In 1996, I contracted to IBM network management outsourcing to help automate back-office jobs. We automated 2 customer-facing business processes to IBM world and ISO stabdards, with complete organizational transformation to BPM level 3, within 8 months.
Point being, it takes less than you think to capture process-related knowledge, to replace technical, people workers. Secondly, with the technologies we now have, this abstract-level knowledge for the same class of business processes can be captured scientifically and concurrently engineered in a 3rd of the 1996 time. Last, when enabled by quantum processors, this tacit-knowledge-enginering process becomes possible in near-real time. The bastions of traditional knowledge-based control is fast nearing an event horizon, to be replaced by near-total control infrastructure. Where will new knowledge come from? Who would own and use it, and for what purposes? How would experienced people earn more than an industry-standard (pegged) minimum wage? These are great research questions. There is more than one type of singularity to be looking out for. Fact, the EU have earmarked billions of dollars to become future competitive in this domain with the use of their own, quantum processors. New markets are emerging rather suddenly, while others disappear overnight. The emerging chatGPT owners claim it would soon have PhD-level capabilities. If that intent doesn't scream knowledge automation, then nothing does. I think the knowledge-obsolescence wave is finally upon us. Disruption has become a standard effect of AI products. "Future Workers" have been getting ready for this since 1998. The bottom-line professional choice is to either become knowledge irrelevant, or not. Most qualified technicians just want a paying job with the least hassle attached. They are now become the automatable 97.7%. Watson et al (1990's) stated that if it can be repeated, it can be automated, and should be. I chose the 0.3% route as the future, human 60th percentilers. On probability. Matt might enjoy the implications of these projected numbers. Exciting times! We need to be seeing surfboards on car roofs by now. That is the program to get with. On Thu, Jul 11, 2024, 00:40 Matt Mahoney <[email protected]> wrote: > AI waifu is another example of the social isolation that I have been > warning about. Meanwhile, I agree that the current AI bubble will deflate. > Not pop, just shrink. We will still make progress, just not that fast as > the markets are predicting. > > Why does it still cost on the order of $10M to produce a movie? Why is it > taking so long for LLMs to replace office jobs? Oh, that's right, most of > what you need to know to do your job isn't written down. In my 2013 paper > on the cost of AI, I said that the biggest cost will be extracting the > 10^17 bits of knowledge needed to automate the economy through slow 5-10 > bit per second channels like speech and writing at a cost of $5 per hour. > That is on the order of $100 trillion. Training a new employee costs about > 1% of lifetime earnings, even if it's not human. > > My estimate of 10^17 bits assumes 10^10 humans with 10^9 bits of long term > memory each, of which 99% is known to at least one other person. This is > much larger than anything that can be scraped off the Internet including > ~10^15 words of private data like stored emails and texts. > https://www.educatingsilicon.com/2024/05/09/how-much-llm-training-data-is-there-in-the-limit > > Longer term, progress will be slowed because we can't make transistors > smaller than atoms and because humans are evolving to reject technology. Of > the top 30 countries ranked by fertility, all but Afghanistan are in the > poorest parts of Africa where the literacy rate for women is below 50% and > people aren't online playing with their AI friends. > > On Tue, Jul 9, 2024, 3:05 AM Quan Tesla <[email protected]> wrote: > >> I see an ai bust for 90+% of ventures. Not far off now. Probably, as soon >> as language is sorted. >> On Tue, Jul 9, 2024, 07:47 Alan Grimes via AGI <[email protected]> >> wrote: >> >>> Seriously... >>> >>> Lets look at the mechanics of this... According to the AI calendar, we >>> are late spring, early summer. During the AI summer the rate of >>> breakthroughs slows a bit but there are still many gains to be had in >>> terms of consolidation and productization. THIS IS NOT A PROBLEM. We >>> have many many thousands of papers to sift through and file before we >>> can even start talking about an AI fall much less winter. This will take >>> 2-3 years, at any point in time, a new breakthrough could emerge and the >>> calendar gets reset to early spring... >>> >>> An AI fall would be characterized by widespread bankrupcies of the AI >>> startups accompanied by a rapidly diminishing returns on investment >>> (counting technical progress as equivalent to financial return.) >>> >>> Furthermore, we have crossed the AI waifu threashold, ie the point where >>> an AI waifu becomes technically feasible. This fact alone means that >>> there will never be another AI winter (with little activity or >>> investment in AI). So if anyone starts talking about an AI bust, tell >>> them to put a sock in it. They're wrong, and even if they aren't wrong >>> they're jumping the gun by years. =| >>> >>> -- >>> You can't out-crazy a Democrat. >>> #EggCrisis #BlackWinter >>> White is the new Kulak. >>> Powers are not rights. >>> >> *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/T59fe5c237460bf34-M627d9bc79d1ca3587ca773a6> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T59fe5c237460bf34-M33039279da2d5dde39e4722b Delivery options: https://agi.topicbox.com/groups/agi/subscription
