Strong disagree. Having lived those times, it really really was different, and there are a bunch of reasons for it.

1. First, back then (90s, early 00s) there was very little financial incentive to participate in discussions. BBSs, IRC, forums etc. were mostly non commercial. People joined without any expectation of making a profit, just for "the fun" of it. And for something new, interesting, evolving. Way less perversion of topics for monetary gain.

2. People back then made a clear separation between being online and offline. We literally had the term IRL coined. So a lot of discussions were "in abstract" and much less prone to be taken literally or seriously. A lot less identity / ideology stuff as well. Having a clear separation made it easier to not confuse your real world self with your online persona. Having an idea debated wasn't about you / your identity.

3. Politics was much less divisive back then. There was political debate, but again a bit more "abstract" and theoretical. I'd say the moment when this changed was 2008s US presidential campaign. Until then the Internet was seen as "not important". It has changed a lot since then.

4. Entry barrier. This might sound elitist or disparaging, but it really was a thing back then. The people online were mostly tech inclined, or curious enough to learn. It was much more educational, and (linked to point 1 above) everyone wanted to learn the cool new thing, without any monetary incentives. Much more sharing of pure knowledge, helping out and so on. It of course changed over time, but the early days were really something beautiful. I have very fond memories.

I gave a simple counterargument to this. Since there are "countless" prior games, many of them released before genAI, your argument is pointless.

The simplest counterargument: since there are already tens of similar games out there, why didn't the previous authors, supposedly grass-fed genuine checkmark blood-through-their-veins humans didn't notice the other 9-8-7-6-5... games, and still released their own version? Maybe because it was still that they wanted the game out there? Maybe because originality really isn't that common? Maybe because each individual had their own idea and spin to it? Maybe because they wanted the game out as they made it?

Same for this author. How they made the game is irrelevant, and nitpicking the "originality" or anything else is silly. Something like this wasn't possible 3 years ago. Now it's possible. Deal with it, and stop trying to find ways to diminish it. It's a huge accomplishment any way you cut it.

10000% LLM fever dream...

Also this:

> The following production-ready Python processing block is integrated into the telemetry readback array to automatically run the five-point global integrity suite against output log paths:

at this point if you see "production-ready" in anything, it's Claude slop 100%. Also the script is hilarious.

The better analogy is phishing. Because that's what's happening here. The "prompt injection" attack is trying to "phish" the LLM into doing something unintended. That's how we should all comunicate it, as it matches better with what's happening. Unfortunately there aren't really good defences for it, as we all know from phishing "education" / "campaigns". Your best bet is to secure it in layers, try to have warnings (i.e. classification models) you try to secure the next step (i.e. capabilities based tool execution) and so on. But it's not foolproof and it should be communicated clearly.

I doubt that will be the case, because of the long tail problem. (same with self driving cars and other ML related problems).

In fact, we have counter-examples today. Newspapers (even reputable ones) can't get it right every time, despite the fact that they have both trained people and in theory they're setup to catch that w/ reporters - fact checkers - editors. And still, from time to time, they get it wrong. (and I'm not talking about purposefully getting it wrong, just honest mistakes.)

What will likely happen with a ruling like this is that the answers will be hedged and legalesed and muddied up the wazoo.

That's not how the deal was announced. You don't pay Bs / year for a licence to gemini to send them your data. You pay that to run it on your own hardware, in your own garden, so the data stays put.

I know the internet is always anti big companies, but this is likely a "not worth it for now, we'll eventually do it" effort from Apple. The EU AI act is a mess, and the effort to simply know what they have to do to comply with it is likely going to take armies of people (not devs) and a lot of time, as the OOP said.

And the saddest part about it, is that Apple has the money and resources to sink into this. Think about all the small players that don't. This is yet again a miss for the commission, with the end result being an insidious form of regulatory capture. It sucks for those of us running small companies. Oh well.

There's also 2-3 other trailing labs in MS, xAI and Meta. All of them are blundering behind, but at one point or the other they've been up there for some verticals as well.

I think this is good. Having one clear winner would be worse than this SotA of the week rotating thing they've got going on. For us as consumers anyway.

I think the more companies there are, the better. Having 3 top labs competing, with 2 more trailing is better for consumers than having a monopoly/duopoly in goog or goog vs. the world. There'll be pressure on innovation, cost, availability and so on.

You can make some educated guesses and find out some limits on inferencing cost by looking at 3rd party providers on platforms like openrouter. You can get some median cost /tok for a given model size. Then make some educated guesses on SotA model sizes, and you can get an estimate on pure cost of serving a model. Error bars and all that, of course. But still a range, with some limits.

I think they mean that specifically in working with 3rd party per-project / freelance designers you usually get a "first draft + one adjustment" price, then every modification costs more. Similar for small design shops. Prices aren't necessarily per hour, as you'd get from developers.

Also, a common misunderstanding of orbital mechanics (probably amplified by otherwise great cinematography, but poor physics depictions movie Gravity) is that after a collision things move to higher orbits and thus remain up there forever / change planes and affect other satellites. But that's not how it works, the orbit gets elongated, but the periapsis remains the same (or slightly lower), so the things / parts / pieces still re-enter the atmosphere. And the satellites are grouped in rings, with different inclinations, making it extremely hard to reach one from the other.

Also also, space is like really really big. Plenty of space (hah) to put lots of rings of satellites and coordinate between themselves up there. The operators are the first ones who care about it, and they're slowly improving the existing systems, in both tracking (and access to tracking) and automated collision avoidance. Having 10k sats up there makes you good at keeping them separated.

They seem pretty close, in both average and "best run" scores. And, in a highly verifiable domain, "best run" or pass@n is what you're looking for.

Anyway, here's what the oficial FAQ [1] says about the process, they are trying to remove them all, but not always succeeding.

> Streak detection is implemented in the Rubin data processing pipelines, and occurs in two main portions of the pipeline. First, pixels associated with detected streaks are identified in the image mask plane and excluded from contributing to the deep coadded images. Second, sources identified in regions associated with streaks or glint trains are flagged during image subtraction and not used to create new difference imaging objects. In addition, alerts are only sent for detected difference-image sources that do not coincide with known satellite sky locations.

> Signatures from satellites may appear in LSST data products despite the Rubin Data Management team’s best efforts. Difference imaging catalogs have some flag columns which are designed to indicate sources that may be affected by streaks or glints (e.g., look for columns with “streak” or “glint” in their name). The IAU CPS SatHub has developed NOIRLab-hosted tools that may be useful for scientists working with Rubin data products.

[1] - https://rubinobservatory.org/for-scientists/frequently-asked...

[1] - https://www.theatlantic.com/science/archive/2024/12/vera-rub...

> Because the real world is not a Hollywood movie.

One interesting thought experiment that I like to do is think about how many years you have to go back for this to be true. In this particular scenario, I think ~25 years is pretty much the sweet spot.

The Internet was beginning to take shape in the late 90s, early 2000s, and security was just beginning to be taken seriously, but it was still nascent. In that timeframe we had the first worms starting to appear, we had slammer, we had blaster, ssh had lots of exploits and so on.

It's not really far-fetched that a mythos equivalent "unit", working in the 2000s could really "take over the world". Especially one without the "safety" tuning. The Internet was really ripe for this in that timeframe, security wasn't up to par, and employing advance techniques that came later (in memory payloads, rootkits, etc) could make it pseudo-invisible to that era's detection tech. (reminder that traces of blaster were found on computers from a nuclear powerplant at that time).

The only question is would the trend continue? Meaning would a ~2050s "mythos" equivalent be able to do today what the one we have today could do in the 2000s. And if true, would that capability come before the 2050s? Could this be reached sooner, with say a dedicated offline DC somewhere where "mythos" could bang its tokens against the network and learn to exploit everything we have today, faster than 25 years? That's probably a bit of a stretch, but maybe not "hollywood" far fetched...

so the people vs. otis, the people vs. IBM608, and so on? Has it ever worked?