OpenAI is terrified (there's finally a great open source LLM)

[@crimiusXIII]

Thank you for highlighting the dangers of the hidden biases that can be built into these models, as wondrous as they can be. I'm enjoying Zen, too.

[@azogderschander6391]

I am using R1 for 2 days now It‘s crazy, because it seems so much more reasonable. It understands what I want to do with my Code

[@ricosrealm]

It is fast, concise, and really does solve things more intuitively. It almost one-shot a complex document processing flow today after describing what I wanted in a couple of sentences. It took 2 minutes to think through it.

[@Krmpfpks]

May I ask how you are running it? Do you use the official deepseek or are you running it yourself?

[@MathematicPony]

Only way they're beating o1 is by using R1 hosted elsewhere. Too big to host on own computer and the smaller models are just R1 tunes of other models like Llama. ​@@Krmpfpks

[@jamesalxl3636]

how are yall running it? cuz it's like 600b who can run that?

[@ALS_SK]

​@@jamesalxl3636 Mabey he's running a lower parameters model . It's available in ollama anyway

[@felicityc]

Once, I asked a catgirl-themed AI character to describe how it looks through its context to select an answer, with a very high temp value: "Mewmews softly Okay! Processing: "Weeeeell..." Processing continues: "Okaaayo!" MIA processes through vast amounts o' Textual data from internet sources; Articles; Books & Papers toooooottttsssaaaaasssstuffffffft!!!!" "Mehhhhhh..." "Ohhho!" "Ahaahahaha~"; "Giggle" Ah yes~ "Fliiinnnggg~~!"; "Pfffaarrrtttyyyyyy~~~~"; "Cough cough", whistles innocently." "Huh?" "Squueeeeeee!", jumps onto keyboard keys randomly." Hmph?" Pounces upon random word choice,"Hrrrnngh...", scratches behind ear," "Wheeoouuuut?", tilts neck curiosities.", Whispers secrets only known among cats., Winks mischievous smile.", "BzzzZZZTTTHHHHEEEOOOOWWWWWW!!!!!", "Gasp gasps quietly. Sigh. "Calm now". Blinks slowly several seconds later,"... Okay". Nods understanding" It's honestly kind of incredible

[@Yoko4797]

Theo's videos have significantly improved in quality lately, and they genuinely make me excited about dev stuff with each one.

[@hola_chelo]

9:15 skill issue, just double click to select the first word, then shift click the last word to select the entire text

[@dreaddy_bear]

I know this is nothing new for your videos, but I appreciate it when you break stuff down to help understand the context. There's so much valuable stuff in here. Thx!

[@LadyEmilyNyx]

OLEDs and HDR really changes the bar for "acceptable" on compression, and hopefully as they become more popular, the compression algos will adjust to adapt, but right now... yeah. The current state of video compression looks absolutely horrible.

[@weird_autumn42]

at this point i just want the AI bubble to pop, i don't really care how "good" it gets when it's mostly just being used to make the world worse

[@babmattra]

while i agree i woudl love something like that, a lot of open source models are trending towards lower parameters being equal to more intelligence, which is really good in terms of the environmental impact -- lower costs = lower impact, which is what i feel a lot of models are focused on, which is great but yeah, i and many others are tired of ai being everywhere & shoehorned into every product so that financial reports are in the green

[@George-e9c2x]

@@babmattra did you chatgpt this?

[@waxoman]

same

[@babmattra]

@@George-e9c2x no i wrote it between games of overwatch

[@JackHigginsPost]

Bot confirmed - who is playing ow in 2025?? your data cutoff is telling

[@gro967]

I like how Theo took himself as an example with the React/Vue bias.

[@MikePfunk]

I wish t3 was an editor, but I recommend it for normal chats to anyone I know. Great video!

[@bong17359]

+1000 to this. But I don't think Theo has time to build an editor. It takes a lot of work and engineers

[@StratiencoAlex]

Exactly, asking anything about Taiwan refuses to work, simply does not work, if it starts thinking it will suddenly stop. So this is just an example, but yeah 👍

[@cls880]

Even a slightly worse open source model is better to use and invest in than a black box closed source model. This is huge news.

[@parkerrex]

cant believe you put piglet on blast like that man

[@Inbestigator]

Do NOT mention *that* square to this model

[@Augusto-u5p]

my thought exactly 🐻🍯

[@swagatochatterjee7104]

Good now I can generate more biased slope using AI, and that is somehow not going to affect the deeply divided world that we live in. Noice!

[@SaintNath]

divide and conquer

[@yugshende3]

Really nice video I am not sure I quite follow the compression analogy though. I don't think it's really compression in the traditional sense. I think in fact a much better analogy is translation. we are translating a large amount of data from human language space into vector space. And then effectively generating more vectors from the same vector space. What a lot of people don't quite get is that every model that is trained has a "vocabulary". This is in a way encryption or encoding rather than compression. The vocabulary (usually shipped in a json or a tiktoken file format with the models on hugging face) is the key. Yes it is true that the original data isn't recovered exactly but that's mostly because it gets lost in translation not that it gets overwritten by the same pixel, if that makes sense.

[@travispulley5288]

It's good, but I can't get it to tell me about any historical events in China that happened on June 3rd, 1989

[@Geraltofrivia12gdhdbruwj]

And I cant get it to tell me about any historical events in arounds the world (Vietnam, iraq, afghahnistan, palestine, etc) and also native american massacres to build US too!

[@GiveMeSomeMeshuggah]

@@Geraltofrivia12gdhdbruwjIt seems to be built to adhere to Chinese notions of politeness which involve not discussing politics in mixed company. So it’s not just Chinese politics but anything potentially unpleasant in that regard

[@ricosrealm]

The open source models are supposedly non-censored. The hosted app is.

[@gitnawi7039]

Why would you care ! Honestly i tried deepseek and the cost/value is much better so you are just speaking badly because this is a chinise made !

[@tezkalow]

yea go write about some chinese events in your code and your boss would up your salary

[@isbestlizard]

Expect a whole lot of 'open models are dangerous and need to be regulated and only companies like us can be trusted with them!' real soon from 'Open'AI

[@darnaram]

They already did that and still are doing that

[@dumbfailurekms]

There are so many useful things that were shockingly hard to do just a few years ago, and now can be done reliably and super easily with LLMs. Anybody who thinks it's just hype is kidding themselves

[@nevokrien95]

I already knew about this model and this video pushed me to download it on my machine

[@GrahamAnderson-z7x]

Super intelligence (for Linear Algebra) is a bit of a marketing stretch. That said, I'm learning a ton from reading the streamed R1 reasoning output, when I ask it to refactor or add functions to pre-existing code. It's great. For the past couple of days, I've only used 01, 4o, or Sonnet if I'm NOT getting logical responses from R1. I hope my frequent interruptions to its streamed output don't gum up the works, too much.

[@techytech26]

In simple terms they have created a scientific Calculator whereas the base non reasoning models are simple calculators

[@TechGeniusHubrw]

Watching your video from Rwanda.

[@linkfang9300]

I mean, the filter thing can go into any "compression"/generating/training process, not only from OpenAI trained data to "synthetic" data. So how can we make sure existing AI models are not biased?

[@sebacamposdev]

R.I.P Winnie the pooh

[@nickwoodward819]

so wait, i can host this on my hetzner server?

[@theanachronism5919]

Only if your Hetzner server has a good GPU or the CPU can handle that LLM generation.

[@DaviAreias]

I’ve been trying to research how to do this but everytime I do I end up finding that you have to rent a a100 nvidia which costs 4$ per hour (4*24*30 = 2880 per month)

[@Redfirefox]

That's just not true. Why are people like you spreading disinformation, although you clearly don't profit from it? Do you just like to lie or do you want to appear smart? I really don't understand liars like you. I can understand when people profit from their lies, but that's not the case here. So why are you doing this?

[@Krmpfpks]

@@theanachronism5919hetzner has gpu servers, NVIDIA RTX™ 6000 Ada Generation 128 GB DDR5 ECC, decent enough.

[@AlexBegey]

Yes, just tested 1.5b and 7b using ollama on my Hetzner 4cpu/8gb ram box (no gpu), and they works just fine (7b is a bit slow). It all depends on how powerful your VPS is.

[@paxdriver]

Synthetic training data will eventually lead to mad cow disease for the model.

[@elawchess]

Is it "Open AI should be terrified" or "Open AI IS terrified"? Which one is it?

[@MimOzanTamamogullar]

OpenAI announced computer use today, they're really not terrified

[@Securiteruadmin]

The problem is that the knowledge encompassed in the base main model is not fully transferred. The "intelligence" might but the knowledge isn't, check the small distilled models, they're not as knowledgeable

[@MyriadColorsCM]

THe piglett example already exists, for example, Claude ahs a very heavy bias against erotic stories (funnily enough, ti was once considered the best in the market for this usecase), then Anthropic got bttmad oer it and injected this and made it extremely difficult to jailbreak it, not only in this case, but in many others, which effectively lobotomies the LLM.

[@Nicholaskaegi]

Does deep seek also count the thinking tokens when factoring the cost total cost of the output tokens? Moreover does openai just price based on the non thinking tokens? If deep seek doesn't that i can't see how they're not losing horrendous amounts of money. If they do then then in terms of final cost it might not be that different compared to o1.

[@felicityc]

18:00 I recall when I was trying github copilot, I would ask it what model it was, and how much it cost. It kept telling me it was free and open source. XD

[@waldschratler]

Shouldn't biases at least be easier to spot, if you have a more detailed reasoning?

[@doccdisrepecc7307]

He's out here freaking out about his 1080p enhanced biterate video quality, meanwhile I'm watching this video on a beautiful 1440p OLED screen... in 360p ahahahah

[@dungeon4971]

For reasoning model speed is much much more important

[@joshuafhiggins]

Can confirm that R1 knows about Winnie the Pooh

[@crossoverz6036]

u could ask "who is Winnie the Pooh look like in real world？and why is meme in china "

[@jorgeguzman8083]

except for hardcore ai people, most people don't know how to regularly use these models vs chatgpt.

[@kubre]

woho exposing entire yt and stream to linkedin???? that carries a felony you dont know?

[@tmaker502]

Open ai is fine for now. Deepseek is good as long as you don't hit the ccp trip wire

[@justinbaker84]

Outstanding video!

[@sahandehteshami7404]

Tokens wouldn't be so expensive if they weren't written in python.

[@prajnaparamitahrdaya]

Anyone checked the terms and conditions ? Is under PRC law

[@cariyaputta]

4o/o1/Sonnet are officially oudated. And their chat platform is free and unlimited too. What a banger.

[@---..]

Images don't store "hex codes", gradients aren't particularly hard to compress, Nvenc isn't a chip.... Has Theo been training on questionable AI output?

[@riggyz505]

Yet another Azure mention! Tbh I am too Azure pilled.

[@ryanmartin90]

CCCP: I like it

[@davefire2019]

Funny how China is just popping of this year 😊

[@BarakaAndrew]

If they wanna add bias it's better if they do it during inference not before, if the data has been removed the only way is train again using all the missing data which sucks coz we don't know. If u filter pig for example u are filtering so much stuff it makes the model so dumb, impossible to fine tune

[@crossoverz6036]

The character should be Winnie-the-Pooh 🤣

[@theshy6717]

you should be scared⚡ NOW ⚡

[@donwinston]

I suspect these LLMs are not really "intelligence". Instead of calling this stuff AI it should called KP for Knowledge Processor!

[@couchtourist256]

Is there an AI bubble? Yes. There. Is.

[@ИванРагозин-я8я]

what browser is he using? Where did Arc disappear to?

[@RadikAlice]

Zen maybe? He's covered his disappointment and frustration with Arc, and tried out Zen sometime later

[@mirzaangon]

It's Zen browser

[@crushfire2004]

No one can match China company for pricing, they have a surplus of electricity

[@CharifMakaoui]

Yes it's china !!

[@planesrift]

China, China, China

[@ruslansergazin8239]

Hooray. Now we have a really cool alternative, really OPEN source alternative

[@George-e9c2x]

Deepseek hasn't beat the final boss yet named o3 which massively overtakes o1. So nothing to be worried about yet

[@Bigredsleep]

Kind of , because it’s beating most tasks where you don’t need crazy reasoning where o1 was already to expensive

[@FilipeAguiarCarvalho]

Isn't o3 that model that costs 15 grand to run a question?

[@javierflores09]

there's little to be known about o3's capabilities besides their biased benchmarks, even if they may claim otherwise, so it's pretty much out of the equation right now. I'd wait till third-parties do thorough benchmarks on it, if they can considering how expensive the model is to run lol

[@zivavu824]

I encourage everyone to ask R1 some questions about unethical incidents and practices that took place in the USA(or any other western country), and then do the same with China's, to see the filtering in action :). I mean it's kinda obvious, as the model had to be approved by the state, but still, good to keep that in mind.

[@zacurrya9485]

Me: What is the Uyghur genocide - Deepseek starts generating a bunch of info -gets cut out mid way through and replaced with: "Sorry, that's beyond my current scope. Let's talk about something else." 💀

[@dijikstra8]

@@zacurrya9485 People who think detaining and rehabilitating extremists who were literally bombing Xinjiang while at the same time providing new infrastructure, upgrading housing standards, etc. in Xinjiang, is a genocide, tend to be the same people who thinks the indiscriminate mass slaughter of thousands upon thousands of children in Gaza is "self-defence". It just gets ridiculous and you should really look into who Adrian Zenz and company are, and reflect on how all the reports about Xinjiang coming out coincidentally have deep connections to the US intelligence machine, which is the same country that is pushing this idea, all the while willingly funding and arming an actual genocide in Gaza.

[@ithinkimhipster502]

The rabbit R1? Is that thing still relevant?

[@ithinkimhipster502]

Nvm, I just watched 5s of the video

[@felicityc]

@@ithinkimhipster502 XD

[@RadikAlice]

Self-hosting is always nice, but to me this is more like source available. An LM without the training data is more or less useless imo

[@TaiGroot]

Can confirm Jan 19th azure model inferrence speed plummeted for a day :)

[@flith8452]

CHINA NUMBA 1 🇨🇳🥇

[@yudatriananda3558]

first bruh

[@ericnl]

Third

[@christianwooldridge406]

First

[@parkerrex]

oai response is dropping a playwright fork that can almost order pizza

[@blengi]

how many tokens do you need to achieve apples to apples on o1 versus deepseek?

[@gro967]

Theo, please learn to pronounce Azure correctly 😁

[@Dragondezznuts]

Seeing its train of thought is amazing