I think the factorization trick should enable even bigger models on the corporate scale systems. These systems will need to be trained how to collect training sets. Too much work for humans to compile!
On May 4, 2023, at 3:39 PM, Roger Critchlow <r...@elf.org> wrote: Didn't read Cory's blog, though I'm still laughing at the blurb for Red Team Blues. But I read Geoffrey Hinton's interview with MIT Tech Review yesterday. https://www.technologyreview.com/2023/05/02/1072528/geoffrey-hinton-google-why-scared-ai It's not hype that chatgpt dazzled everyone with a model which is much smaller than a human brain, even though it took a fairly huge budget for OpenAI to build it. And I read this posting from an anonymous googler today via hackernews. https://www.semianalysis.com/p/google-we-have-no-moat-and-neither It's not hype that the open source community has rapidly figured out how to produce equally dazzling models with drastically smaller budgets of resources, and is continuing to iterate the process. -- rec -- On Thu, May 4, 2023 at 10:11 AM Gary Schiltz <g...@naturesvisualarts.com<mailto:g...@naturesvisualarts.com>> wrote: I love the graphic! I've had the misfortune of twice jumping on that roller coaster just before the Peak of Inflated Expectation - once for the AI boom/bust of the mid 1980s and once for the dotcom boom/bust of the late 1990s. Jumped on too late to make a killing, but didn't get too badly damaged by the Trough of Disillusionment either. On Thu, May 4, 2023 at 10:34 AM Steve Smith <sasm...@swcp.com<mailto:sasm...@swcp.com>> wrote: [https://miro.medium.com/v2/resize:fit:720/format:webp/1*aG2GG56d_C6DJnGzZn0d3Q.jpeg] https://doctorow.medium.com/the-ai-hype-bubble-is-the-new-crypto-hype-bubble-74e53028631e I *am* a fan of LLMs (not so much image generators) and blockchain (not so much crypto or NFTs) in their "best" uses (not that I or anyone else really knows what that is) in spite of my intrinsic neoLuddite affect. Nevertheless I think Doctorow in his usual acerbic and penetrating style really nails it well here IMO. I particularly appreciated his reference/quote to Emily Bender's "High on Supply" and "word/meaning conflation" in the sense of "don't mistake an accent for a personality" in the dating scene. A lot of my own contrarian commments on this forum come from resisting what Doctorow introduces (to me) as "CritiHype" (attributed to Lee Vinsel)... the feeling that some folks make a (a)vocation out of kneejerk criticism. It is much easier to *poke* at something than to *do* something worthy of being *poked at*. I appreciate that Doctorow doesn't seem to (by my fairly uncritical eye) engage in this much himself... which is why I was drawn into this article... I also very much appreciate his quote from Charlie Stross: corporations are Slow AIs, autonomous artificial lifeforms that consistently do the wrong thing even when the people who nominally run them try to steer them in better directions: https://media.ccc.de/v/34c3-9270-dude_you_broke_the_future I could go on quoting and excerpting and commenting on his whole article and the myriad links/references he offers up but will curb my enthusiasm and leave it to the astute FriAM readers to choose how much to indulge in. It was a pretty good antidote for my own AI-thusiasm driven by long chats with GPT4 (converging on being more like long sessions wandering through Wikipedia after the first 100 hours of engagement). -. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. . FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ -. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. . FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ -. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. . FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. . FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
