Artificial Intelligence
- Started
- Last post
- 2,368 Responses
- grafician0
^
- the CoT feature is really something, a lot of people will think this is true AI alreadygrafician
- The reasoning is more interesting that the actual answer TBHSalarrue
- It's like a full page of text even for simple questions like "Why 2+2=4?"
But very interesting feature - reads like the model narrates its thinkinggrafician - try it, the model is free and VERY capablegrafician
- OKyuekit
- grafician1
Spline3D launching AI 3D generation model Spell
https://spell.spline.design/
https://spell.spline.design/expl…- Meshy AI has been making good progress too: https://www.meshy.ai…NBQ00
- The problem with Spline3D for now is the "painterly" gaussian-like results, it's not very usable other than just for conceptual stuff.NBQ00
- webazoot0
"A harmonious blend of AI selected grape varietals, curated by cutting-edge AI algorithms." Erm.
- yuekit3
OpenAI demos "Operator"
- !Nairn
- It's impressive but I feel like the hype has greatly exceeded the reality.yuekit
- but this is where they're going...Nairn
- $200 a month...jonny_quest_lives
- https://openai.com/i…grafician
- and a good review here
https://every.to/cha…grafician - Hey man, I got five kids to feed!utopian
- kool.. but i guess they couldnt find any women in their companny to have a seat at the table...neverscared
- I played a bit with Anthropic's Computer Use, which is the same thing. I would say it was not ready for prime. Maybe this version is an improvementzaq
- grafician0
- They are getting scared.
Deepseek cooked with their new R1 model!grafician - ^ it's free here https://chat.deepsee…grafician
- you mean to tell me overcharging for low lying fruit isn't a good business model? who woud have thunk?jonny_quest_lives
- They are getting scared.
- yuekit4
How Chinese A.I. Start-Up DeepSeek Is Competing With Silicon Valley Giants
- oops
https://x.com/kimmon…YakuZoku - Crazy shit, they built their entire platform for under $20 million and stole all the US data to build it.utopian
- Literally killing it !_me_
- oops
- yuekit1
AI identity crisis...
- grafician0
https://github.com/synth-inc/oni…
"Onit is an open-source AI chat assistant that lives in your desktop!
It's like ChatGPT Desktop, but with local mode and support for other model providers (Anthropic, GoogleAI, xAI, etc). It's also like Cursor Chat, but everywhere on your computer - not just in your IDE!"
- How to know the RAM requirements...Salarrue
- Google recommends at least 8GB of RAM
https://www.reddit.c…utopian
- jonny_quest_lives2
So to recap the week... OpenAi and a a bunch of other western Ai companies have been selling their AI secret sauce for years to investors as a proprietary mystery formula that only they could understand... They also claimed they needed infinite compute to achieve the next horizon of either AGI/ASI driving up the valuations of their companies as well as Nvidia.
Now a group of young Chinese dudes who used to mine crypto just wrote a really good algorithm that performs equal to or better than OpenAi’s $200 a month offerings made it open source and claim it was trained for WAAAY less is threatening to burn down the whole Ai Silicon Valley venture capital grift?
Interesting week.
- Chan GPT_niko
- Yes. Exactly this.grafician
- Highly doubt this. This story, like the open source release, is more likely intended to dry up private investment in AI in America.monNom
- It saps revenue from the companies selling access to private models, drys up interest from investors. Facebook was following the same strategy with Llama releasmonNom
- LOL the entire market is DOWN because of DeepSeek
Even tho' it was known for weeks
Odd Mondaygrafician - Chinese dudes stole all of the data from US and western countries, go China!utopian
- Probably a buying opportunity for NVDA. Not so much for anyone trying to sell chat bots.monNom
- I mean.. a neural network is a fundamentally different technology to a simple algorithm but I appreciate the sentiment.kalkal
- @grafician known and known are two different things in the world of AI research, largely in part to too many projects optimising for the benchmarkskalkal
- only when the benchmarks are released can we find out whether it's all the real deal or not.kalkal
- *benchmarks based on actual user testing and scrutiny, not the creators benchmarks.kalkal
- they aren't just "Chinese dudes who used to mine crypto"sted
- If it’s open-source, benchmarking it should be straightforward. My take is if it is competitive with state of the art, unlikely it was not state supported.monNom
- https://arxiv.org/ab…
look at the names on the paper, looks more like an entire class than just a bunch of kids.sted - Or if it truly is 1/10th the resources to run, unlikely to be state of the art. Rare that some crypto-bros come in and produce an order of magnitude step changemonNom
- their first product from a year ago:
https://deepseekcode…sted - And because their huggingface pretty much everybody knew that they are working on this.sted
- "young Chinese dudes" is funnier than Quant Hedge fund manager hiring dozens of top Chinese PhD's and buying 10,000 H800 chips to topple the US Ai industry...jonny_quest_lives
- i tend to sarcastically shitpost through tumultuous times... i didn't mean to offend any of the PhD's who co wrote their open source documentation:jonny_quest_lives
- https://arxiv.org/ab…jonny_quest_lives
- https://i.imgur.com/…jonny_quest_lives
- NBQ000
Holeee phuk.
- NBQ00-3
Is this the AI bubble popping?
Markets red in pre-market trading (NVIDIA -5%'ish)
- YakuZoku4
- USA #2
USA #2
USA #2
USA #2utopian - It's ok Trump will add a 25% tariff on Chinese AI and all will be well._niko
- lolYakuZoku
- the excess of "USA #1 in AI" got wiped, nothing interesting. nvidia lolsted
- but trump said golden age...not silver or bronze age... but looking at trumps fat face..its def. bronzer age...neverscared
- AI is simply the newest cash grab in an overhyped market by greedy investors.utopian
- USA #2
- i_monk1
Deepseek's privacy policy page is a 404: https://platform.deepseek.com/do…
- utopian5
I'm considering closing my $20 ChatGPT account and moving over to Abacus AI for $10 a month. Abacus AI offer so many models under one roof.
Abacus Models Included:
LLMs
- GPT-4o
- GPT-o1
- Deepseek R1
• Claude Sonnet 3.5
• Claude Haiku 3.5
• Gemini 2.0 Flash
• Gemini 2.0 Flash Thinking
• Gemini 1.5 Pro
- Abacus AI Dracarys
- Abacus AI Smaug
- Llama 3.3 70B
- Llama 3.1 405B
- GrokImage Generation
- FLUX 1.1 Pro
- FLUX 1.1 Ultra
- Dall-E
• Recraft
• IdeogramVideo Generation
- Kling Al v1.6
- Luma Labs
- Runway
• Hailuo AI- me too, but why not just use Deepseek for free?_niko
- Abacus has more stuff and image/ video generation stuffNBQ00
- Because look at all the various models that are on this platform. You get ChatGPT, Google, Meta, etc..and all of the image and video generators as well.utopian
- this one? https://abacus.ai/grafician
- "$10 per user, billed monthly (first month free, minimum 2 month subscription Learn more)"grafician
- https://abacus.ai/ch…grafician
- oh shit, glanced over that, so you get the latest chatGPT as well plus all the other stuff so worse case even if you only use GPT you save 50% each month_niko
- There is a token limit though. It's great if you are a moderate userShenanigansTV
- That's the URL and they're a U.S. company that doesn't share data with China. They also offer the Deepseek LLM within their API environment.utopian
- So both OpenAI and Anthropic are allowing this company to cannibalize their margins on already money losing models?monNom
- I’d be highly suspicious that you weren’t actually getting GPT-o1 but rather some smaller/cheaper model fine tuned to say it was GPT-o1.monNom
- Also, all of the AI data can be integrated and indexed into: Google Drive, Slack, Microsoft Teams, Salesforce, etc.. You also can build AI agents & chatbots.utopian
- According to information available on Abacus AI's website, their $10 per month plan does not have a strict "token limit".utopian
- I'm probably paying around $500 a month easy for AI shit for work, probably closer to $700YakuZoku
- YakuZoku5
US AI companies right now
- nvidiasted
- nah they knew what they had, the retards are the investor bros taking these stocks to the moon,_niko
- i think 1) they knew what they had 2) they knew the real costs 3) everything else was investor fugazijonny_quest_lives
- i just don't think they saw an overseas competitor revealing what was behind the curtainjonny_quest_lives
- puts the whole Sam Altman board of directors of OpenAi drama in a a whole new light...jonny_quest_lives
- for profit vs. open source... real costs vs. speculative hypejonny_quest_lives
- also all the dank tech memes really cheer me upjonny_quest_lives
- I guess the US based AI companies did think that the Chinese were going to steal of the data and models and then recreate all of them for pennies on the dollar.utopian
- I mean one could argue they are more transparent than the Sam Altman led investment vehicle called "OpenAi"jonny_quest_lives
- Deepseek is either being truthful in their documentation and OpenAi has been lying or...jonny_quest_lives
- Deepseek is is being less than truthful and OpenAi has been honest with their cost structurejonny_quest_lives
- i am here for the lulz either wayjonny_quest_lives
- We've had to sit through 2 years of Sam Altman talking about rewriting "the social contract" for the sole benefit of his company...jonny_quest_lives
- this couldn't happen to a better bunch of silicon valley lizard peoplejonny_quest_lives
- https://i.imgur.com/…jonny_quest_lives
- OpenAi employees crying for "protectionism" is biblical schadenfreude: https://i.imgur.com/…jonny_quest_lives
- Biblical: https://i.imgur.com/…jonny_quest_lives
- https://i.imgur.com/…jonny_quest_lives
- utopian0
DeepSeek built its product using roughly 50,000 Nvidia H100 chips it can't mention because it would violate U.S. export controls that ban the sale of such chips to Chinese companies.
- Source: "Alexander Wang of Scale AI"jonny_quest_lives
- who to trust? Western Ai Oligarchs attempting to rewrite society or test Deepseek's documentation independentlyjonny_quest_lives
- https://c.tenor.com/…jonny_quest_lives
- I doubt this claim that was everywhere do discredit these guys.
They released the papers, everything points that they did it on the cheap as mentionedgrafician - < so this is probably coping liesgrafician
- The model is really cheap TO RUN ALSO so yeahgrafician
- jonny_quest_lives0
"DeepSeek claimed it used just 2,048 Nvidia H800s and $5.6mn to train a model with 671bn parameters, a fraction of what OpenAI and Google spent to train comparably sized models.
Ritwik Gupta, AI policy researcher at the University of California, Berkeley, said DeepSeek’s recent model releases demonstrate that “there is no moat when it comes to AI capabilities”.
“The first person to train models has to expend lots of resources to get there,” he said. “But the second mover can get there cheaper and more quickly.”
Gupta added that China had a much larger talent pool of systems engineers than the US who understand how to get the best use of computing resources to train and run models more cheaply.
Industry insiders say that even though DeepSeek has shown impressive results with limited resources, it remains an open question whether it can continue to be competitive as the industry evolves.
Returns at High-Flyer, its big backer, lagged behind in 2024, which one person close to Liang blamed on the founder’s attention being mostly focused on DeepSeek.
"
- neverscared0
The Less People Know About AI, the More They Like It
You might assume that tech-savvy people are the most open to using AI, but research suggests it's actually those who are least familiar with it.
