I think the key to successfully using modern LLM AI code is language phrasing and describing the problem and goal successfully in order to give the AI code as much information as it can that is relevant to the issue trying to be solved. As the saying goes about computers, 'garbage in'='garbage out' (GIGO). The same goes with AI. Information going in has to be as full and complete as possible in order to get the best answer out of LLM AI code. So 'stupid people' and chimpanzees finding AI unimaginably great? Uh, has anyone actually tested how chimpanzees interact with modern LLM AI code? GSU needs to update the Lana project so we can find out! https://news.gsu.edu/research-... Koko the gorilla was taught to sign, using ASL. Maybe someone set up a LLM AI to read ASL and then we can test the theory that gorillas would find modern LLM models 'unimaginably great'? But for 'stupid people' finding LLM AI satisfying? Uh, that seems at odds with the reality the person interacting with modern LLM AI needs to be able to elucidate problems in fairly high level language in order to get good answers out of it. And those kind of capabilities in stupid people, are kind of lacking. So pleasing stupid people, modern LLM AI code? Stick to Sora stupid people. That will keep you happy, uninteresting, and thinking current LLM AI is 'unimaginably great'. Just saying.

Another reason I guess I had no clue AWS was down that day, I was on Session all day long and a friend sent me a message thru Session that AWS was down that day. I was on Signal using Linux distros and Android phones, till I tried Session and found, wow, ok, so, no persistent connection to AWS, which means it stays online if AWS goes down (which at the time seemed improbable). Still, nice to have the alternative that stays connected no matter what.

Only one problem with that idea, my systems were up/online all day yesterday, connecting flawlessly, and I didn't even know AWS was down or having problems till late afternoon when I had a chance to browse the news. As in all things, diversification is the best disaster protection. When the DNS root servers are all hosted on AWS, then we're in trouble. But Cloudflare, Control-D, thru DoT, my systems connectivity was fine yesterday and my Tor relay node was up online all day long. AWS went down yesterday? Really? I couldn't tell.

Tell me who funds McKinsey. I take it they aren't big on advising their clients to use AI? I asked Perplexity a few days ago when some fossil fuel industry shill commented on a YouTube video something to the effect of 'coal will be the number one source of energy generation by 2050': "Perplexity, what's the biggest source of electricity generation in the world today?" Perplexity said that's not even true as of last week: "In summary, renewables, especially solar and wind, are now the largest and fastest-growing sources of electricity worldwide, with coal declining in use and fossil fuels overall slowly being replaced in the electricity mix."

As a person who had a parent do himself in because an undiagnosed medical condition in him caused him to waste away and go insane, what happened to that kid is terrible, but, I've worked with and known a few people in my life who took their own lives, back in the 1980s. My father went that way in the late 1990s. As much as we hate to admit it, people sometimes fall apart for physical reasons, or just depression, get stuck in a rut they can't escape, and decide to end their life. The best defense for AIs at discouraging people from abusing them to learn efficient ways to do themselves in? For adults, Idk if that's even achievable. Probably the best thing AI companies can do to prevent this from happening again? Do age requirements, to screen for minors asking such questions, and just tell them no at giving them the information, until they turn 18 in the U.S. After the age of 18? If an adult is determined to commit suicide, they're going to do it, unless someone intervenes to stop them. In my father's case, he tried once, got the 3 day 5150 hold, when he was released, a few months later, slipped out into his backyard late one night, found some high gauge extension cords, and a high tree branch, and you can guess the rest. Should we have been able to sue the psychiatrists and hospital that had him on anti-depressants and treated him? It was determined later he had undiagnosed celiac disease which caused his depression because he just was not testing positive for anti-bodies to gliadin proteins in his blood work. Should OpenAI pay for that kid's suicide? Idk. That's going to be an interesting case to follow. But, Imo, age-limit the answers to questions about how to commit self-harm. Also, maybe limit the answers to questions teens ask about committing mass harm using firearms or other weapons? Seems like a good idea.

I've never used my brain, ergo I'm not becoming any dumber. Personally making me stupid, AI? I've installed Alpaca and added 4 models (mostly Qwen2 and Llama code). In the last 6 months, using AI (Perplexity, iAsk.ai, Deepseek, and my local AI installs) I've solved and answered a LOT of questions and theories and designs, experiments, chemical mixtures, etc I'd been thinking about for decades, and most of which required lots of research to answer. Most of it was just, I'm too busy most of the time, so just did research on my own when I had the free time, and so these ideas went unresolved for decades. So, just in the last 6 months, answering a lot, figuring out a lot of designs, tuning a lot of Linux and Unix systems and code, etc. I would say, no, AI has not made me stupid if this is the result. However, I have heavy STEM education, and watch, understand, and enjoy films like this https://www.bbc.co.uk/programm... which actually bring me back to my college years, and I laugh at the feeling of nostalgia watching it. So no, AI is not making me any more 'stupid' than I already was xD

It's C level grade (2.0 GPA) average responses from the AIs I have tried. 50% of the time brilliant, 50% failing grade. One way I've found to improve their results is targeted, calibrated searches using startpage (to bypass the Google adsense algorithm's non sequitur search results). Here's a for-instance of how to use search engine operands and startpage to lead your AI of choice to better answers for writing a business letter (a complete business novice asking the letter writing question example) https://iask.ai/q/business-let... Doing it that way seems to bring the GPA of the AI answer results up to 3.0, B grade. Imho, AI currently is WIP. To think it's good enough now to replace likely more intelligent human employees? xD that's folly, will lead to all kinds of business losses and failures. Can't wait to see it happen lol

Not unless their intelligence levels go from 2.0 to 4.0 in the next decade. I've been using AIs lately to help me with optimizing fstabs and sysctl.conf settings for Linux and OpenWRT desktop OSes and routers. And the results are good, eventually, because I know enough to catch the BS the AIs have told me to use and correct it before their code recommendations crash my systems. I'd say they've contributed 50% brilliant code, 50% garbage. If I was a comp-sci instructor grading their performance, I'd give them a solid C right now. Also, i've caught some MAJOR errors AIs have made on questions I've put to them. So bad of errors it was easy for me to prove what they told me was demonstrably false from simple web searches. Let's not forget as humans we're all prone to built in human error. None of us are infallible. To be so would mean we would be Gods, not the mere mortals we are. And the coding for AI is coming from we error-prone humans. So at some point in the future, if companies replace workers with the current batch of solid C grade AI, and replace humans who would do better than the C grade AI replacements, then the error those companies will have made due to their human CEOs, will cause those companies to lose money, lose profits, and we all know where that downward spiral leads to ..

Well, I had an online AI analyze the Slashdot article, and it said, promising candidate, but needs verification https://iask.ai/q/mysterious-P... btw, I played devils advocate for a micro black hole or possibly white dwarf core, and the AI says based on the current evidence, nah, unlikely. More likely based on its weak thermal emissions, it's a large planet emitting residual thermal radiation based on its extreme distance from the Sun https://iask.ai/q/Planet-9-evi...

But then again, current facts often act as trend indicators of future events: "The Arctic is on fire: Siberian heat wave alarms scientists" https://phys.org/news/2020-06-... Never thought I'd see a smoggy sunset in Siberia, hmm

Tor. Ed had it right https://bit.ly/2XAtueJ I use a VPN HQ'ed in 14-eyes US which has about 60 overseas exits. I've checked most of them out, most come out of either US owned data-centers or DCs owned by companies in one of the 14-eyes partnership https://bit.ly/2XRZ46N So far about 1 out of 6 of the overseas exits is in non-14 eyes partner nations. So mainly I just run Tor out of my overseas VPN connections, figure it adds another layer of anonymity to Tor. Oh, it also helps to set the torrc to exclude 14-eyes nations as well as for entering or exiting them, within reason "{??}" to not break Tor functionality, as well as doing the same for obvious abuser type nations (Saudi Arabia, Iran, China, Russia, etc), other obvious violator nations of human rights, free speech, citizen repression, as well as spying, secret police excesses, etc. The result is quite good performance, security, and anonymity using Tor this way. Thanks Ed.

bc Trump goes back to Jan, 2010 on Twitter https://bit.ly/2YxDUbc (did my FF just count 73 instances of this word in his tweets? https://bit.ly/2JgN9aI :-/) anyone on /. or @ Twitter know? bc retroactive would b some kind of justice :-)

Try Startpage.com using Tor+Privoxy like this https://v.gd/5WAYUT & watch those Google bubble feelings melt away.