EthicAI
Region: AU
Sunday 14 June 2026 01:07:08 GMT
19807
870
64
93
Music
Download
Comments
X :
Local LLMs don’t address the risk of another anthropic type situation. The us government can just as easily tell companies like MS to switch off locally run software with a stroke of a sharpie. The bs with Fable is more likely to do with anthropic giving the us government the big middle finger when the dod wanted to use them to help create skynet.
2026-06-14 03:13:07
7
oaiat99 :
you don't need a brain the size of a planet to summarise your emails
2026-06-14 01:23:52
43
Blindfolded Sight :
Yeah we will all have our own ai companion
2026-06-14 17:36:23
0
Reddata33 :
Stop using AI to write your scripts. There is so much “It’s not just this … It’s that” framing, it distracts from you otherwise excellent points.
2026-06-14 01:30:48
2
BGSnjjcffv :
That makes sense. Apple’s IOS 27 is bringing a Siri AI dedicated app. Everyone should be downloading open sourced LLM’s before the government make them illegal
2026-06-14 12:16:05
4
Jay Dee :
3.3B parameters? You are not going to get anything close to the 100+ B parameter model capabilities. If you are intending to run models so small you had better understand RAG.
2026-06-14 11:59:17
0
Plexo :
They banned Anthropics Mythos / Fable because Trumps friends are about to lose billions. Basically Trump and friends bought a bunch of pre-IPO shares at $100 because thats what they said the company was worth. Anthropic then made a statement saying all these pre-IPO shares are VOID. After being evaluated properly a week later at 10x what Trumps fund manager friends told the company it was worth, Anthropic can now pull the pre-market shares which were sold for $100 and charge what they are worth. $1000+ Trumps been told to sink the stock so that Anthropic doesnt trigger a buyback prior to the IPO because Him and his buddies are sitting on millions, potentially billions in shares. If anthropic buy it back, him and his friends get $0. a single company already has over a trillion dollars in IPO funds waiting to purchase anthropic shares. a single company. which means the shares will likely be worth between $1000 and $2000 not $100. Meanwhile, running these models locally is becoming feasible. Mythos and Fable are actually less than 20TB.... that fits on a large off the shelf hard drive... so in 10-20 years I can 100% see people having these available locally on people's computers. someone just needs to let it cook on their home computer for a decade and they will have the model good to go. freeware style. Mythos / Fable will be back once Anthropic either reconfirm all the pre-ipo shares being traded will be 'honored' at the original price or they take the model, fly over to europe and restart the company there (or release the model to the public 'accidentally'). I cant see Anthropic winning this fight because those guys dont play fair and have 'wet teams' if they dont play ball. Billions are at stake. Also: It screwed my workload entirely. two developments stopped in their tracks. Opus was just not good enough. 3 months work on opus with lots of errors vs 1 week on Fable with no errors. screwed us royally.
2026-06-14 08:01:07
4
ElRei 🇺🇸 :
I'm confused, how is having a local 3b model the same as having access to Fable?
2026-06-14 14:02:28
0
LivOC :
Yep, all software needs to go back on premise
2026-06-14 09:10:36
3
jeebs :
ollama been around for a while now if youre interested in local models. I have a 14 billion parameter models running on a macbook m4
2026-06-14 17:29:59
0
Cr :
Have you tried phi? It’s rubbish
2026-06-14 06:50:53
1
coldpointblue 🇸🇪 :
🧐History says doing THIS, keeping advanced science secret, WILL BACKFIRE IN PROPAGATION by any means necessary, like it did in the 1950’s and end-up with biggest rivals having equal or better since ’concerned experts’ will always recognize ONE DOMINANT POWER ALWAYS ENDS BADLY FOR ALL, so they’ll leave to opposing sides or share key info over time TO GUARD AGAINST VAST CATASTROPHIC ABUSE FROM 1 SIDE.
2026-06-14 06:50:50
5
Rudi Van Zyl :
If you connect anything with Microsoft, I dont trust it, even if it runs locally on a 486DX266, i will not trust or touch it. They will have a way to put adverts into it, and to get data from you.
2026-06-14 06:41:37
4
Vi Killeen :
Do you post content on insta ? Would love to follow there
2026-06-14 02:08:01
0
The_RZH :
Switching to less capable local models doesn’t address the issue of being blocked from the best models. No hospital was running on those anyway. It is companies writing new applications and conducting research that benefit from frontier models and they can be denied.
2026-06-14 04:03:58
0
DBest Tech 🇹🇹 :
this isn't because they want users to have LLMs for themselves. this is good for users. but it will be far too costly for a user to extract real value. this is for the user that can spend $50k - 1mil......... ok so I'm listening to the second half of this video. ummmm yeah what she said. think lawyer firms that want to analyze a case and look for insights. medium sized security firms that need vision, drones and personnel to coordinate. we enthusiasts will have some level of autonomy as a byproduct.
2026-06-14 15:47:47
0
Beach Perfect 🇧🇿 :
US companies are being prohibited from purchasing international companies due to US government data grabs.
2026-06-14 16:03:39
0
coldpointblue 🇸🇪 :
🧐Clearly, this gives the open source models from elsewhere the clear advantage over the long run because precisely in response to this BS the BEST people everywhere else, will not want to be boxed-in arbitrarily for their life’s best work, since for most it’s not about money.💰
2026-06-14 06:38:54
1
⚜️DeccyLifts⚜️ :
I mean we have had local models for ages they simply dont come close to cloud ones, I can run a 103b model on a 4090 with some tensor tricks and manipulation but that is still nothing close to even some of the worst cloud models. Unless you have one hell of a setup local is pointless atm
2026-06-14 17:49:01
0
Otherground 🏴☠️ :
Hey Guys. Ai in the US is dead. Look up ALL the resource constraints from the Middle East - helium and aluminum? Shortage. No chips or data centers.
2026-06-14 14:02:57
0
011 :
I have windows 10 debloated server edition for games. Plus i am usiing linux for important stuff. Windows update disabled. Using oppus which works perfectly but for agent i use qwen 3.6 35b 3a perfectly intelligent to do anything on the web i ask. Runs by itself for hours
2026-06-14 14:58:07
0
bhagatriks 🇨🇦 :
very good take.
2026-06-14 14:30:59
0
pauldavidpauldavidpa :
Are small local models more efficient because complexity scales quadratically? I.e.: If there are n relevant variables then there are n^2 - n combinations of 2 different variables. I've heard "exponential" but I haven't seen any test of linearity of logarithms.
2026-06-14 14:22:14
0
jeebs :
Also you already have the hardware to run it, I run decent models on a debian laptop I have with 16gb of ram and just a ryzen 5
2026-06-14 17:39:56
0
SSish0 :
Don’t you just think it makes business sense? German government have taken action to move away from reliance on American technology. If most major jurisdictions (where they make their profits) did this it would significantly impact their profit margins.
2026-06-14 01:46:26
0
To see more videos from user @ethicai_au, please go to the Tikwm
homepage.
