even heard of this?) was written by people who are also employees at anthropic, so I don’t trust this result.

Anonymous at Thu, 7 Mar 2024 04:05:20 UTC No. 16060833

>>16060573
Yuzu has been around since early 2018. The original Claude release (so 2 full generations ago) was March of 2023 after slightly over a year of development (at least if Anthropics own development blog timeline is to be believed).

Tell me how an LLM that wasn't even in develop until (at the earliest) early 2021 was involved in the development of Yuzu in 2018?

Anonymous at Thu, 7 Mar 2024 04:15:41 UTC No. 16060844

>>16058986
>7th year PhD student