No, in case you wonder, we haven't changed our minds.
-
@benroyce @petrichaos @YurkshireLad @Vivaldi Such flawed attempts at dialectic on social media result only in the burden-of-argument tennis; see Schopenhauer. https://burdentennis.com/ PS <30 days to real actual Wimbledon with yellow balls.
@bms48 @benroyce @YurkshireLad @Vivaldi Off thread topic, but referencing the above link, I agree on most points ; notably the not-really intelligent AI - LLM is explicit imo: advanced code that can easily eat LARGE inputs then puke less output while requiring massive compute capacity
The more you scale, the less LLMs are relevant. Exponentialy inflating token consumption as context grows:
https://thenextweb.com/news/microsoft-claude-code-retreat-ai-costBut yeah one day, it would be faithful to the "AI" terminology
-
@petrichaos @YurkshireLad @Vivaldi
it infects social media
perfectionism, toxic idealism, purity
as an entitled basis to complain about every. goddamn. thing.
and always zero appreciation from these assholes shown, for the only thing we are ever going to get, from any entity in this world, on any topic:
progress
it's simply a basis for edgelord trolling on baseless malcontent
or it's a personality disorder: stunt vibing ego masturbation
it's a parasitical social interaction on us all
@benroyce @YurkshireLad @Vivaldi just ignore lol
The most pathetic are the ones requiring finished product but wouldn't contribute. "I'm not a coder" is not an excuse: join beta programs, support alpha builds, etc etc
Anyways.. -
@petrichaos @benroyce @YurkshireLad @Vivaldi So the weaponization of the LLM as a geared DNN is very real and with us. They might not do so on particular use of human language. Chomsky pointed out that they fall down as they can't differentiate impossible languages; those humans would not acquire or understand ever, but this does not hold for opcode streams intended purely to bring software to its knees. There are defensive counters. I am actively researching specific applications of such.
-
@petrichaos @benroyce @YurkshireLad @Vivaldi So the weaponization of the LLM as a geared DNN is very real and with us. They might not do so on particular use of human language. Chomsky pointed out that they fall down as they can't differentiate impossible languages; those humans would not acquire or understand ever, but this does not hold for opcode streams intended purely to bring software to its knees. There are defensive counters. I am actively researching specific applications of such.
@petrichaos @benroyce @YurkshireLad @Vivaldi It should go without saying that token economics will directly influence the outcome of what is essentially now a cybersecurity arms race. Both state and non state actors have keen interest in this. Genie cannot be put back in the bottle. Whether the effort to sloppify across the board were to succeed or not, there will still be the Gatling Gun of Exploitation -- for a price. I wonder what Dan Geer thinks of this, after "Cybersecurity as Realpolitik".
-
@bms48 @benroyce @YurkshireLad @Vivaldi Off thread topic, but referencing the above link, I agree on most points ; notably the not-really intelligent AI - LLM is explicit imo: advanced code that can easily eat LARGE inputs then puke less output while requiring massive compute capacity
The more you scale, the less LLMs are relevant. Exponentialy inflating token consumption as context grows:
https://thenextweb.com/news/microsoft-claude-code-retreat-ai-costBut yeah one day, it would be faithful to the "AI" terminology
@petrichaos @benroyce @YurkshireLad @Vivaldi Thing is, LLMs are really really good at exploit writing to trigger CVEs in cybersecurity. That's the brute force search space exploration approach right there; Takanen et al covers the what and why of fuzzing, but not the how of what orgs like AISLE are doing. But it's very compute expensive and they still rely on breadth first search strategies optimized for deductive "inference" where that word is used with a very specific ML context.
-
@bms48 @benroyce @YurkshireLad @Vivaldi Off thread topic, but referencing the above link, I agree on most points ; notably the not-really intelligent AI - LLM is explicit imo: advanced code that can easily eat LARGE inputs then puke less output while requiring massive compute capacity
The more you scale, the less LLMs are relevant. Exponentialy inflating token consumption as context grows:
https://thenextweb.com/news/microsoft-claude-code-retreat-ai-costBut yeah one day, it would be faithful to the "AI" terminology
@petrichaos @benroyce @YurkshireLad @Vivaldi Oh god you had to post a link with the "growth mindset" "business idiot" no1. ICkick I really did not need to see that... It does come down to slop vs token cost. Every crank of the one-eyed bandit gives you a hit. Even stochastic parrots occasionally hit gold. Its like holding UK premium bonds; stochastic investment. The brute force approach makes far more sense for offensive cybersecurity, but threat detection with patterns vs heuristics can scale.
-
@petrichaos @benroyce @YurkshireLad @Vivaldi Oh god you had to post a link with the "growth mindset" "business idiot" no1. ICkick I really did not need to see that... It does come down to slop vs token cost. Every crank of the one-eyed bandit gives you a hit. Even stochastic parrots occasionally hit gold. Its like holding UK premium bonds; stochastic investment. The brute force approach makes far more sense for offensive cybersecurity, but threat detection with patterns vs heuristics can scale.
@bms48 @benroyce @YurkshireLad @Vivaldi some AV providers already using LLMs fot the mentioned usecase for a while now, same applies for pentest, redteaming, etc etc.
While it's useful, that's imo more pattern recognition rather than intelligence in pure sens of the word. -
@bms48 @benroyce @YurkshireLad @Vivaldi some AV providers already using LLMs fot the mentioned usecase for a while now, same applies for pentest, redteaming, etc etc.
While it's useful, that's imo more pattern recognition rather than intelligence in pure sens of the word.@petrichaos @benroyce @YurkshireLad @Vivaldi Absolutely. It does not resemble the "AGI" that Clammy and Dario like to shine the media on about, which is pathetic. When one knows how the sausage is made...If I had a really cynical view of humanity, if I really didn't value people or even my own skills I'd just "Pivot to AI", do deep learning for money, I have the aptitude not inclination. I specialized in networking early on, did time at ICSI in Berkeley for it. Why screw myself over for money?
-
@petrichaos @benroyce @YurkshireLad @Vivaldi Absolutely. It does not resemble the "AGI" that Clammy and Dario like to shine the media on about, which is pathetic. When one knows how the sausage is made...If I had a really cynical view of humanity, if I really didn't value people or even my own skills I'd just "Pivot to AI", do deep learning for money, I have the aptitude not inclination. I specialized in networking early on, did time at ICSI in Berkeley for it. Why screw myself over for money?
@petrichaos @benroyce @YurkshireLad @Vivaldi I suppose the same ethical-moral non-dilemma does not recur for genuine psychopaths, e.g. a human with a propensity for killing people and enjoying it, and then doing it for money. But if I know why the "AI" is broken and can express that rationally, instead of believing in AGI/ASI fairy stories, what then? I am absolutely disgusted by what this whole thing has visited upon the world. Trying to destroy professions with highly specious arguments? aargh
-
@petrichaos @benroyce @YurkshireLad @Vivaldi Absolutely. It does not resemble the "AGI" that Clammy and Dario like to shine the media on about, which is pathetic. When one knows how the sausage is made...If I had a really cynical view of humanity, if I really didn't value people or even my own skills I'd just "Pivot to AI", do deep learning for money, I have the aptitude not inclination. I specialized in networking early on, did time at ICSI in Berkeley for it. Why screw myself over for money?
@bms48 @benroyce @YurkshireLad @Vivaldi tbh I'd say let whoever do whatever makes them happy/feeling useful/whatever else
-
@wafflesies @benroyce and you don't have to, but it would also be good not to spread misinformation around. We make no money whatsoever off Google, and we never have.
-
@Vivaldi I'll try it out if it stops using Chrome as a base, but it's good to see that some browsers aren't jumping on this bandwagon of doom.
@gudenau it's not Chrome, it's Chromium, and there's huge differences
-
@Vivaldi That's good to heard you won't cave into AI like the other browsers.
What about my Vivaldi cat ear logo change? I think it would be a hit!! :3
@catsalad I'm not mad about this customization
-
I wouldn’t mind someone else to go out to dinner with my friends
@MikeyMcFilms we won't tell your friends
-
@Vivaldi now please make the address+search bar work like Firefox and then never touch it again
@deutrino could you explain further on how you'd like your address bar to behave?
-
@Vivaldi That's good to heard you won't cave into AI like the other browsers.
What about my Vivaldi cat ear logo change? I think it would be a hit!! :3
-
-
-
@Vivaldi@social.vivaldi.net Stop baiting users into using your proprietary Chromium fork with your CEO stanceslop and instead do something useful like release the source code zo your users can have freedom.
@SuperDicq @Vivaldi Just the source code wouldn't do - it'd need to be the source code, validly licensed under a free license.
