t_mann

9 ms·

91.
by t_mann 1y ago
For a lot of eg security-related applications where you want random numbers you don't want them to be publicly known. I wonder whether there are any risks to a naive approach of turning those numbers into private random numbers, eg ha
92.
by t_mann 1y ago
Does it? Could you explain which mechanism you suggest using? Because the main results from social choice theory about ranked choice voting that come to mind seem to be all about impossibility of fair elections (eg Arrow) or even paradoxica
93.
by t_mann 1y ago
I was kind of hoping this would be in the 'Dreambooth mold' of finetuning open weights models. I have used that with some success some ~2 years ago, does anyone know what improvements there have been in that direction since Dreamb
94.
by t_mann 1y ago
Windows 10 with WSL(2) is/was peak Windows for me. You could build stuff and edit MS Office documents in the same place. Sadly, it wasn't meant to last. I have no intention of giving W11 a try, not yet decided what I'll be us
95.
by t_mann 1y ago
> do adults really need friends? There's substantial evidence that it makes people live happier, longer and healthier lives, eg [0]. [0] https://mcpress.mayoclinic.org/healthy-aging/a-surprising-ke...
96.
by t_mann 2y ago
I don't get what the AI agent is for - the way I know of DocuSign is to use it to legally effectively sign documents such as contracts. Those would typically be prepared by (human) legal experts. I'm wondering what niche has docum
97.
by t_mann 2y ago
> Good luck with spreadsheets which have formulas in anothe language. Excel always shows the formula names in your system language. If happen to be on a computer set to Japanese and prefer English, just set it to that. Why does the whole
98.
by t_mann 2y ago
It is a 1-to-1 mapping of keywords. It's not the only choice of German words one could have used for those concepts, but neither are the English ones. For some people the particular choices made in German here will make more sense, f
99.
by t_mann 2y ago
I wonder whether LLM programming will make such projects more (easier to get assistance) or less (assistance might not be as good as in the original version) common.
100.
by t_mann 2y ago
An empirical study could start by quantifying model quality and available compute and plot them over time, see how they correlate, investigate whether there might be confounding factors,... in essence, putting numbers behind the qualitative
101.
by t_mann 2y ago
Has anyone empirically assessed the claims of the Bitter Lesson? The article may sound convincing, but ultimately it's just a few anecdotes. It seems to have a lot of 'cultural' impact in AI research, so it would be good to h
102.
by t_mann 2y ago
At first glance it seems that it not only has functional whitespace, but has overloaded it (used for grouping numerator and denominator or for pre- and suffixes). Also, it seems inconsistent in how it handles commands, some seem to be reser
103.
by t_mann 2y ago
Big LLM is too long as a name. We should agree on calling them BLLMs. Surely everyone is going to remember what the letters stand for.
104.
by t_mann 2y ago
> 174 Czech adults listened to 5 minutes of Māori... They were then tested on new audio clips from either Māori or Malay ... and asked to say if they were hearing the same language as before or not. The test phrases were acoustically fi
105.
by t_mann 2y ago
If the context implies that it's hand as the instrument used in fight . I don't see how that's relevant here.
106.
by t_mann 2y ago
To the OP, I say forget this advice. There's a headhunting company called Monster, arguably the worst possible name, especially for that business. Last time I checked they were doing fine. The truly worst thing is not being talked abou
107.
by t_mann 2y ago
AI is fundamentally much more of a danger to the fraudsters. Because they can only calibrate their obfuscation to today's tools. But the publications are set in stone and can be analyzed by tomorrow's tools. There are already sta
108.
by t_mann 2y ago
Is it? Take a look at the bot accounts filling up social media (the non-obvious ones). It wouldn't seem to hard to make one that makes 2am posts about '[next product] feels like real AGI' or tells stock analysts that their qu
109.
by t_mann 2y ago
What you should really aim for is companies that can win in the market. "Found European" instead of "Buy European" if you will. Right now might be a good time to try and reverse some of the brain drain into US tech.
110.
by t_mann 2y ago
I get why people might feel this impulse, but they should realize that it follows the same logic as tariffs. You do not gain by closing yourself off from the world. Having the choice among the best products in the world is prosperity. Wheth
111.
by t_mann 2y ago
Thanks. That does sound like it could be a pragmatic solution, if they go through with it.
112.
by t_mann 2y ago
The math also tells you what percentage of your wealth to bet: https://en.wikipedia.org/wiki/Kelly_criterion
113.
by t_mann 2y ago
Almost nothing of relevance can be done overnight, but the first step is always the most important one. I haven't heard of that comment by vdL yet, maybe they're ready to take it, but the proof is in the pudding.
114.
by t_mann 2y ago
Without knowing anything about the arrangement, I would assume that as far as these matters are concerned, whoever controls the launch button effectively owns the thing. Those things are made exactly for the kind of situations where treatie
115.
by t_mann 2y ago
The UK also has a nuclear arsenal [0]. A bit smaller than France's, but the reality is both pale in comparison to the US and Russia. Europe has no natural leader, it's an ensemble of what some would now call random countries that
116.
by t_mann 2y ago
The article is about stablecoins, which mainly live on smart-contract-enabled blockchains (ie, not Bitcoin). I don't think there are many (if any) major smart contract chains out there still using PoW, which I assume your concern is ab
117.
by t_mann 2y ago
Without knowing the detail, it's almost certainly a statistical model. Call it ML if you like. GDP nowcasting models take contamporaneously observable variables like foot traffic as features and GDP - which is observable only with a se
118.
by t_mann 2y ago
Hallucinations themselves are not even the greatest risk posed by LLMs. A much greater risk (in simple terms of probability times severity) I'd say is that chat bots can talk humans into harming themselves or others. Both of which have
119.
by t_mann 2y ago
I do remember it fondly, but I don't remember ever thinking any of what they show you was necessarily based on usage stats. Iirc there were pretty obvious editorial decisions like the front page having a Halloween theme around Hallowee
120.
by t_mann 2y ago
I get how that works on desktop, but on mobile, I can add a local file as a vault in Obsidian, but I don't think that file could be tracked by my cloud sync app. Does the Nextcloud app support that? Not sure how you use git here, could

More ›