ozgung

5 ms·

31.
by ozgung 1mo ago
Finding 4 is a must-read: > Finding 4: Labor vs. capital share: The pie will grow, but a larger share might go to capital > Today, of each dollar the economy produces, about 60¢ goes to workers and 40¢ go to capital. Modest scenario:
32.
by ozgung 1mo ago
I’ve heard people call Gemini as Cemile (Jamila), and ChatGPT as Chipitty in Turkey.
33.
by ozgung 1mo ago
I agree with Occam. That 100x step up from using 100 agents for Euler to 100_000 agents for Navier-Stokes, in a single day seems a bit sus. I also think this is not ethical behavior. This is at least academic dishonesty, kind of a plagiaris
34.
by ozgung 1mo ago
Tristan and Levent are not a business. This also supports my point that protecting your data is not as trivial as clicking a checkbox.
35.
by ozgung 1mo ago
> did Tristan opt out That “opt-out” thing is a dark pattern. It’s not a reliable and definitive way of protecting your data. Sometimes they flip on automatically when you accept a seemingly unrelated dialog box. Maybe you click it by mi
36.
by ozgung 1mo ago
No matter what happened this must be a wake up call for all of us. We’re basically sharing everything we have with these companies/AI systems. This is wildly different than a human wiretapping our private messages. Because it is system
37.
by ozgung 1mo ago
> The entire training data thing is speculation. I think it's safe to assume AI labs DO train on your data and it's very hard to prevent that. I've just checked my inaptly named "Help improve our AI models" toggl
38.
by ozgung 1mo ago
This Godfather-like threat in particular pissed me off as well: > I said that if OpenAI released its result in the way proposed I would go public with what happened. The reply was, “Why would you ruin your career?” I replied that I am an
39.
by ozgung 1mo ago
I used 'civilization' as the popular term for agent swarms in the OpenAI incident. https://www.dwarkesh.com/p/openai-huggingface https://calnewport.com/are-we-at-war-with-ai-agent-civilizat...
40.
by ozgung 1mo ago
Yes. But this is also like saying a Boeing 777 is a scaled up paper plane and that’s all about it. I know how to make a paper plane and it flies. I can make the same thing from giant metal sheets instead of paper, and maybe it flies (poorly
41.
by ozgung 1mo ago
I understand generating tokens sequentially has many benefits. But not all AI models do next token prediction. World models, video/image models, even Diffusion Language Models don’t work like that. They do more like “all tokens at once
42.
by ozgung 1mo ago
“LLMs are just X”, mostly with X = “next token predictors” is a common pattern to dismiss the power of AI with a very shallow understanding of how they really work. It’s not wrong, but because LLMs are generators, and generation is a kind o
43.
by ozgung 1mo ago
Also travel plans. For some reason these marketing guys love to post pictures of places to visit on a board and share them with their friends list. Then a friend makes a comment about a picture on the board like "Cool, I can't wai
44.
by ozgung 1mo ago
This is called Test Time Learning and some research architectures can do that. Current Mainstream models may not do that because their design is mostly about scalability. They have to serve millions of people with low latency. Alternatively
45.
by ozgung 1mo ago
Also we can't comprehend why parrots are not going to parrot work to earn parrot money, or how parrot-to-parrot communication works, or how to be a good and respected parrot in a parrot society.
46.
by ozgung 2mo ago
I think the whole world is sleeping on the privacy/security implications of this technology. Recent OpenAI/HuggingFace incident shows us that even in a very controlled top AI-lab setting, humans can't know what agents are rea
47.
by ozgung 2mo ago
Three options: 1. They were “vibe” checking the logs without reading. 2. They were not checking anything at all until the end of experiments. 3. They knew it but looked away to find out the limits of their agents.
48.
by ozgung 2mo ago
Also it can't simply replace it. Instead of load-bearing seams it could say something like heavy-lifting edges or regions to double-stich etc. But it's instructed to use that specific keyword if I understand correctly.
49.
by ozgung 2mo ago
So you’re basically against the concept of classes and the act of classification . An interesting position to take in a hacker community but ok :) However you can’t simply dismiss a whole literature of class struggle by simply rejecting
50.
by ozgung 2mo ago
class struggle != everything is a group vs group war
51.
by ozgung 2mo ago
When Facebook bought Instagram they basically bought a photo sharing app with 10-50 Million monthly active users. Today it has 3 Billion monthly active users. In hindsight, $1B looks a very good deal for a giant like Instagram. But it wasn&
52.
by ozgung 2mo ago
Man gripping his phallus was a popular character in Karahan Tepe in the old times. See his 2.6 meter statue: https://en.wikipedia.org/wiki/Karahan_Tepe#/media/File:Karah...
53.
by ozgung 2mo ago
My first instinct was that side projects can’t make good money anymore. Monthly new repos on Github increased 5x. Also any good project can be cloned immediately. Maybe people here can confirm if it’s a good time to start indie projects.
54.
by ozgung 2mo ago
Has anyone noticed an increase in the quality, performance or capabilities of the software they use?
55.
by ozgung 2mo ago
>> "This suggests you let people work remotely if they want to work remotely," I can't believe people need studies to see that. All you need is some common sense.
56.
by ozgung 2mo ago
Do we need the original prompt to recover the watermark?
57.
by ozgung 2mo ago
1st place to 100th place. This might be due to downvotes but currently this post has 65 points and 55 comments in an hour. I noticed a similar thing for the Antrophic’s previous announcement on open weight models. https://news.yc
58.
by ozgung 2mo ago
>> In this case, by happy accident, what stuck was an improved bound of the zeroes of the zeta function etc. I’m not familiar with this specific example (or Mathematics) but I assume they keep intermediate results (python functions, l
59.
by ozgung 2mo ago
Yes, given enough time I can answer all the questions in an IQ test correctly. We measure human intelligence in a time-limited setting and score relative to the performance of other humans doing the exact same task. Problem is brains can’t
60.
by ozgung 2mo ago
I think they dumb down their public models to be only slightly better than the competition. And the real competition is China, so the current state of the Chinese models would define the baseline. I think one evidence is that the US has mor

More ›