I found a very interesting article about when the AI intelligence explosion will occur it's at:
AI takeoff Speed <https://astralcodexten.substack.com/p/davidson-on-takeoff-speeds?utm_source=substack&publication_id=89120&post_id=127386375&utm_medium=email&utm_content=share&triggerShare=true&isFreemail=true> I have picked out a few quotations from it that I like: "The term “slow AI takeoff”, Davidson is a misnomer. Like skiing down the side of Mount Everest, progress in AI capabilities can be simultaneously gradual, continuous, fast, and terrifying. Specifically, he predicts it will take about 3 years to go from AIs that can do 20% of all human jobs (weighted by economic value) to AIs that can do 100%, with significantly superhuman AIs within a year after that. [...] It seems like maybe dumb people can do 20% of jobs, so an AI that was as smart as a dumb human could reach the 20% bar. The compute difference between dumb and smart humans, based on brain size and neuron number <https://slatestarcodex.com/2019/03/25/neurons-and-intelligence-a-birdbrained-perspective/> , is less than 1 order of magnitude so this suggests a very small gap. But AI can already do some things dumb humans can’t (like write coherent essays with good spelling and punctuation), so maybe this is a bad way of looking at things." "It takes much more compute to train an AI than to run it. Once you have enough compute to train an AI smart enough to do a lot of software research, you have enough compute to run 100 million copies of that AI. 100 million copies is enough to do a lot of software research. If software research is parallelizable (ie if nine women can produce one baby per month - the analysis will investigate this assumption later), that means you can do it really fast." "Around 2040, AI will reach the point where it can do a lot of the AI and chip research process itself. Research will speed up VERY VERY FAST. AI will make more progress in two years than in decades of business-as-usual. Most of this progress will be in software, although hardware will also get a big boost. My best guess is that we go from AGI (AI that can perform ~100% of cognitive tasks as well as a human professional) to superintelligence (AI that very significantly surpasses humans at ~100% of cognitive tasks) in 1 - 12 months." "It intuitively feels like lemurs, gibbons, chimps, and homo erectus were all more or less just monkey-like things plus or minus the ability to wave sharp sticks - and then came homo sapiens, with the potential to build nukes and travel to the moon. In other words, there wasn’t a smooth evolutionary landscape, there was a discontinuity where a host of new capabilities became suddenly possible. Once AI crosses that border, we should expect to be surprised by how much more powerful it becomes." "Sometime in the next few years or decades, someone will create an AI which can perform an appreciable fraction of all human tasks. Millions of copies will be available almost immediately, with many running at faster-than-human speed. Suddenly, everyone will have access to a super-smart personal assistant who can complete cognitive tasks in seconds. A substantial fraction of the workforce will be fired; the remainder will see their productivity skyrocket. The pace of technological progress will advance by orders of magnitude, including progress on even smarter AI assistants. Within months, years at most, your assistant will be smarter than you are and hundreds of millions of AIs will be handling every facet of an increasingly futuristic-looking economy." John K Clark See what's on my new list at Extropolis <https://groups.google.com/g/extropolis> bs8 -- 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 everything-list+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/everything-list/CAJPayv1W0EVSjounHzM%2BsLeVnPE_GTXW3ZMHLO%2BdiXE%2B68S5%3DQ%40mail.gmail.com.
