Path: csiph.com!weretis.net!feeder9.news.weretis.net!i2pn.org!i2pn2.org!.POSTED!not-for-mail From: D Newsgroups: comp.os.linux.misc Subject: Re: GIMP 3.0.0-RC1 Date: Tue, 11 Feb 2025 12:01:51 +0100 Organization: i2pn2 (i2pn.org) Message-ID: References: <655acbf6-05e5-69ff-8a44-9f7075aafa2e@example.net> <9f3323c0-5035-1172-e0d9-13ab2544b321@example.net> <_4mcnYIMzv0paDT6nZ2dnZfqnPWdnZ2d@earthlink.com> <9Vudndwpf_KGPDf6nZ2dnZfqn_idnZ2d@earthlink.com> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="8323328-1989619997-1739271713=:16712" Injection-Info: i2pn2.org; logging-data="3722016"; mail-complaints-to="usenet@i2pn2.org"; posting-account="w/4CleFT0XZ6XfSuRJzIySLIA6ECskkHxKUAYDZM66M"; In-Reply-To: <9Vudndwpf_KGPDf6nZ2dnZfqn_idnZ2d@earthlink.com> X-Spam-Checker-Version: SpamAssassin 4.0.0 Xref: csiph.com comp.os.linux.misc:65395 This message is in MIME format. The first part should be readable text, while the remaining parts are likely unreadable without MIME-aware tools. --8323328-1989619997-1739271713=:16712 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 8BIT On Mon, 10 Feb 2025, WokieSux282@ud0s4.net wrote: > On 2/10/25 4:50 PM, D wrote: >> >> >> On Mon, 10 Feb 2025, rbowman wrote: >> >>> And here we are now with AI.  Depending on how you count this is the third >>> cycle of promising the world, falling on your ass, and going back to the >>> drawing board for a decade or two. >> >> This is the truth! I wonder how spectacular the crash will be? What gives >> me >> hope is that all of the AI startups, sucking in billions are private. I >> hope >> that this will shield the stock market _somewhat_ from the worst effects of >> collapse. > > > Yesterday, kinda-broke France put 100 billion into 'AI'. > > Modern "AI" does have its uses, but also seems to have > limitations that will cap its utility. Not sure the > current approaches will ever graduate from "AI" to "EI". > > Of course MAYBE that's a good thing ... As good a sign as any, that maybe we are approaching the end of the hype cycle when the EU will throw away 200 billion EUR at building "AI-factories". A lot of professional con men will surely flock to suck at the public teat to see how much of that 200 billion they can get away with, without having to promise (or deliver) any kind of result. =/ > >> Today I heard about the first EB opportunity in storage. I remember when I >> sold >> my first PB deal, and I thought that was incredible. Today, the first EB >> opportunity. I hope I will get the chance to sell a EB solution. That would >> be >> nice! =D >> >>> But, it all started with a branch of psychology: how does that wetware >>> work? > > > You want "physiological psychology", not the freshman-friendly > Psych-101. > > To emulate the wetware you need something a lot like neural > networks. This may not be the best way - Nature made do with > what it had - but there may be some functional equations > hidden down in all that goo which can be used effectively. > NNs are 'getting better', after a LONG time, but LLMs were > the 'easier' faster way to get pretend IQ so that's where > the money went. > > I've always been impressed with things like baby cows > and horses and even elephants. They pop out and within > an hour are trotting around and acting all appropriate. > Where did human ancestors go wrong ??? Elephants are > smart large-brained critters too - not like 'born ready' > is limited to pinheaded things. NNs should aim at being > E-lephants when you press the "ON" button, with a lot > of How To and How To Do Better already burned in there. > > Current NNs ... well ... we're MISSING SOMETHING, some > kind of "it" ... and clearly it's not easy to identify. > > Hmmmm ... idea ... put the LLMs onto improving NNs :-) > Maybe LLMs don't have the tilt/bias/blindspot that keeps > us from seeing how to make NNs great ? LLMs might be > able to do 1000 years worth of experimental tweaks > over a long weekend and eval the results. > --8323328-1989619997-1739271713=:16712--