maeil

33 ms·

391.
by maeil 2y ago
By doing what Luigi did, until suddenly those who do have the power to change things start realizing that fixing things is a better option than having an angry mob.
392.
by maeil 2y ago
Even if we see 20 Luigi Mangione copycats in 2025, this isn't doing to start a violent revolution nor a civil war in the US. It would actually do the opposite - the great positive effects from those 20 Luigis would reduce the chance of
393.
by maeil 2y ago
This is dealing with them. It's just not dealing with them in a way that you like.
394.
by maeil 2y ago
The size isn't stated, not necessarily a given that it's as small as 1.5-Flash.
395.
by maeil 2y ago
I actually have built such a tool (two AIs, each with different capabilities), but still cringe at calling at agentic. Might just be an instinctive reflex.
396.
by maeil 2y ago
No, and they haven't been for at least half a year. Utterly optimized for by the providers. Nowadays if a model would be SotA for general use but not #1 on any of these benchmarks, I doubt they'd even release it.
397.
by maeil 2y ago
> i don’t think they need to win the on device market. The second Apple comes out with strong on-device AI - and it very much looks like they will - Google will have to respond on Android. They can't just sit and pray that e.g. Sams
398.
by maeil 2y ago
People act based on incentives and disincentives. As of 2024, there are big incentives to cause great damage to society at large for pure personal gain in a manner like this CEO, with no disincentives. This brings back a potential disincent
399.
by maeil 2y ago
That only shows today, and the keyword has already been removed for a while.
400.
by maeil 2y ago
To my understanding, as long as they're not using a VPN, yes. Now as another commenter pointed, if it's a website that the GFW blocks already anyway (e.g. Reddit), there's no point. So first thing to do would be to check if t
401.
by maeil 2y ago
To my knowledge, as much as this tactic is made fun of, it does work to prevent the page from being accessible from China without a solid VPN. Is there anyone with enough experience in China to confirm or deny this?
402.
by maeil 2y ago
> especially now, if only to use chatGPT. Aren't the Chinese models competitive? > thinks reddit style Tianamen copypasta would get a PRC netizen's computer to explode and get abducted by MSS Bit of a weird exaggeration. As
403.
by maeil 2y ago
There's no one-to-one comparison, but a decent one would be a drug cartel making and selling fentanyl. Sure, the orders come from civilization at large, I guess, as if no one were to buy the stuff then they wouldn't produce it. IG
404.
by maeil 2y ago
Potential rise in racist incidents must overrule accurate reporting on and research into what may have spawned a global pandemic directly killing 6+ million people (more than the Vietnam war and Korean war combined) and indirectly killing m
405.
by maeil 2y ago
It's still a choice. They don't do it when there's an article on the dangers of the rise of Neo-Nazism, or on the horrors of IS. "Now we interview Paul (32), who doesn't think it's dangerous at all, in fact he
406.
by maeil 2y ago
I'm late to the party, but expect Simon to still read this. > Anthropic’s decision to raise the price for Claude 3.5 Haiku isn’t looking great, given the continued competition at the bottom end of the market. Just a few months ago C
407.
by maeil 2y ago
> Was your status “Productive” most of the week? Awesome, you know you are improving. What is the next workout? Done. Right, but to us monkey-brained seeing a boolean "productive" isn't a great motivator, we need to see a
408.
by maeil 2y ago
I consider myself qualified enough to determine there are orders of magnitudes of difference between them, but not nearly enough to place them at an exact location on the spectrum. I'd have to devote months and work with plenty of othe
409.
by maeil 2y ago
Turns out China (or here, Russia) infiltrated the country, waged an enormous disinformation campaign and succeeded by getting their chosen candidate elected? "Tough luck, it's too late now, should just stand by and watch the count
410.
by maeil 2y ago
I think most of suggest that large-scale targeted foreign interference, and benefiting from it, should be illegal qnd have consequences which may include overturning results. And that's the case here, as it is in most countries. I don&
411.
by maeil 2y ago
Super helpful, going to look into those! I too only care about the fitness part, in an even narrower scope - literally only the usecase mentioned here, giving some VO2max-ish or other "stamina level" number (even if of questionabl
412.
by maeil 2y ago
> We are essentially all complicit Empty statement that shirks responsibility when some are complicit orders of magnitudes more than others. Even among those not directly involved in fossil fuel industries. This matters. In Nazi Germany
413.
by maeil 2y ago
We're hundreds of miles off from going too far down this road. This is taking one step on a road that spans the globe. It's about as relevant to bring this up now as it would have been to do so when people celebrated the death of
414.
by maeil 2y ago
Which watch do you use? The user you replied to has convinced me to get a smart watch as 1. I've had awful stamina since age 10 or birth 2. Stupid "number going up" gimmick generally works well at motivating me. But I'm
415.
by maeil 2y ago
No, hence we're now seeing cases like Brian Thompson's where citizens rightly take law enforcement into their own hands because the actual judiciary has been completely captured by the wealthy class.
416.
by maeil 2y ago
> IMO no big enterprise will adopt chatGPT unless it's all hosted in their cloud Plenty of big enterprises have been using OpenAI models for a good while now.
417.
by maeil 2y ago
1/2: Great points, thank you for educating me on this and renders my point largely moot, at least when compared to ICEs 3: I'm incredibly skeptical. We've heard this from every industry with enormous negative externalities. C
418.
by maeil 2y ago
> "An AI assistant was tasked with {task}. The relevant information for their task was {context}. Their answer is {answer}. The correct answer should be something like {ground truth}. Is their answer correct?" If you have a gro
419.
by maeil 2y ago
Yep, point being it should be part of the title, as that's the degree to which it impacts results, and the title is what has the biggest impact.
420.
by maeil 2y ago
> At scale, you are doing the same thing with humans too. LLMs seem to have an error rate similar to humans for the majority of simple, boring tasks, if not even a bit better since they don't get distracted and start copying and pas

More ›