Hacker News: nostradumbasp

New comment by nostradumbasp in "Exposed DeepSeek database leaking sensitive information, including chat history"

nostradumbasp — Thu, 30 Jan 2025 00:33:34 +0000

Because you can download the models yourself and run them on your own hardware bypassing all of those concerns. Also, I would be far more worried what my government might do with my chat logs then a foreign one.

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 22:55:01 +0000

I completely agree with you and apologize for cheapening both the nuance and complexity where I did.

My personal take is this. What deepseek is offering is table scraps for the CCP's actual ambitions with what we call AI. China's economy is huge on industrial automation, and they care a lot about raw materials and manufacturing efficiently than say the US's interests.

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 21:29:03 +0000

Yea but I think the OPs point is something along the following lines. Not everything you buy from China, or every person you interact with from China is part of a clandestine CCP operation. People buy stuff everyday from Alibaba and its not a CCP scheme to sell portable fans, or phone chargers. A big chunk of the factories over there are US funded after all... Just like how it's not a CCP scheme to write a scientific paper, or create a ML model.

Similarly, I see no evidence (yet) that DeepSeek is a CCP operated company anymore than saying any given AI start up in the US is a three letter agencies direct handiwork or a US political party directive. The US has also supported genocides and a bunch of crazy stuff, but that doesn't mean any company in YC is part of a US government plot.

I know of people who immigrated to China, I know people who immigrated from China, I went to school with people who were on visas from China. Maybe some of them were CCP assets or something, but mostly they appeared to me to be people who were doing what they wanted for themselves.

If you believe both sides are up to no-goodery thats in the face of the OPs statement. If you think it's just one, and the enemy is in complete control of all of its people doing all of their commerce then I think the OP may have a point.

New comment by nostradumbasp in "SmolGPT: A minimal PyTorch implementation for training a small LLM from scratch"

nostradumbasp — Wed, 29 Jan 2025 19:53:15 +0000

Cute! Keep making fun things.

New comment by nostradumbasp in "The protester's guide to smartphone security"

nostradumbasp — Wed, 29 Jan 2025 19:48:15 +0000

They actually sell those now. Well they use "faraday fabric". So that's another solution, but your phone one your head and put a foil hat on it.

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 19:42:31 +0000

Sounds hilarious but... https://techstory.in/trump-is-accused-of-using-ai-to-compose...

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 19:41:42 +0000

Probably more like specialized tools to help spy on and forecast civilian activities more than anything else. Definitely with hallucinations, but that's not really important. Facts don't matter much these days...

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 19:37:14 +0000

There's no evidence for almost any of this, and even when there is, we won't see it. Just like 95% of posts on here.

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 19:36:14 +0000

Maybe not $10 unless they are loss-leading to dominance. Well they actually could very well do exactly that... Hm, yea, good points. I would expect at least an order or two of magnitude higher to prevent an inferno.

Lets be fair though. Replicating TSMC isn't something that could happen quickly. Then again, who knows how far along they already are...

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 19:26:30 +0000

https://openai.com/global-affairs/introducing-chatgpt-gov/

And don't forget the billions in partnerships...

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 19:15:27 +0000

No reason why you were downvoted. This is completely valid.

New comment by nostradumbasp in "OpenAI says it has evidence DeepSeek used its model to train competitor"

nostradumbasp — Wed, 29 Jan 2025 19:14:42 +0000

I really doubt it. If that's the case the US GOV is in serious shit. They have a contract with OpenAI to chuck all their secret data in there... In all likelihood they just distilled. It's a start up company that is publishing all of their actual advances in the open, with proof. I think a lot of people run to "espionage" super fast, when reality is, the US probably sucks at what we call AI. Don't read that wrong, they are a world leader obviously. However, there is a ton of stuff they have yet to figure out.

Cheapening a series of fact checkable innovations because of the country of origin when so far all that they have showed are signs of good faith is paranoid at best and propaganda to support the billionaire tech lords saving face for their own arrogance at worst.

New comment by nostradumbasp in "Cali's AG Tells AI Companies Almost Everything They're Doing Might Be Illegal"

nostradumbasp — Wed, 29 Jan 2025 15:42:22 +0000

Fine of five hundred dollars and a personal donation to a re-election campaign required.

New comment by nostradumbasp in "LinkedIn removes accounts of AI 'co-workers' looking for jobs"

nostradumbasp — Tue, 28 Jan 2025 23:49:27 +0000

More than anyone could ever count.

New comment by nostradumbasp in "Cleveland police used AI to justify a search warrant. It derailed a murder case"

nostradumbasp — Tue, 28 Jan 2025 22:00:20 +0000

It couldn't lie at scale though. The amount of effort it took to doctor someones face believably onto film in the star wars era was tremendous. Now it can be done on the terabyte per second scale in an individualized fashion.

Ring camera feed could show your actually nice neighbor kicking your dog. Could show a lowly employee that pissed off a manager stealing the company stapler. Could show a spouse having an affair, fake messages (one of the highest incidents of murder in the US is due to this). Swatting on a new level by impersonating your voice. All of these are things a somewhat motivated highschooler can now do.

Imagine what an agency or enterprise could do? Especially if they intercepted your network activity to turn you into any level of puppy kicker you could imagine because of your politics, your position, or for testing out the success rate of such a program.

New comment by nostradumbasp in "Cleveland police used AI to justify a search warrant. It derailed a murder case"

nostradumbasp — Tue, 28 Jan 2025 18:03:39 +0000

Mostly agree. We all know how awful swatting has been for many victims. Also personalized scams and deception at scale. I also worry about homebrew automated weaponry used for cowardly terroristic purposes and how that could imaginably be around the corner. It's a pretty scary topic to consider the depths of what can be done by individuals using these tools without regulations.

New comment by nostradumbasp in "Cleveland police used AI to justify a search warrant. It derailed a murder case"

nostradumbasp — Tue, 28 Jan 2025 17:58:23 +0000

It's a little different now. Someone can deepfake or even face swap videos/images/audio and send an angry mob to someones house. Or be convinced to handle matters themselves...

"Pictures/videographic evidence doesn't lie" isn't as true now as it was 30 years ago at the amateur level. Especially with coercion and human flaws.

New comment by nostradumbasp in "Qwen2.5-Max: Exploring the intelligence of large-scale MoE model"

nostradumbasp — Tue, 28 Jan 2025 17:50:34 +0000

They are being attacked as well.

https://apnews.com/article/deepseek-ai-artificial-intelligen...

New comment by nostradumbasp in "Cleveland police used AI to justify a search warrant. It derailed a murder case"

nostradumbasp — Tue, 28 Jan 2025 13:48:16 +0000

It's definitely not looking good. Now that we have more open source models, for better and for worse, we have to worry about politically motivated amateur's and vigilantes too.

Startup DeepSeek says it was hit with 'large-scale malicious attacks'

nostradumbasp — Tue, 28 Jan 2025 13:03:04 +0000

Article URL: https://apnews.com/article/deepseek-ai-artificial-intelligence-be414acadbf35070d7645fe9fbd8f464

Comments URL: https://news.ycombinator.com/item?id=42851832

Points: 5

# Comments: 1