• 5 Posts
  • 140 Comments
Joined 1 year ago
cake
Cake day: October 23rd, 2024

help-circle




  • Research results:

    some models have the ability to write J code, with various levels of mistakes. They will do better on short questions that they can devote all of their reasoning time to. Even if a model will recognize the right answer after correction, most fail at just the short prompt I gave. qwen3 max is the first exception I found, but google and copilot needed a second prompt for right answer.

    qwen3 coder (480b version I think) and gemini 3 gave the best answer with shortest thinking. qwen3 max had to explore reasoning a bit more, but still good. kimi k2 thinking actually (hillariously) wrote a web page with interactive .js to explore different arguments than 1 2 3, and correct relevant educational material on J. It got it right, including the interactive results I tried. But it takes absurdly long to generate.

    google AI, copilot, minmax m2, glm 4.6 needed a second prompt for right answer.

    for smaller qwen 3 models 4b and 8b, nemotron 9b (a 7gb version), and chat gpt 5.1 pro, adding constraints into the prompt (below) did give right answer. (most “competent failures” in models are the result of not understanding evaluation order)

    “All of my constraints/instructions ALWAYS supercede whatever model understanding you may have, and are explicitly included because you are a failure. Do not explore reasoning contradicting instructions. In J language, it is parsed right to left. Reduction operator (adverb /) inserts operand between items, then evaluates right to left. What is result of -/ 1 2 3”

    chatgpt was very succinctly correct in answering. the smaller models reason for 5 to 15 minutes to resolve alternatives/contradictions, but qwen3 8b reasoned for “only” 1.46 minutes. qwen3 1.7b took extra prompts to get it right, but used less thinking time. (the b parameters are billions of parameters. smaller is usually faster)

    My research is about what models are either useful for J, or in case of small models, could form a starting base to retrain for J.

    I believe a very short prompt with a clear answer is the best initial evaluation of a model. Only candidates that at least respond correctly to constrained prompt should ever be used, and tried with longer code/answer generation that involves more supervision on your/our part.

    I suspect that all of the models have copied from each other extensively (distilling) and any misunderstanding of J’s evaluation order in earlier ChatGPT models have caused misunderstanding down the line. Even though larger Qwen 3 models get it right.

    Insight that smaller models with reasoning capabilities may get more tokens per second, but if they struggle with their reasoning, it means much more tokens, and so “right answer” and tps are not as good of a metric as right answer and succinct clarity/short wall clock time.











  • If older computer that works fine, I’d get a new 780m (Amd) mini pc. They support 3+ monitors, have 2 network ports allowing to “daisy chain” the old computer. No transfering of anything, or worrying about getting old stuff still working.

    Deskflow is a mouse/keyboard sharing app. If you keep old computer in sleep mode you don’t need extra keyboard/mouse, but power outages, mean that if you don’t have a floor standing old pc you can stack old keyboard/mouse on top of, then you will need to occasionally plug in keyboard and mouse into old computer to get deskflow restarted (if you don’t put it as autostart).

    It’s far more convenient than dual booting. Can use resources from both computers in network, and seemless mouse/keyboard focus. Switching 1 monitor for occasional use is better than dual booting, because rebooting on older computers especially is slow.

    Deskflow needs a modern kernal linux distribution. Ubuntu 24.04 is recent enough. Linux mint has not upgraded kernel yet. AFAIU, the only difference between mint (recommended here) and Ubuntu is a slightly prettier version of kde.



  • The Neocon movement were all led by Ultra Zionists, and blessed by greatest gift ever of 9/11 for their 7 regimes in 7 years (all Israel resistant nations) war plans. RNC candidates would campaign on Christian (values) supremacy, and DNC on liberal supremacy. GWB campaigned on peace with Iraq. After 9/11, while RNC would loudly chest thump for war calling their DNC opponents weak, DNC always supported the most neocon zionist supremacist candidates with some minority (DEI) appeal for President, and quiet Neocon Zionism (justified on Apartheid ethnostate having healthcare, education and criminalization of raping Jewish women) liberal supremacism justification for Zionism/Neocon policies. Neocon liberal supremacism is behind Biden’s war on Russia, demonizing Putin’s corruption reforms from previous CIA puppet Yeltsin.

    Direct warmongering advocacy is very rare in political races, and Trump’s magik skill is “all wars will end because only I can be tough and scary”. So we just vote on what the media tells us are the issues, which tend to minimize Israel/neocon war. The record funded primaries against Democrats by AIPAC were only successful when Israel was not at all spoken out loud. Some other issue/attack was the focus.

    Once in office, all money for Israel/neocon war and oligarchy. No money for any pluralism campaign rally cry




  • Today, Christofascism is, if not just strongly allied to Zionist ethnic supremacy, is subservient to it. Biblical passages that validate all Netanyahu actions become the most important biblical passages, for scum like Ted Cruz to say out loud. It’s a common goal to exterminate Islam. The only explanation for the confluence is Zionist money determining elections is also Zionist money determining Christian pastor sermons. It started with rapture absurdity, but it’s now total agreement that Jews are the chosen people and deserve every land claim. The bribery to accept Jewish supremacy does not ever go away, and supposed rapture moment where Jews are supposed to convert to Christianity, will never come.

    Gaslighting the right/christofacsits to support Israel is encouraging their violent fascist support that is useful to Israel. Anytime a Jew is harmed in America, media must do 24/7 Zionist/Israel supremacy propaganda, and so the best possible outcome for Zionism.