motoboi

This is what Banality of the Evil was all about. The idea that people who do evil things is different from people who do not is what allows those awful things to happen. Because that idea stops us from examining our own cognition for signs of conformity and groupthinking.

If we start recognizing that we too are capable of awful things all while believing we are doing good, we can start self examining our ideas, motivations and behavior, to catch ourselves before we embark on such grouptrips.

Nah, not like this. It's mostly brainwashing into thinking you are doing good things.

Banality of the evil. Hannah Arendt.

They probably link libraries written in C, given the scope of the pdf and image parsing.

Too huge attack surface. Good candidate for victim trust (who would not open a pdf file on an iPhone?).

So pdf and images are the best vector for iPhones.

Once you escape the parser you have to reach and escape the sandbox and from there reach the kernel.

> Instead, they chose hubris and self-interest.

Well, you accidentally described a top mathematician before age of 41.

Mathematicians should web search the arguments when artists were hit by the wave. Scientists in general and people that think humans know better.

This is probably the first time in history deep thinking is being replaced by a machine, but the overall arguments are the same in all cases. Graphical artists saw it first, then programmers, now mathematicians.

No, there is not stopping the wave. No, society won't probably be able to support the same number of mathematicians as before. But yes, there will still exist mathematicians after the wave break, just probably completely different than the old way.

And yes, some people will still try to do artisanal mathematics and say it's better, it probably is, if the artisanal mathematician in question is doing deep creative and out of the distribution work.

> At the same time, OpenAI had a disastrous dev day that led a lot of users to jump ship to Claude.

Can you elaborate on this?

How many cicles before pelicans become photoreallistic SVGs and them come back to stylistic artistic abstract renderings?

Also, that first tweet contains a commentary that, for me, explain a lot about the public misconception about how current AI works:

> wait, does this mean you guys are actively training svg generation? i was mostly amazed by it thinking the models doing better at it by itself due to becoming better at reasoning.

No, it's not better reasoning that drives current AI, it's better examples we have been giving them. Whole army of people generating content for the frontier labs.

Also, reasoning improving the out of distribution part, like computer use, tool use etc.

A nice side effect is that it transcribes in English an audio in Portuguese. And it actually make sense.

But the overall transcription quality is very low.

That's my experience too. sol-6.1 goes straight to solution, like it has done it 100 times before. Deepseek will explore and insecurely overthink like it's an intern made CEO.

The strategic advantage, by definition, is something you, as a company, don't know what it is. If you know, then people leaving your company take the information to competition and the it's not strategic nor advantage anymore.

So, given that you don't know what makes your company better then the competition, it's better to hide everything you can so you don't accidentally disclose useful information. God forbid competition discover your own strategic advantage before you do.

I can’t actually understand the incapacity of the richest country on the world to build infrastructure. Why would people be fighting over resources like that when you can just build a hell lot of wind, solar and gas plants along with batteries and whatnot.

From what we see on the TV is not that people in rural areas in the US are concerned with the environment. It’s just that they fear energy will become more expensive.

As a non American, I find this incomprehensible.

Just make it too expensive. Wanna buy one internet link per hour to download videos. Or maybe rotate between source IPs (google will block the whole ASN ip block in less than an hour). Or maybe pay people to download videos for you? They can download maybe 20 in a day, for a week, because tarpit and blocks.

It's a numbers game.

Before 2025 yes. Currently not. YouTube implemented severe restrictions to scraping and downloading of videos. You can get away with maybe 5 videos, after that you enter a tar pit.

If you continue, you need to log in.

If you continue, the account is blocked in less than 30 minutes.

If you try to use cloud machines you are blocked before the first video. If you try to use known proxies, there are already tar pitted or blocked because everyone is trying.

You forgot YouTube. The source of world model data. THAT is the mother of all moats, if they can survive until capable of taping into it.

that would probably be a great companion to 'Command and Control: Nuclear Weapons, the Damascus Accident, and the Illusion of Safety'.

New models are distill from the actual unrelease frontier models. They are just giving us better checkpoints.

Jensen is borderline comical on his instances of high GPU usage. OpenClaw? "let's do it, it's amazing!", selling GPUs to china "it's of uttermost importance, USA can't afford not selling GPUs to china because other will do", chinese firms burning GPU cycles like hell on american datacenters from openai and anthropic? "this is just competition, let's do this, this is amazing".

Laya, the open source model, it’s actually modernBERT based.

Is for when you want an AI to make a decision. If you have been using gpt or claude or open source models for that, than it’s a way cheaper alternative.

And if you have not been, it’s for when you have to extract the context from text. When you have numbers or fixed options, it’s just a matter of code.

So if you find yourself having to decide if a given user comment is a refund_request, that’s for that.

It’s not perfect, you still have to fine-tune (or calibrate) using examples you have (and keep those examples updated over time). But it’s way better than trying to parse text with regexes.

Have you tried Gemini on them?

It's actually just a matter of how much do you want to pay for the task. Obviously if the task it's hard, then also a matter of infinity.

already at azure foundry and copilot

I'm sorry to prolong this thread, but what I mean is: networks are just software layers, the operational systems, the driver code, the firewalls, switches, the other server, the sandboxes, the TPM. That and all security policies we put on them. All software layers waiting to suffer a buffer overflow.

this is nice, basically virtual threads for kubernetes.

You have an unreasonable trust in software layers.

If the machine doing tool call can reach via network the machine hosting the weights then it’s just a matter of time.

Maybe not the current models, maybe not this year. But even a almost perfectly aligned model will misbehave one day.

When I think of a GPU I think of an Nvidia rack kit. What do you think when you think of a GPU?

And what if they create a bot net with decentralized command and control and hack for nodes with GPU and nodes with compute?

The only way to kill that is making plugging AI accelerators on the internet a crime. Good luck air-gapping them.

I suppose you don’t read sci fi?