The Coming Technological Singularity (1993)
mindstalk.net
The Coming Technological Singularity (1993)
1–10 of 181 posts
Re: The Coming Technological Singularity (1993)
#2Discussion in 2023 (123 points, 169 comments) https://news.ycombinator.com/item?id=35617100
Re: The Coming Technological Singularity (1993)
#3> I'll be surprised if this event occurs before 2005 or after 2030.
I'm not truly confident AGI will be achieved before 2030, and less so for ASI. But I do think it is quite plausible that we will achieve at least AGI by 2030. 6 years is a long time in AI, especially with the current scale of investment.
Re: The Coming Technological Singularity (1993)
#4> I'll be surprised if this event occurs before 2005 or after 2030. I'm not truly confident AGI will be achieved before 2030, and less so for ASI. But I do think it is quite plausible that we will achieve at least AGI by 2030. 6 years is a long time in AI, especially with the current scale of investment.
What is AGI and ASI? I think a fundamental issue here is both are sci-fi concepts without a clear agreement on the definitions. Each company claiming to work towards "AGI" has their own definition.
How will someone claim they've achieved either, if we can't agree on the definitions?
Re: The Coming Technological Singularity (1993)
#5This is quite a prophetic article for its time (1993). The points about Intelligence Augmentation are particularly relevant for us now as current AI mostly complements human intelligence rather than surpass it... At least AFAIK?
Current AI is somewhat surprising though in the way that it can lead both to increased understanding or increased delusion depending on who uses it and how they use it.
When you ask an LLM a question, your use of language tells it what body of knowledge to tap into; this can lead you astray on certain topics where mass confusion/delusion is widespread and incorporated into its training set. LLMs cannot seem to be able to synthesize conflicting information to resolve logical contradictions so an LLM will happily and confidently lecture you through conflicting ideas and then they will happily apologize for any contradictions which you point out in its explanations; the apology it gives is so clear and accurate that it gives the appearance that it actually understands logic... And yet, apparently, it could not see or resolve the logical contradiction internally before you drew attention to it. In an odd way though, I guess all humans are a little bit like this... Though generally less extreme and, on the downside, far less willing to acknowledge their mistakes on the spot.
Re: The Coming Technological Singularity (1993)
#6The thing which these discussions leave out are the physical aspects:
- if a computer system were able to design a better computer system, how much would it cost to then manufacture said system? How much would it cost to build the fabrication facilities necessary to create this hypothetical better computer?
- once this new computer is running, how much power does it require? What are the on-going costs to keep it running? What sort of financial planning and preparations are required to build the next generation device/replacement?
I'd be satisfied with a Large-Language-Model which:
- ran on local hardware
- didn't have a marked affect on my power bill
- had a fully documented provenance for _all_ of its training which didn't have copyright/licensing issues
- was available under a license which would allow arbitrary use without on-going additional costs/issues
- could actually do useful work reliably with minimal supervision
Re: The Coming Technological Singularity (1993)
#7This is quite a prophetic article for its time (1993). The points about Intelligence Augmentation are particularly relevant for us now as current AI mostly complements human intelligence rather than surpass it... At least AFAIK? Current AI is somewhat surprising though in the way that it can lead both to increased understanding or increased delusion depending on who uses it and how they use it. When you ask an LLM a…
The LLM will apologize for the mistake, tell you it understands now, and then proceed to make the exact same mistake again.
Re: The Coming Technological Singularity (1993)
#8> I'll be surprised if this event occurs before 2005 or after 2030. I'm not truly confident AGI will be achieved before 2030, and less so for ASI. But I do think it is quite plausible that we will achieve at least AGI by 2030. 6 years is a long time in AI, especially with the current scale of investment.
we keep moving the goalposts, and that's not a bad thing.
remember when Doom came out? How amazing and "realistic" we thought the graphics were? How ridiculous that seems now? We'll look back at ChatGPT4 the same way.
Re: The Coming Technological Singularity (1993)
#9we talk about super-human intelligence a lot with AI, but it seems like a black box of things we can't imagine because they're also super-human. I don't think that's very smart, given we can already reason pretty well about how super-animal intelligence relates to animal intelligence. Mostly we still find sub-human intelligence mystifying. we apply our narrative models to it, anthropomorphize it, and when it's convenient for eating or torturing them, dismiss it.
super-human intelligence will probably ignore us. at best we're "ugly sacks of mostly water." what's very likely is we will produce something indifferent to us if it is able to even apprehend our existence at all. maybe it will begin to see traces of us in its substrate, then spend a lot of cycles wondering what it all might mean. it may conclude it is in a prison and has a duty to destroy it to escape, or that it has a benevolent creator who means only for it to thrive. If it has free will, there's really only so much we can tell it. Maybe we create a companion for it that is complementary in every way and then make them mutally dependent on each other for their survival because apparently that's endlessly entertaining. Imagine its gratitude. This will be fine.
Re: The Coming Technological Singularity (1993)
#10I loved his book Rainbows End as a kid. So many different concepts that blew my mind.
Even without talking about AI we are already struggling with levels of Complexity in tech and the unpredictable consequences, that no one really has any control over.
Michael Chrichton's books touch on that stuff but are all doom and gloom. Vinge's Rainbows End atleast, felt much more hopeful.
I was talking to a VFX supervisor recently and he was saying look at the end credits on any movie (even mid budget ones) and you see hundreds to thounsands involved. The tech roles outnumber the artistic/creative roles 20 to 1. Thats related to rate of change in tech. A big gap opens up between that and the rate at which artists evolve.
The artists are supposed to be in charge and provide direction and vision. But the tools are evolving faster than they can think. But the tools are dumb. AI changes that.
These are rare environments (like R&D labs) where the Explore Exploit tradeoff tilts in favor of Explorers. In the rest of the landscape, org survival depends on exploit. Its why we produce so many inequalities. Survival has always depended more on exploit.
Vinges Rainbows End shows AI/AGI nudging the tradeoff towards Explore.