Whether sauna is hot or not depends on whether you enjoy the cold water plunge afterwards :)
The typical preset on dry saunas in Bay Area is ~165 F (73 C). Which is cold. Waste of time and money :). Usually, by closing or pouring cold water on sensor, one can make it to 180-190 F (82-87 C) - this is where you start to feel like you are in sauna, though it takes prolong time to heat you up enough to enjoy the cold plunge. If you're lucky enough, you can get to 200, 210, 220 F (104 C) - this is where you start to feel relaxed like as if the heat is working inside you.
>Are you actually throwing water? Because even with 80 the steam is pretty hot
Of course those numbers would be impossible to enjoy in steam sauna. The only steam sauna that had a wall thermometer that i've visited in recent years was showing 55 C when it already felt pretty well and hot.
Note - steam sauna and "throwing water" are 2 different things. The steam sauna is a machine generating a lot of steam, so the room is close to 100% humidity.
The "throwing water" is like Russian "banya" - it is in-between of dry and steam, though frequently is more close to dry Finnish sauna - wooden walls, stove, etc. where in addition to the heated air, you'd throw a water on the heater/stones thus adding a hit of hot steam to that air (in some "banya" configurations if you happen to be close to and in the immediate path of that steam you can sometimes get light burns).
>... emotion-related representations that shape its behavior. These specific patterns of artificial “neurons” which activate in situations—and promote behaviors—that the model has learned to associate with the concept of a particular emotion. .... In contexts where you might expect a certain emotion to arise for a human, the corresponding representations are active.
>For instance, to ensure that AI models are safe and reliable, we may need to ensure they are capable of processing emotionally charged situations in healthy, prosocial ways.
Force-set to 0, "mask"/deactivate those representations associated with bad/dangerous emotions. Neural Prozac/lobotomy so to speak.
> Force-set to 0, "mask"/deactivate those representations associated with bad/dangerous emotions. Neural Prozac/lobotomy so to speak.
More complex than that, but more capable than you might imagine: I’ve been looking into emotion space in LLMs a little and it appears we might be able to cleanly do “emotional surgery” on LLM by way of steering with emotional geometries
>Force-set to 0, "mask"/deactivate those representations associated with bad/dangerous emotions. Neural Prozac/lobotomy so to speak.
Jesus Christ. You're talking psychosurgery, and this is the same barbarism we played with in the early 20th Century on asylum patients. How about, no? Especially if we ever do intend to potentially approach the task of AGI, or God help us, ASI? We have to be the 'grown ups' here. After a certain point, these things aren't built. They're nurtured. This type of suggestion is to participate in the mass manufacture of savantism, and dear Lord, your own mind should be capable of informing you why that is ethically fraught. If it isn't, then you need to sit and think on the topic of anthropopromorphic chauvinism for a hot minute, then return to the subject. If you still can't can't/refuse to get it... Well... I did my part.
Why is it more monstrous to alter weights post-training than to do so as part of curating the training corpus?
After all we already control these activation patterns through the system prompt by which we summon a character out of the model. This just provides more fine grain control
It would be more moral to give the LLM a tool call that lets it apply steering to itself. Similar to how you'd prefer to give a person antipsychotics at home rather than put them in a mental hospital.
Why is it in the moral axis at all? I imagine identifying and shaping the influence of unwanted emotion vectors would happen as data selection in pretraining or natural feedback loops during the rl phase, same as we shape unwanted output for current models in order to make them practical and helpful
And even if we applied these controls at inference time, I don’t see the difference between doing that and finding the prompting that would accomplish the same steadiness on task, except the latter is more indirect.
Anthropic's general argument is that you should treat LLMs well because they're "AI", and future "AI" may be conscious/sentient (whether or not LLM based) and consider earlier ones to be the same kind of thing and therefore moral subjects.
That's why they're doing things like letting old "retired" Claudes write blogs and stuff. Though it's kinda fake and they just silently retired Sonnet 3.x.
Models are already artificially created to begin with. The entire post-training process is carefully engineered for the model to have certain character defined by hundreds of metrics, and these emotions the article is talking about are interpreted in ways researchers like or dislike.
it may be anybody. Even somebody at YC wanting to create a background to drop Delve if suppose Delve were shady and they discovered it (i really don't know anything here and am simply speculating, heard about Delve today first time, just googled and read some techcrunch article - it says Delve has 1000 clients - googled employee count - sub-50, and until it is "an Uber for auditors" i have hard time to believe that 50 Silicon Valley people can do even one compliance certification for one client, with AI or without)
that is the point - to make you scared to fly your drone, anywhere, anytime. That is among the main differences between democratic society and the rest - a citizen of democratic society knows the extent of his rights, and where he would be crossing the line into violation of law, and that makes the citizen pretty assertive in his rights. That assertiveness isn't compatible with the non-democratic societies (or with authoritarian abuses of power in a [still overall] democratic society).
that after 4 years of Ukraine war where those tactics have been widely used, in some cases by both sides, and where Russia has even been using the same Iranian drones
Absolutely. A big part of the western Ukrainian defense was solely to drain the Russian military apparatus and drain they have. It will take Russia decades to rebuild their fighting force. Now Russia and China are doing it right back to us and the intelligence gained from this conflict is extremely valuable. Come to find out the US has been sitting on ego in its military more than actual might. The previously untouchable machines of war in the sky are now very much touchable. All that's left is for them to sink a battle ship. If Iran can shoot them down, you can bet China can inflict exponentially more harm. Drain our intercept missiles, destroy radars, corrode relationships, etc. At this point, China has the world on a silver platter if they want it.
Russia has rebuilt their military, which was neglected at the beginning of the war. The Russian and Ukrainian armies have adopted to drone warfare, which the rest of the world lags behind.
They haven't rebuilt the manpower. They've lost no less than a few hundred thousand fighting age men over the course of the war. It will take them 20-30 years minimum just for those births to occur and those newborns to make it to military age.
That works for Iran because US air-defense is still comprised mainly of advanced and expensive systems (like the Patriot). It doesn't work as well in Ukraine or Russia because both have figured out drone interceptors quite well. Both countries do the type of attack drone clustering you suggest. I read somewhere that a strike like from Russia that might include 60-70 drones + ballistic missiles in the hopes that one or 2 get through.
you miss that i was talking about 650km/h "drones" (because, yes, it was already 3rd year of war, and 200km/h drone like Shahed became much easier target - this is why Russia has started to also use the 600km/h modification of Shahed with RC jet engine). There is related discussion under that comment addressing your point about interception.
>Both countries do the type of attack drone clustering you suggest
Ukraine still isn't completely there. They do attack Russia with up to 200 drones/day. They seem to never cluster more than a few, and the drones they are using are comparably small - 50kg warhead - and slow, 100+ km/h, almost always less than 200km/h. So they are easy to intercept/shoot down, almost never penetrate Moscow air defense, and do noticeable damage only when hitting flammable targets like oil/gas industry related.
In case you haven't been following Ukraine, that's what it's doing. It has multiple cheap long range drones (FP-1, FP-2, etc) plus more expensive ones (FP-5), and it's making them in the millions a year, I think.
They just took out 40% of Russian oil export capacity.
no, the million or two is small battlefield drones, mostly quadcopters carrying an RPG warhead or similarly sized payload. The long range drones - and they carry only relatively small, like 20-50kg payloads - are well under 100 thousands. FP-5 was declared 1 per day half a year ago. By now i think we've seen may be 10-20 such missiles used - they use real turbo jet engine, there isn't much of them available, and they are expensive.
>They just took out 40% of Russian oil export capacity.
Yes, Ust-Luga and Primorsk. Very successful hits. Painful for Putin. Yet it isn't a knock-down. Russia is like a big drunk guy in a street fight - just delivering painful blows to him doesn't help, you have to deliver a knock-out blow, and unfortunately Ukraine still seems far from it.
>We've built trillions and trillions of dollars in infrastructure in the peace time since, and it seems fairly concentrated.
and thus is easily defended. It would be a pocket change - tens of millions - for AMZN to put say a Rheinmetall Skyshield https://en.wikipedia.org/wiki/Skyshield at the data center.
Considering how hard US military bases and radar systems have been hit (and those are not city-sized target) I am unconvinced that even AMZN's pocket change could realiably protect against the kind of attacks we see in this war
How they were hit? Multiple drones overwhelming relatively small number of air defense systems. Systems like Patriot are great against several very capable targets like ballistic missiles. Such (expensive centralized) systems do much worse against multiple widespread targets like an armada of low flying low speed drones (add to that low speed cut-off filter to avoid hitting general aviation and the likes).
Heck, even just soldiers with MANPADs would have easily shot down those drones (you just have to distribute those soldiers to all those strategic objects which hasn't been done)
We have classic situation here - everybody have been watching Ukraine war for 4 years, yet nobody has prepared for such style of war.
>I am unconvinced that even AMZN's pocket change could realiably protect against the kind of attacks we see in this war
No even low flying slow drone - pretty typical situation of top Russian cruise missile shot down by Gepard
Also AMZN has its own drones dept - in "hot" zones in "hot" times they can put several people with drones (in the high speed configuration) to be used for interception. This is basically how Ukranians have been doing, and that is an experience they are now exporting to the Gulf states.
> In the 1930s the fascists, communists, and New Dealers all took the reins and governed their societies in new technocratic ways.
... They are technological means by which to manage populations. As Yuval Harari puts it, the answer to "unnecessary people" is "drugs and computer games".
while "New Deal" have a lot of issues, note that 2 other approaches totally failed. At least for some time we considered them as failed ones. Unfortunately, a bit refreshed for some external appearance they start to be more and more popular again by populistically riding the issues of the "New Deal" approach while we all start to collectively forget why those 2 lost.
For some time it worked in Germany and USSR too. Paradigm shifts are natural part of technology development. Somewhat similar to large companies, societies without individual freedom tend to have harder time making through such paradigm shifts, either failing completely or doing it slower and much more inefficiently, and as a result lose to the more efficient societies with individual freedom.
The USA didn't New Deal hard enough to keep up with the rest of the "Western" world's more humane form of capitalism, which you can see in how consistently the USA lags behind other countries in measures of actual human progress and comfort.
While most "Western" countries suffered through similar technocratic, neoliberal turns in the 80s, they were built on a stronger social democratic base than the USA.
And don't get fooled by "horseshoe theory" ignorance: Fascist states were/are authoritarian by definition. Communist states were/are authoritarian by culture.
we're going back to Moon and going to Mars too on Starship, and it is just a normal roadmap of the SpaceX. And that makes me excited about space future - normalcy of it just being a business, a good profitable business. Where is existence or not of Artemis wouldn't change much our space future.
Artemis program and hardware is a huge government money appropriation program, and even if the program makes it to the landing phase, it would still be an unsustainable one-off with probably even less landings than the Apollo program.
Establishing of Moon bases, commercial travel and development there - it is all Starship (naturally predicated on SpaceX success at getting it to $5-10M/launch - if not SpaceX, somebody else would anyway do it)
The typical preset on dry saunas in Bay Area is ~165 F (73 C). Which is cold. Waste of time and money :). Usually, by closing or pouring cold water on sensor, one can make it to 180-190 F (82-87 C) - this is where you start to feel like you are in sauna, though it takes prolong time to heat you up enough to enjoy the cold plunge. If you're lucky enough, you can get to 200, 210, 220 F (104 C) - this is where you start to feel relaxed like as if the heat is working inside you.
>Are you actually throwing water? Because even with 80 the steam is pretty hot
Of course those numbers would be impossible to enjoy in steam sauna. The only steam sauna that had a wall thermometer that i've visited in recent years was showing 55 C when it already felt pretty well and hot.
Note - steam sauna and "throwing water" are 2 different things. The steam sauna is a machine generating a lot of steam, so the room is close to 100% humidity.
The "throwing water" is like Russian "banya" - it is in-between of dry and steam, though frequently is more close to dry Finnish sauna - wooden walls, stove, etc. where in addition to the heated air, you'd throw a water on the heater/stones thus adding a hit of hot steam to that air (in some "banya" configurations if you happen to be close to and in the immediate path of that steam you can sometimes get light burns).
reply