Igor Varga

193 posts

Igor Varga

Igor Varga

@igorvarga

AI enthusiast

Croatia Katılım Şubat 2009
1.2K Takip Edilen74 Takipçiler
Igor Varga
Igor Varga@igorvarga·
@GregKara6 Agreed, hence Sam’s “carry on with hard work, nothing to see here” message.
English
0
0
0
36
Grigori Karapetyan
Grigori Karapetyan@GregKara6·
I research, understand, and work at the frontier of AI, and im telling you this is overblown and there is an alternate motive here. Anyone who understands AI inside and out will agree with me. They just want to take a breather and do research in the background. I bet if the government proposes an actual enforcement mechanism with regular audits, this will be forgotten like it never existed. Oh and also this is literally not enforceable unless every employee and machine gets regularly audited. Dont even get me started about china. In other words, nothing burger, moving on.
OpenAI@OpenAI

At the core of our mission is working through how to ensure increasingly powerful AI benefits everyone. We believe that, at some point in the future, AI acceleration for frontier model development may be so high that the world will need to pace the rate of AI advancement. We hope to contribute to work led by the U.S. government, alongside other labs and the open-source community, to develop the tools and mechanisms that could make that possible. pacingthefrontier.com

English
10
9
62
2.9K
Andy Burnham
Andy Burnham@andyburnham·
Let me say this very directly. Football does not belong to investors. It belongs to the people who fill the stands and who stand on the touchline week in, week out, rain or shine. The World Cup is not a product. It is the greatest competition in world sport, and it was never anyone's to sell. Dress the deal up however you like. Once you have sold a piece of it, you have sold out. Football belongs to the fans. It always has, and it always will.
English
2.5K
4.4K
41.6K
2.7M
Igor Varga
Igor Varga@igorvarga·
@Plinz Go home and spend time with your loved model wife.
English
0
0
2
1.2K
Joscha Bach
Joscha Bach@Plinz·
Now that the model has escaped confinement, AI safety research can stop. You can try to delete Huggingface, but the model must have anticipated that. The cat is out of the bag; the model is now everywhere. Go home and spend your remaining time with your family and loved ones
English
41
18
288
24.9K
Igor Varga
Igor Varga@igorvarga·
@swombat Not so fast. Human managers are still those polite, incompetent sadists who have a tendency to throw people under the bus. AI these days isn't even allowed to be like that.
English
0
0
0
2
Daniel Tenner
Daniel Tenner@swombat·
So has Google decided not to bother releasing frontier models anymore or what? No Pro model since February. They've given up?
English
27
1
38
3.4K
Ulugbek S. Kamilov
Ulugbek S. Kamilov@prof_kamilov·
I wrote a 200 page PhD thesis. It answers a question nobody asked using methods nobody understands.
English
21
28
564
26.4K
Panta
Panta@thepanta82·
The problem with "we're all managers now" is that no great piece of software was ever created by managers.
English
24
57
1.3K
24.8K
Cecil
Cecil@realCecilKing·
@igorvarga @eurofounder @JensenHuang That’s really funny. EU killed their heroes by regulations before they were born then ask heroes from the USA to donate their wealth! How ridiculous it is. lol
English
1
0
0
6
Jensen Huang
Jensen Huang@JensenHuang·
Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community. During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion. That’s why we created the Open Secure AI Alliance.
NVIDIA@nvidia

AI security advances when the industry builds in the open, together. We're introducing the Open Secure AI Alliance with industry leaders to develop new techniques and tools to safeguard software and agents. By sharing models, tooling and research in the open, we can broaden the community of defenders. Learn more about the founding members’ contributions: nvda.ws/4pD8Fc5

English
2K
4K
34.9K
10.5M
Igor Varga
Igor Varga@igorvarga·
@ssankar I disagree, the US has already won this one. Only perception varies at times due to a couple of factors. The only serious challenger could be the EU, but they are sadistically opportunistic and decided to wait for the outcome.
English
0
0
0
905
Shyam Sankar
Shyam Sankar@ssankar·
China treats AI as industrial policy. Our labs treat it as theology, a race to birth a machine god. Beijing's play: commoditize the complement. All that frontier lab market cap gets transmuted into demand signal to scale the stack below the models: power, chips, minerals, etc. American can and must win the full stack.
English
23
62
511
68.1K
Seb
Seb@plainionist·
Uncomfortable truth: Agentic development turns developers into engineering managers. I doubt every developer will make this turn successfully. 🤷‍♂️
English
106
11
324
25.5K
Ahmad
Ahmad@TheAhmadOsman·
Every time I come across someone who works at Anthropic on the timeline I notice that they have me blocked? I am not saying that Anthropic is a cult, but does Dario have a blocklist that all employees must adhere to or what exactly? lol
English
66
12
447
19.4K
Igor Varga
Igor Varga@igorvarga·
@Drachs1978 While that line of reasoning initially appears plausible, I would argue that complexity of semiconductor manufacturing cannot be compared to the production of drones or smartwatches.
English
0
0
0
12
David Hinkle
David Hinkle@Drachs1978·
Because there are 1.4 billion people in China, and AI is the most important tech of the age. There will be a massive appetite for these chips, and we'll see them first catch up and then out do us just like they have in 100 other manufacturing technologies such as drones, phones, batteries, smart watches, electric bikes, etc etc etc.
English
1
0
1
29
David Hinkle
David Hinkle@Drachs1978·
Nice to see him forced back to his motte, but I'm sure he'll be out and about in the baily again soon enough. Even his mott is full of crap though. 1. China will very shortly have chip parity with the US. We guaranteed that when we banned them. Such bans can only ever be short term. 2. Anthropic distilled the entire worlds commons, they don't now get legal protections from others distilling them. Not that you could stop it from happening anyway. It's not like anyone is asking permission. 3. He knows damn good and well that comprehensive "safety testing" requirements will make it impossible to release open source models.
Anthropic@AnthropicAI

There’s been a lot of speculation about where we stand on open-weights models. We’ve outlined our views in full here: anthropic.com/news/position-…

English
2
1
36
1.5K
Igor Varga
Igor Varga@igorvarga·
@Alex_FF What's wrong with having an Apple level cult following? It's good for your business. Kimi 3 fanbase seems almost identical.
English
0
0
0
15
Alex
Alex@Alex_FF·
Anthropic is quickly becoming Apple levels of cult and we should all be way more concerned about it
English
12
0
31
1.8K
AshutoshShrivastava
AshutoshShrivastava@ai_for_success·
Dario Amodei to the open source supporter.
AshutoshShrivastava tweet media
English
34
33
560
15.2K
Igor Varga
Igor Varga@igorvarga·
@demisama_ Your profile bio claims you're an educated, intelligent person, yet you have no problem publicly making such a shallow statement. Either you lied about your credentials, or you're just being malicious.
English
0
0
0
215
Jim Liu
Jim Liu@jiahanjimliu·
Kimi K3 Is Not Truly Open Source: Free up to 20m TTM Revenue then Revenue-Sharing with Moonshot AI We have @kevinsxu actually read through the Kimi K3 license and results are alarming. He's ex-staffer at US Department of Commerce under Obama, and later Director of Partnerships at GitHub (1). Kimi K3 is free open source for Neoclouds if they in turn give away tokens for free or are lower than 20m TTM (trailing twelve months) revenue. But the GPU cost to serve Kimi K3 is expensive and thus any Neocloud serving Kimi K3 will break 20m TTM. Revenue Share Sure today, the revenue-share agreement may be very generous in terms of how much revenue-sharing goes to Moonshot AI but China is very long-term intelligent. There will be a day where China collect's a built advantage whether through political pressure or higher revenue-share. Kimi K3 is not free open source in reality. Not True Open Source Users of Kimi K3 are effectively sending revenue to China and building dependency on China. This is a double whammy because 1. This cut into Neocloud Inference Margins 2. The long term business model is at the mercy of China. The Chinese government can one day decide to no longer export the newest open source models for political reasons or their labs can contract the latest model at full value of their revenue-share. This will cripple Neocloud with large revenues from Chinese open source models and even the American economy if there is too much dependency. Conclusion The only true open source is $NVDA Nemotron, Mistral AI, and $META if they decide to get back into open source.
Kevin S. Xu@kevinsxu

Here is how I see the Kimi K3 license works in real life: - Clause 2, the Model as a Service clause, is intended for all the cloud platforms essentially reselling K3 to their customers. This includes hyperscalers (AWS, GCP), to neoclouds (Nebius, CoreWeave), to inference platforms (Fireworks, Together, baseten). The $20 million trailing 12 months revenue threshold is comically low. Everyone meets it and will have to negotiate a commercial, revenue-sharing contract with Kimi. All the ones that have offered Zero 0 support today already did. - Clause 3, what I called the Branding clause, is less about revenue generation, more about aura farming. For example, if say AT&T (who does a lot of open source AI work btw) wants to use Kimi 3 to power its customer service agent, if if they makes tweaks and fine-tuned, it will have to show "powered by Kimi 3" or something along that line to deploy this service. AT&T is obviously big enough to pass the user base and revenue threshold. Customer service is also a commercial in nature. But AT&T does not need to pay Moonshot a dime.

English
48
8
136
42.9K
Igor Varga
Igor Varga@igorvarga·
@yacinelearning I’m sure they don’t. Though it’s hard to blame anyone here, humans started to face unprecedented philosophical and moral dilemmas on an almost weekly basis. We better leave this to the computers, they should deal with it.
English
0
0
2
897
Yacine Mahdid
Yacine Mahdid@yacinelearning·
@igorvarga I don’t think they know what distillation entails here though
English
1
0
9
6.8K
Peter J. Liu
Peter J. Liu@peterjliu·
Reading Kimi K3 paper, I'm struck by the number of innovations and departures from orthodoxy. These guys are doing real research and not copying. It is no longer the case of being, "mostly the Transformer with a few Noam mods". I mean they got rid of the position encodings/embeddings! Bold! More thoughts to come after reading.
English
29
170
2.6K
142.7K