hnfong

5 ms·

61.
by hnfong 3mo ago
Only those who intended to use the new technology as a tool to achieve dominance and control are freaking out. And it's the only viable tool the US has left. It's reasonable they are freaking out.
62.
by hnfong 3mo ago
> European models are so far behind because they don't resort to these tactics on a massive scale. Basically every other country is entirely dependent on 2 countries for frontier AI. You may or may not be factually correct in your o
63.
by hnfong 3mo ago
You two are talking about different things. You are pointing out that OpenAI can cancel user's accounts for almost any reason, and nobody can really force them to serve customers that they suspect are distilling their models. That&#x27
64.
by hnfong 3mo ago
Well, then let's hope you're wrong and the AI bubble won't also blow up private equity and wipe out people's retirement funds...
65.
by hnfong 3mo ago
It sounds vaguely similar to what Cerebras.ai is doing
66.
by hnfong 3mo ago
Still cheaper than Cola
67.
by hnfong 3mo ago
It means there's a non-trivial chance a future version of the model will know private information about you. Maybe you're super careful with this stuff, but with agents and harnesses being given access to user data and accounts, I
68.
by hnfong 3mo ago
The markets can be irrational for a while, but if the Chinese models are about 90% of the performance of OpenAI and Anthropic models, and the Chinese companies are < 10% of OpenAI and Anthropic's proposed IPO valuations, something h
69.
by hnfong 3mo ago
Supplementing with evidence: https://x.com/stevibe/status/2026227392076018101
70.
by hnfong 3mo ago
Claude Sonnet 4.8 reproducibly identifies itself as DeepSeek when asked in Chinese: https://x.com/stevibe/status/2026227392076018101 I mean, people can point fingers however they want, and the fact is nobody actua
71.
by hnfong 3mo ago
People have been able to run DeepSeek v4 flash with a high spec Mac.
72.
by hnfong 3mo ago
In an extreme sense, yes. If the free stuff was given for a prolonged period of time with the expectation that it would continue indefinitely. In modern terms, it's enabling others' pathological dependency on your free stuff. The
73.
by hnfong 3mo ago
At least one of those is true.
74.
by hnfong 3mo ago
I agree, but you think OpenAI and Anthropic are not already at their absolute peaks of hype and investment? At valuations of about trillion each, I don't know where else they'd go..
75.
by hnfong 3mo ago
Also potentially a performance (in terms of output quality) cost. DeepSeek is cheap on a per token basis but lags behind in the benchmarks, perhaps it was a calculated tradeoff.
76.
by hnfong 3mo ago
Ironically Windows XP was the first "consumer" (i.e. not "workstation") Windows version to not crash several times a day . In the 98/ME era, it was pretty common to reboot every couple hours or the system would ran
77.
by hnfong 3mo ago
So I clone a repo from Github and try to do due diligence before compiling and executing the code. I open the project with my IDE/text editor, and boom, I get pwned while doing the due diligence. Unless it's Cursor's stance t
78.
by hnfong 3mo ago
If we take your logic to the extreme, then there are never any legitimate code execution vulnerabilities in web browsers since the user is not supposed to visit untrusted websites.
79.
by hnfong 3mo ago
Same here. The git commands were confusing to me until I actually saw a Linus video on youtube explaining how the git data structures worked under the hood. It was dead simple, and one could easily mentally map the tools' functions ont
80.
by hnfong 3mo ago
I always feel obliged to link to this when contextually appropriate: https://en.wikipedia.org/wiki/Whataboutism
81.
by hnfong 3mo ago
I'm not sure what you're talking about. Europe dismantled their own nuclear plants and are actively trying to ban electric vehicles from China. Getting the rest of the world on board would be much easier when you (presumably are i
82.
by hnfong 3mo ago
Well, computing and AI technology is going to be nationalized and state owned in a couple years. Good luck.
83.
by hnfong 3mo ago
> At the Embassy of the Free Mind in Amsterdam, we’ve created https://SourceLibrary.org , a collection of over 15,000 translations of Renaissance and premodern books in NeoLatin, Chinese, Sanskrit, etc. Wow, this is like... ex
84.
by hnfong 3mo ago
I think it's not really much of "hitting a ceiling", but more like plateauing, with growth and improvement slowed down so much that it seems to stagnate. It's all relative. In computing we're used to Moore's la
85.
by hnfong 3mo ago
This is kind of like the old school captchas where they distort the letters and numbers so badly that humans can't really read them, and specifically trained AI models might surpass humans in performance...
86.
by hnfong 3mo ago
You can still be competitive in "casual mode" using total time used to solve all the words.
87.
by hnfong 3mo ago
I think the issue here is that there's no objective difference between "alignment" and "actively trying to shape its replies to fit a political narrative", beside the fact that the latter refers to the kind of polit
88.
by hnfong 3mo ago
No, but technically, all models go through an alignment phase where you feed data that's aligned to your goals and values to train the model so that it will exhibit the kind of behavior you want. There's no "politically neutr
89.
by hnfong 4mo ago
Also https://xkcd.com/217/
90.
by hnfong 4mo ago
Everyone distills/copies training data. A couple months ago when Anthropic was complaining about Chinese distillation, people found that Claude self-identified as "DeepSeek" when asked in Chinese: https://x.com&#x2

More ›