+- Hack Forums (https://hackforums.net)
+-- Forum: General Topics (https://hackforums.net/forumdisplay.php?fid=7)
+--- Forum: World (https://hackforums.net/forumdisplay.php?fid=445)
+---- Forum: News and Happenings (https://hackforums.net/forumdisplay.php?fid=89)
+---- Thread: New AI Surpasses Both ChatGPT & Deepseek v3 (/showthread.php?tid=6291078)
New AI Surpasses Both ChatGPT & Deepseek v3 - Foxxy Proxies Ltd - Jan 31, 2025
https://allenai.org/blog/tulu-3-405B
Just received an email regarding this AI outperforming both chatgpt and deepseek v3.
RE: New AI Surpasses Both ChatGPT & Deepseek v3 - Foxxy Proxies Ltd - Jan 31, 2025
(Jan 31, 2025 05:04 PM)Lowkey Wrote: I will read this article and study more about it but ain't no way something called Tülu 3 is outperforming chatgpt. (I'm biased)
Its not impossible. Deepseek already surpassed chatgpt. And as a long time chatgpt user, I feel your pain. :)
RE: New AI Surpasses Both ChatGPT & Deepseek v3 - The Pact - Jan 31, 2025
(Jan 31, 2025 05:04 PM)Lowkey Wrote: I will read this article and study more about it but ain't no way something called Tülu 3 is outperforming chatgpt. (I'm biased)
I wonder why people are such followers. People have acted like ChatGPT is the only LLM, and are prepared to sit and wait during over usage, when there is a ton of better stuff out there attached to search engines so they don't die after 2023, and the Deep Seek did some Temu style marketing and was suddenly the new next best thing when my couple of times using it it is crap and there's much better out there. People are biased just because they've become fond of ChatGPT from always using it. But I've been using Ai that has search and source material past 2023 attached and that can do more, so I'm not attached. I guess I'm glad that other people are dealing with wait times and leaving my better Ai tools alone.
RE: New AI Surpasses Both ChatGPT & Deepseek v3 - Foxxy Proxies Ltd - Jan 31, 2025
(Jan 31, 2025 08:45 PM)The Pact Wrote: I wonder why people are such followers. People have acted like ChatGPT is the only LLM, and are prepared to sit and wait during over usage, when there is a ton of better stuff out there attached to search engines so they don't die after 2023, and the Deep Seek did some Temu style marketing and was suddenly the new next best thing when my couple of times using it it is crap and there's much better out there. People are biased just because they've become fond of ChatGPT from always using it. But I've been using Ai that has search and source material past 2023 attached and that can do more, so I'm not attached. I guess I'm glad that other people are dealing with wait times and leaving my better Ai tools alone.
I use a wide variety of them myself. I've used a ton over at theresanaiforthat and what can be found over at huggingface.
RE: New AI Surpasses Both ChatGPT & Deepseek v3 - The Pact - Jan 31, 2025
(Jan 31, 2025 05:04 PM)Lowkey Wrote: I will read this article and study more about it but ain't no way something called Tülu 3 is outperforming chatgpt. (I'm biased)
That wasn't really directed at you by the way although why be biased proChatGPT when there is so much else out there. Or perhaps you just think that Tulu sounds like an eleven year old girl with annoying parents, Idk.
RE: New AI Surpasses Both ChatGPT & Deepseek v3 - Bastille - Feb 1, 2025
Can we rest for like a week or two ? 😭 the AI race is crazy
I thought about studying a little bit about AI field but it's moving so fast I don't know if it's worth getting involved now or to wait
RE: New AI Surpasses Both ChatGPT & Deepseek v3 - Virtue - Feb 1, 2025
(Feb 1, 2025 01:05 AM)Bastille Wrote: Can we rest for like a week or two ? 😭 the AI race is crazy
I thought about studying a little bit about AI field but it's moving so fast I don't know if it's worth getting involved now or to wait
Yes, you should check it out. Do you have any powerful graphics cards? Do you have an m1 mac or other high bandwidth ram + cpu setup?
You can install ollama easily on mac windows or linux, run any of the llama models in one terminal command.
There is python bindings to interact with the ollama server, so if you have enough compute you can start a few agents simultaneously and put them together where each one has a responsibility as part of a prompt for example, or for fun you can assign them a task and they have to work together.
If you are interested in learning the basics of LLMs, you should go nowhere else than Andrej Karpathy. He is a foremost AI engineer who was part of OpenAI.
He made the best lectures and released them for free, I didn't understand it until I watched these a couple of times:
https://www.youtube.com/watch?v=VMj-3S1tku0
--
Virtue: Be nice, don't ever steal, do good work for other people and they'll pay you!😎