For anyone not familiar with Fujitsu's CPUs, worth noting their overall long history with HPC, and more recently, ARM. FugakuNEXT is "NEXT" because the current Fugaku (#1 on TOP500 for a time, and still in the top 10) also used an ARM CPU from Fujitsu.
A starting note: I don't disagree with you (about systemic issues), but I want to explain what I understand as the perspective you are responding to.
A "scapegoat" is someone who is incorrectly blamed for someone else's errors or sins. The perspective you're responding to is this: They built the system, they run the system, they have continuously warned "This system is dangerous!", and yet persisted. That is not being incorrectly blamed, not being a scapegoat, and instead is a collaborator.
So I think you mean to ask: "Why do people focus so much on finding someone to blame?" It's not merely semantic, because the answer to that is more straightforward: Consistent accountability is a major factor in deterring bad behavior. It is not the only factor, but it is a major one.
That is my Steel Man understanding of the people searching for individual blame.
This is a good comment and I'd like to add that I sometimes am astonished how many comments on HN are on the topic of liability or blame for this or that. I see liability as a 2nd or even 3rd order phenomenon of most of the subjects under discussion, but for whatever reason, liability always seems to come up. Maybe because it's sort of easy to reason about? Thirst for justice? Natural reaction to feeling disempowered? Whatever it is, I'd like more focus on the actual topic.
Sometimes in "normalization of deviance" situations, there isn't anyone specifically to blame. I wouldn't make an assumption that you can find anyone particularly blameworthy without doing an investigation first.
The rate at which these "labs" are creating these incidents is simply staggering.
Imagine we made nuclear weapons a private industry, had CEOs bragging how they have enough warheads to blow the Earth to smithereens, and then they "accidentally" nuked three cities over a short period of time each, saying they lost control, or rather couldn't contain their semi-autonomous weapon. All somehow managing to turn the PR around from their abject incompetence and towards SciFi visions of mankind hunted by self-replicating bombs.
There is also a disconnect between "customers" and "players". I am under the impression that RTS games have proven harder to monetize in an ongoing manner, or at least harder relative to the effort required to build and maintain them. That doesn't mean there isn't a solution, and this is just one person's limited perspective... but it seems to be the case thus far.
Hm. That's a situational rule. In this case it makes it funnier. A person could write a whole encyclopedia of lore tracing Kallax back to its mythical roots.
At least in the US, and as of a few years ago (when last I tested), address verification can also be a bit flaky. Zipcode/postalcode is solid, but street address matches don't reliably account for the various ways people can colloquially write them. It's not awful, just not good. We should re-test it some time, but as of a few years ago, street address checks were not a major improvement in fraud prevention compared to just Zipcode and CVV checks.
Address verification only ever checks the numbers in the address, nothing else. You can pretty much skip all the words that don't spell numbers and it should be the same iirc.
I read a prediction back in 2023 about there being some epoch date, where any data after it is untrustworthy for training, as it might result in a sort of fixed point, with LLMs training on LLM-generated content rather than human-generated content.
In truth, I don't think that's quite how it works, but I wouldn't be surprised if teams who scrape data for training apply weights in favor of data that is known to be generated prior to the GPT-3 era.
The people who curate training data use all kinds of tricks, yes. They also constantly try new ideas and see what helps.
Btw, learning from your own generated data isn't necessarily bad, if you have an external source of validation. See how AlphaGo learned from self play for a simple example.
I'd say at least 99% of all new youtube videos are AI generated video, whether taking snippets from other videos and voiceovering them, or just making entire fake personalities up from scratch. That number will raise to 99.99% soon.
There is no veracity to these videos, and within a year I'd say no one will ever be able to detect AI or not. This is exasperated by the fact that normal, real productions are being AI edited, and metadata or other non-visible means to detect AI or not will be meaningless.
Bluntly put, most people currently get their 'news' from Tiktok and Youtube. When I discuss current events with random, not-tech people, they all "believe" that if it's video, and on youtube or tiktok? It's real.
I cannot express how much the average Joe thinks, in their mind, "Something published on the Internet" is high in trust. I'm verbose, I talk to a lot of people, and there is an inverse relationship to computing skill, and Internet trust.
And using a computer to "do things" is not "computing skill", any more than driving makes one a mechanic, or ensures you understand a car engine.
I guess my point is, I suspect that the continued degradation of reporting, and news, conjoined with alternate sources being 'news sites', will continue to the point that "reliable source" is non-existent.
For example, imagine if every single post from a phone, such as "I see troops in City X' are seen as almost certainly faked. Or even, "I don't see troops in City X'. And if you think of it, certain sectors of society preferred the curated, more controlled news of the past. There may be attempts to purposefully destroy the credibility and believability of every and all independents on any platform.
Create a personality, have it trusted for months or years, and then show that its data was false. Doing this is easy with the tools today, doing it by creating one million such personalities with the same goal, is just as easy.
The end of truth for all is here now.
There is no point of trust now on the Internet. The cut off isn't just for AI, and training.
>I... am bearish on non technical people building software.
I can't bring myself to be bullish or bearish in such broad strokes.
Over the years, I have met and worked with a number of people who were "non-technical", but who nonetheless built their own internal rough software tools with a cobbled-together combination of spreadsheets and SaaS services, connected by something like Zapier. They are a particular kind of person, but I would not call all of them them "technical". They are experienced in their business domain, are detail-oriented, and are persistent in making improvements by any means they could.
I have seen those people be supercharged by AI app builders like Replit and Lovable, even if I don't personally have a need for either.
This might be a matter of definitions: Is such a person "technical" because they value systematizing human processes and put effort into it? Or is such a person "non-technical" because they are not experienced in software engineering best practices, or programming in the conventional sense?
In addition to the nature of the user, I suspect the fulcrum that success and failure rests on for these kinds of apps and tools is long-term maintenance. How frequent? How involved? How breaking are the changes? On this topic, there is a quote I think about a lot (paraphrased, and I forget who said it, or I would credit them):
>AI app builders mean that anyone can experience the joy of building software, but also that they will experience the joy of maintaining software.
I have seen and worked with a number of “non technical” people ship apps and automations over the last year that are making real money so it’s past the point where you can be bearish. Sure these people are probably just a class in themselves and not “the average person” but they never would have done this if they had to write code by hand
I mean 200k/MRR for one company most of these are not pure tech but people in a domain (the 200k one, while I don’t feel comfortable outright stating it is a car buying application you could likely find with a cursory search) who use tech to scale. So I have worked with lawyers, therapists, people from other countries etc to bring apps to that market using their domain/market knowledge as an edge.
Every time I peek at the code from something by Mr. Endoh, I end up down a rabbit hole and learn something. I look forward to what this one brings. Probably looking into this Piet language: https://www.dangermouse.net/esoteric/piet.html
Another awe-inspiring quine of his, from about a decade ago, is the 128-language ouroboros quine:
The last one is astonishing: "Okay, this one is really wild. Piet J. (yes, that's his real name) was browsing art in a small gallery and saw a work which reminded him of a Piet program. He spoke to the artist, who claimed to know nothing about the language. Piet took a photo of the artwork (left), converted it into a clean image file using close colours from the Piet palette (right), and tried running it.
reply