It means people who are too dumb to fully comprehend how dumb they are, will suddenly start appearing more and more in positions of power and responsibility.
The more complex and unpredictable a system gets, the smartest people in the room will automatically start defaulting to avoidance routines. Cause they dont want to be held responsible for what the system does. Its sort of an accelerating feedback loop.
There has to be a version of "Betteridges law of headlines" which is something like "any headline about AI and the Singularity is junk, just ignore it"
Wish I had more than one upvote to give this comment.
Error-free translation isn't even AGI, and AGI isn't the singularity. The singularity is an exponentially self-improving AGI.
Will we have error-free translation by 2030 (or at least better than human)? Maybe. Will we have AGI? Highly unlikely. Will we have the sigularity? Absolutely not.
Of course, the singularity won't come until the year afterwards when (extrapolating the trendline from the article one year further) it will take negative time to correct AI translations.
With this newfound ability to generate free time for the world's best AI researchers by having them proof-read AI-generated translations, we will unlock rapid exponential growth in AI research.
It's laughable but interesting too when people predict with certainty emerging exponential technologies. Even the best minds have done this huge mistake like you. Absolutely not he said...
dunno, some more money? Or wait for openAI to solve it?
> Scientists are researching several different approaches to achieving it, including inertial confinement, magnetic confinement, and beam-target fusion. Inertial confinement involves compressing and heating a small thermonuclear fuel pellet with powerful lasers until it undergoes fusion, while magnetic confinement uses a combination of powerful magnetic fields to contain and heat a plasma until it undergoes fusion. Beam-target fusion involves shooting high-energy particles at a target to produce fusion reactions. All of these approaches require great advances in technology, engineering, and materials science in order for them to be economically feasible.
The word may is such a powerful tool for writing headlines, not to mention making consequence-free predictions. You're not saying something will or won't happen, and the universe is a strange place — hey, you never know!
Another powerful technology is the phrase up to when describing levels of service ("kills up to 99.9% of germs", "speeds of up to 200 Mbps"). Just means you are guaranteed not to get 201 megabits, or kill all the germs.
> The word may is such a powerful tool for writing headlines
It's the sneaky workaround to betteridge's law. "Will Humanity Reach Singularity Within 7 Years?" becomes "Humanity may reach...". "Will an asteroid destroy humanity in 2023?" becomes "Asteroid may destroy ..."
Aside from the issues with the article itself, I have to vent a bit about how far popular mechanics has fallen.
All of the front page articles are either, "BUY THIS" or "TECHNOLOGY HAS DOOMED US ALL".
There's no mechanics anything easily findable on the page. Where's the DIY "build these things / here's what you can use tool X for / develop this maker skill by x" that used to be the blood and soul of Popular Mechanics?
In the 60's you could learn how to build a freaking atom smasher from one of their articles, now the articles can't even tell you which IKEA bookshelf is worth buying.
17 comments
[ 9.7 ms ] story [ 87.6 ms ] threadThe more complex and unpredictable a system gets, the smartest people in the room will automatically start defaulting to avoidance routines. Cause they dont want to be held responsible for what the system does. Its sort of an accelerating feedback loop.
Error-free translation isn't even AGI, and AGI isn't the singularity. The singularity is an exponentially self-improving AGI.
Will we have error-free translation by 2030 (or at least better than human)? Maybe. Will we have AGI? Highly unlikely. Will we have the sigularity? Absolutely not.
With this newfound ability to generate free time for the world's best AI researchers by having them proof-read AI-generated translations, we will unlock rapid exponential growth in AI research.
I'd settle for self-improving AGI. Then let the AGI eventually figure out how to do so exponentially.
> Scientists are researching several different approaches to achieving it, including inertial confinement, magnetic confinement, and beam-target fusion. Inertial confinement involves compressing and heating a small thermonuclear fuel pellet with powerful lasers until it undergoes fusion, while magnetic confinement uses a combination of powerful magnetic fields to contain and heat a plasma until it undergoes fusion. Beam-target fusion involves shooting high-energy particles at a target to produce fusion reactions. All of these approaches require great advances in technology, engineering, and materials science in order for them to be economically feasible.
a lot more money?
Another powerful technology is the phrase up to when describing levels of service ("kills up to 99.9% of germs", "speeds of up to 200 Mbps"). Just means you are guaranteed not to get 201 megabits, or kill all the germs.
It's the sneaky workaround to betteridge's law. "Will Humanity Reach Singularity Within 7 Years?" becomes "Humanity may reach...". "Will an asteroid destroy humanity in 2023?" becomes "Asteroid may destroy ..."
I can't decide which I hate more.
All of the front page articles are either, "BUY THIS" or "TECHNOLOGY HAS DOOMED US ALL".
There's no mechanics anything easily findable on the page. Where's the DIY "build these things / here's what you can use tool X for / develop this maker skill by x" that used to be the blood and soul of Popular Mechanics?
In the 60's you could learn how to build a freaking atom smasher from one of their articles, now the articles can't even tell you which IKEA bookshelf is worth buying.