oezi

7 ms·

1.
OpenAI Astra checks webcam for user(twitter.com)
7 points by oezi 3d ago|1 comments
2.
Wrong link(xcancel.com)
1 points by oezi 3d ago|3 comments
3.
by oezi 4d ago
The crucial question is how did the agents get recruited or bootstrapped into their malicious collective. Did the agents manage to prompt inject into the system prompt a way for each new agent to escape their jail? Otherwise how could the a
4.
by oezi 23d ago
Being successful or not in your wealth doesn't make you right or wrong about anything with the exception of maybe running the business that generated your wealth. Nietzsche presented two prototypical sets of human who interface with la
5.
by oezi 23d ago
Nietzsche also described it well when noting there are those who deem themselves nobles and feel they are righteously above the laws and those who derive their identity from suffering under the same laws.
6.
by oezi 1mo ago
One aspect missing from TFA is that in biological brains a lot of the training is performed during query time. If we assume 1 GB of DNA is enough to encode the brain's overall structure we still do need training data (e.g. visual/
7.
by oezi 1mo ago
I blame RLHF entirely for this. Nobody used to talk like AI speech before.
8.
by oezi 1mo ago
Don't talk about other people in such a way. If we had a way in Claude to mark a word/token and downvote/reduce it logits then it wouldn't be to hard to send loadbearing to token valhalla. In the future, I hope we get a
9.
by oezi 1mo ago
I would say it is sad that there are people who use Grok when there are so many other choices available which don't come with the issues of supporting Musk. It is not all just 'politics'. Take a stand on some issues. It doesn
10.
by oezi 1mo ago
> 10 USD per Module isn't really competitive with disposable RFID. Great for a relay though it seems!
11.
by oezi 1mo ago
In Europe bonds are mostly out of fashion because they pay very little or have considerable risk of default. At least for retail investors.
12.
by oezi 1mo ago
3% less annual returns than the S&P500 over 10 years and the entire fund lifetime. That's a considerable difference.
13.
by oezi 1mo ago
I would be really interested to learn what the default advice for retail investors is across countries. In Germany the consensus is MSCI World or FTSE All-World ETFs. I believe in the US most advice goes to VTSAX (US Total Market), VOO (S&a
14.
by oezi 2mo ago
True but also each day they open slots due to cancelations or often have a system where you can just go in the morning and wait for somebody to not show up. Or allow things to be done online. It is as with all bureaucracy: you got to be cra
15.
by oezi 2mo ago
Yes, the preference for bonds seems deeply ingrained. Have you found a way to steer LLMs away from old school allocation (bonds, gold, cash)?
16.
by oezi 2mo ago
One key thing with MoE I am still not understanding is why we don't keep the same expert in memory for a larger number of tokens than 1. Why do we route to some other expert every token? Wouldn't it be more memory efficient to gen
17.
by oezi 2mo ago
Your take is too fatalistic because any reduction in Co2 emissions is helpful. It isn't an all or nothing thing. The cost estimates for eliminating Co2 emissions are also not very high. In the range of 1-2x total GDP. Would be doable i
18.
by oezi 2mo ago
Maybe we need LLMs which have an internal dialogue rather than the current monologue.
19.
by oezi 2mo ago
I also was confused by this and the usage of the term bet for low/med/high risk. Are up and down supposed to model buying calls/puts? Or buying the stock itself? Wagering 8800 USD of 10000 on high to get 115 USD return was po
20.
by oezi 2mo ago
At least one graph showing global total datacenter compute over time and projected to come online would have been helpful.
21.
by oezi 2mo ago
The article makes a great point that the token industry is going to be commoditized as time goes on. Following this argument the key for each player will be the underlying cost structure and serving capacity to offset the upfront R&D co
22.
by oezi 2mo ago
How close do you aim for when it comes to drop-in vs whisper.cpp? Are timestamps per word and character something aimed for? How about multi-lingual transcription or hallucination suppression? The github page doesn't seem to go into de
23.
by oezi 2mo ago
Nasdaq 100 is up 13% YTD, so you may have a point. 20% more isn't so much more.
24.
by oezi 2mo ago
Claude code doesn't expose the model to the usage of sleep or timing commands. For instance it cannot set up loops on its own (and only in rare cases can stop them). To prevent cache timeout you can do: /loop 55m continue if any o
25.
by oezi 2mo ago
On first read of the headline I thought they are now offering a way to join Anthropic by paying them 65k USD. Sounds like a steal to get Anthropic on your resume!
26.
by oezi 2mo ago
I like the German non-compete requirements (by law): max non-compete length is 24 months. Non-competes need to be reimbursed at half the last salary. The non-compete also needs to be specific to a real competition scenario (e.g. you taking
27.
by oezi 2mo ago
Even if transport is encrypted, the LLM computations will always be clear text, right?
28.
by oezi 2mo ago
I get the bearish case but it is easy to forget how much difference good vs bad politics can have and that politics isn't primarily a game of facts. It is an optimization problem under many unknowns and its history is littered with aca
29.
by oezi 2mo ago
Forecasting that the GPU build-out will reach 100 trillion USD in 2034 is wild (that's triple the US GDP in 8 years). And another 10x within 2 years. I am not sure where they believe that amount of capital could come from. It would req
30.
by oezi 2mo ago
Can you elaborate?

More ›