marcyb5st

With the rank and file I am not 100% sure. I mean, they get paid a fraction of execs and potentially don't have the whole picture.

Additionally, it is really hard to draw a line. For instance, I can develop some logic that could take a screenshot of your phone. It is perfectly fine for debug builds or with explicit user consent while using an app to provide some additional context on some issue. The same logic used to spy on users is not.

I "enable" them, but I won't consider myself responsible or even remotely involved.

Yeah, fully agree. Fines for companies bigger than a certain threshold should be percentages of their global annual income and potentially include jail time for the decision makers that enabled them.

As long as these are the incentives companies/execs will continue to do this shit

To be fair, the referendums for Nuclear were so poorly timed. Just after Chernobyl and Fukushima.

I don't know if this was a deliberate choice of the governments leading the country those years (the tinfoil hat in me can see big oil pushing for those referendums at those particular times) or pure bad luck. Anyway, public opinion was highly biased which led to "No" winning with fair margins. With that said, I believe NIMBY is a contributing factor and not the whole explanation.

Ok, I am having troubles imagining that a one trick pony company like OpenAI will generate 350B of revenue. It's in the same ballpark as today's Google that has Search, YouTube, GCP, Android, Chrome, Maps, Gemini and probably more stuff I forget.

> Iran can't have a nuclear weapon. It's an intolerable premise and if the price to be paid is global oil catastrophe it's a price well worth paying

Says you.

All other nuclear nations behave (even Pakistan and India, which are at each other throat all the time). What makes you think that Iran won't? It will also have Russia and China keep a lid on them, same as North Korea. And I don't buy the "Iran is different" gospel. North Korea and China have completely different mindsets than the western world and still play ball.

Additionally, if Iran really is that adamant in getting nukes, I can see them asking some enriched material to Russia as a payment to the 100s if not 1000s of drones they are giving them. They share a "land" border through the Caspian sea and smuggling a few 10s of kilos of enriched material would be relatively easy.

Finally, it is not a good proposition if the hypotethical benefits will be reaped by US/Israel while the suffering is globally spread.

Yeah, it is and the math get's really counterintuitive to grasp since humans (at least me) have trouble thinking in exponential terms.

If you accelerate at 1g constantly for 1y you travel 0.5 light years. You do that for 10.5 years and you reach the center of the milky way. You do that for another 4 (~14 total) years and you are in the Andromeda galaxy, and you do that for another 10 years (~24 years total) and you reach what today is considered the edge of the observable universe.

By the time you get there you are basically traveling at a rounding error from C.

I really hope they do and slap them with the 4% of their total global annual turnover.

Educated or not, people that have been using the same system for many years will have a tough time IMHO. Looking at my mom and dad interactions with any UI, it seems to me they don't really read what that menu is about, they just use "muscle memory" to navigate it. So I expect a drop in efficiency until that memory is rebuilt.

Do the per Capita calculation. Why can US and Canadian folks emit 4x as much as the average Chinese citizen? Do you believe you have the right to do so only because you are born in the West?

I believe that if it wasn't producing shit for US its emissions would be vastly lower.

Never seen anything of the sort apart from the Tsunamis in Japan and Indian ocean.

This event is the "big brother" of what happened in Italy in 2022 with the Marmolada glacier collapse which was tiny compared to this. Additionally, the orography in Tibet/Nepal is much more dramatic (higher elevations, steeper slopes, narrower valleys).

I am really astonished by the videos circulating online.

Time to become a bidet exporter

Just did my internet contract for my apartment in rural Italy (a town of around 2k people in the Verona province). 1Gb symmetric is indeed 25EUR/month. I added a few stuff on top like static IPv4 and reached 32EUR/month.

Time travel breaks causality if you assume Copenhagen interpretation of quantum mechanics. Multi-world interpretation allows it as you just end up in a self-consistent "branch" and you would be stuck there.

Like you can travel back in time and kill one of your ancestors before he/she had children. In that branch you wouldn't be born, but since you come from another branch the system remain consistent.

(If you are interested look at David Deutsch’s quantum model of Closed Timelike Curves).

I think it is subjective. I prefer being left alone and having to talk to the staff only when I need something.

Being interrupted every 2 minutes with a "Is everything ok?" is annoying.

But if the AMOC collapses the drought you talk about will become the a year round problem. Climate models predict a drop in rainfall between 20 and 50% depending on the region. So agriculture will be fucked.

And the US will be hit by super hurricanes since all the moisture/hot water will linger there supercharging them.

My understanding is that watermarking in prose is basically a bias when sampling tokens. For a system that knows the average probability for each possible token in the LLM vocabulary it is possbile to quantify said bias given enough text.

For a human that doesn't reason in tokens and therefore doesn't know anything about their probability distribution, it should be impossible to tell. Relying on fancy words/constructs within sentences should not give you any signal as well, since you don't know if the the prompt included instructions for that.

Ah, so I am not the only one struggling with deciphering Opus writing style. At times I feel dumb as a rock because I read the same passage like 5 times and I still don't get it.

I was reading somewhere that we are already fucked even if we were to completely halt greenhouse emissions today. Apparently we kickstarted some positive feedback loop (I think it was the thawing of the permafrost in Siberia/Canada) which releases enormous quantities of Methane and that with the higher temperatures we already reached we decreased Earth albedo as there is less snow/ice reflecting light into space.

I really hope that particular article was wrong, because the alternative is that we have to resort to geo engineering. Considering that the governments can't really coordinate in preventing this, I don't see them coordinate in fixing it, especially because the scale and costs of the "fixes" are astronomical.

Honestly, despite having a wife from the US (not an US citizen anymore), I agree with you.

It is really weird to me that such a thing as lobbying can be legal. Or the fact that you can donate to an election campaign so that they have more money to influence people.

And it is even more surprising that the vast majority of US people don't even question these things. I have the theory that American exceptionalism hinders their critical thinking when it comes to these kind of subjects. Ie "If we decided to do it like this, it means is the best way possible".

On top of that, isn't Bending Spoons famous for enshittify anything they buy to squeeze every single cent they can? Meaning that their fame would push even more companies/people to do a vibecoded replacement with just the features they need.

You literally made me Laughing out Loud :D

The exact training run is basically impossible anyway. Randomness plays a role. Even if you fix your RNG seed, in a distributed training scenario like this one some weight updates might come at different times and be included in different update steps. Should have minimal impact on the final outcome, but would still be a different model as some of the weights will differ in the end.

Apparently that is what Taalas did though. Not an hardware person, so take this with a pinch of salt

No, if you burn your model into the silicon you don't need memory as the output of a layer flows through circuitry directly into the next one. No I/O to memory of any kind. You still need a bit of memory for in flight answers, but that's it.

You don't need memory because the output of the previous layer flows directly into the input gates of the next. So no I/O to/from HBM or similar

You still need some memory for the context, in flight answers, ... but not for the model weights and for the output of the intermediate layers. I found taalas demo here: https://chatjimmy.ai/

I believe part of the LLMOps (I don't like the term, but it is what it is) should be building a failover plan with proper testing that check tools trajectories and such. If you have these then you can sort the good enough models from cheaper to more expensive and have the failover you mentioned.

I saw people bulding a mapping of model->{{prompts}, {tools descriptions}, ...}, but that, to me, it feels extreme. I believe it is the model that needs to adapt to your prompts after a certain point. Models that fail to do so won't get our api requests as they will be out of the failoever roster.

Good thinking. Didn't think of that approach. Gonna start sending huge walls of text to DOS them then if this thing lands

How do they scan e2e encrypted messages? Will they force apps/OSes to have master keys/institutional backdoors to have access to the private keys?

It is the difference between giving an LLM an epic and say "You figure it out" and giving the single tasks' breakdown you envisioned and build incrementally on top of it.

With the latter you can, for example, say "Wait, this should be an interface because later on we need different concrete implementations". With the former, the agent doesn't do that, gets to the point where you actually need the flexibility interfaces give you and refactors everything to handle that. That is at least 2x the work/tokens. Multiply this for all the decision points you have to do to deliver a big piece of work and you have your bagillion tokens consumed.

If you like hiking and tough, but rewarding, trails, consider the one to the base camp of Everest, or the roundtrip around the Annapurna. I did both and they are truly amazing without the craziness that has become climbing those Mountains. When I reached the basecamp it really felt like being in a weird Venice or Florence, with "tourist groups" and their guides. Crazy stuff.

Especially the Annapurna one, you still climb up to 5+k meters in altitude, and seeing the actual mountain still towering over you is crazy once you realize.