Hallucinations are IMO a hard wall. They have gotten slightly better over the years but you still get random results that may or may not be true, or rather, are in a range between 0-100% true, depending on which part of the answer you look at.
OpenAI's o3 was SOTA, and valued by its users for its high performance on hard tasks - while also being an absolute hallucination monster due to one of OpenAI's RLVR oopsies. You'd never know whether it's brilliant or completely full of shit at any given moment in time. People still used o3 because it was well worth it.
So clearly, hallucinations do not stop AI usage - or even necessarily undermine AI performance.
And if the bar you have to clear is "human performance", rather than something like "SQL database", then the bar isn't that high. See: the notorious unreliability of eyewitness testimonies.
Humans avoid hallucinations better than LLMs do - not because they're fundamentally superior, but because they get a lot of meta-knowledge "for free" as a part of their training process.
LLMs get very little meta-knowledge in pre-training, and little skill in using what they have. Doesn't mean you can't train them to be more reliable - there are pipelines for that already. It just makes it hard.
I love your UDP packet flow tool, kudos for making that! I've always wondered how packets move through an OS. Also interesting how many gotos I see all over the place, even though everyone says "goto is the devil". Then again maybe this code was written long before that "proverb" came into existence.
If we're talking about state funding, that's not a problem. You just send a national to live in a residential area and then a team can proxy through that connection.
Commercial VPNs are relatively easy to block, because they use known IP ranges that companies can blacklist. But it's trivial to set up a private VPN with unique IPs such that VPN blocking becomes much less straightforward and much more resource intensive, for example by using traffic pattern analysis or behavioral fingerprinting.
I am currently adding passthrough mode to my game (DodgeALL) to spawn portals on your walls letting you dodge things coming out of the walls. Planned release within the next month (in time to enter in the Meta VR competition). A friend of mine made a game (Loop One: Done) that is Factorio-lite which lets you build infinite factories in your flat, which is loads of fun.
That's not quite true - when did you get your free Quest 1? Only January of this year did Meta officially stop allowing devs to support those devices which IMO is not nice, but probably necessary to put resources towards newer devices since it was extremely outdated and very hard to keep supporting. The Quest 1 launched in May 2019, so it got almost 6 years of updates and if you have one, you can still install older versions of existing apps that choose to support it (which admittedly is very rare). I shut off support for my game back in 2024 when they recommended it, since the device is less than half as powerful as the Quest 2, very few users still had one, and the Q1 was a hard target to hit performance-wise vs newer devices. If you spend $50 to get a Quest 2 you'll get a couple years of updates or even better, spend $299 to get a 3S which is an amazing piece of kit and will probably be supported for at least 5 more years since it just came out.
Forcing shorter answers will definitely reduce their quality. Every token an LLM generates is like a little bit of extra thinking time. Sometimes it needs to work up to an answer. If you end a response too quickly, such as by demanding one-word answers, it's much more likely to produce hallucinations.
This is how they do it on the Mars Trilogy after setting up a colony on Mars. Of course the trilogy is entirely fictional, but the idea sounds like it couldn't be worse than the current situation, where those with narcissistic/psycho/sociopath tendencies winning elections i.e. those who seek power and/or those who have proficient oration abilities, not necessarily those who are most capable.
reply