0 0 0 On Thursday, 24 April 2025 at 17:41:28 UTC+3 John Clark wrote:
> > > On Thu, Apr 24, 2025 at 10:30 AM 'Cosmin Visan' via Everything List < > [email protected]> wrote: > > > *> Given that AI has 0 intelligence...* > > > *... and thus just before the last human being in existence was vaporized > he turned to Mr. Jupiter Brain and said "I still think I'm smarter than > you". * > > *John K Clark See what's on my new list at Extropolis > <https://groups.google.com/g/extropolis>* > 243 > > > > >> >> >> On Thursday, 24 April 2025 at 15:40:29 UTC+3 John Clark wrote: >> >>> *The AI futures project has updated their forecast ( AI 2027 >>> <https://ai-2027.com/?ending=slowdown#section-slowdown-0> ) **that came >>> out about a month ago, things are moving so fast frequent updates are >>> necessary: * >>> >>> *Beyond The Last Horizon* >>> <https://blog.ai-futures.org/p/beyond-the-last-horizon?utm_source=substack&utm_medium=email> >>> >>> [image: image.png] >>> *Progress is so fast that just a few days after the updated report there >>> was enough information to make the graph seen above; the AI project people >>> conclude from it that "OpenAI’s newest models’ time horizons seem to be on >>> the faster curve we predicted, rather than the slower seven-month doubling >>> time". Keep in mind that the above graph is an exponential graph, so a >>> straight line means exponential growth, and anything above that line means >>> super exponential growth.* >>> >>> *The authors conclude that today "AIs are able to do quick tasks, but >>> not long tasks. But they’re getting better at long tasks! So far it looks >>> like they can do tasks of about 15 - 60 minutes, and that number doubles >>> somewhere between every 3 - 7 months. These numbers are most applicable to >>> coding, and uncertainty increases the further we go from that domain. AIs >>> seem on track to master coding before they master any other comparatively >>> economically important sector. That's interesting because if there’s an >>> intelligence explosion, superhuman coder AIs are probably when it starts in >>> earnest. Our best guess is AIs will become as good or better than the very >>> best human computer programmers sometime in 2027, although with big error >>> bars and multiple asterisks. We think these human-level coders will be >>> enough to start the intelligence explosion, which is why we place it in >>> 2027 or 2028."* >>> >>> *It looks like Ray Kurzweil's prediction that the Singularity will >>> happen in 2045 is way too conservative.* >>> >>> *And to think, some people are still getting all hot and bothered over >>> trivialities like illegal immigration and transsexual men in women's >>> sports. * >>> >>> *John K Clark See what's on my new list at Extropolis >>> <https://groups.google.com/g/extropolis>* >>> akf >>> >> >> -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/everything-list/8ab02972-7074-48c1-b07e-c1cb5ed5e55an%40googlegroups.com.
