It’s getting annoying several people pretending there is an exponential progress that doesn’t exist, AGI and ASI are both years away, not a thing in this decade. Current architecture have strong limitations. I use AI everyday i know it is not really rationalizing it is go and useful. But what happened is that AI is in its teenhood development, where it was a child in the 2010s. I think people confuse them for being intelligent because the data is so big they have strong patterns do to most things. But in terms of inventing something dealing with new situations and complex environment. It’s just not there. Just look at the self driving cars. And ASI been achieved before AGI also doesn’t make sense, since the true definition of AGI is that ASI must came from AGI and not being made

    Transformer architecture came in 2017, ChatGPT by 2022. Something will probably happen. We will see some plans for AGI this decade and then the next or later it will be a thing. Since the hype is so big they will probably release a good AI and people will think it is AGI but then they will see the truth after several weeks using it. When nanobots/claytronics and molecular assemblers (several decades away technology according to experts) are not achieved under supposed AGi or even ASI, people will start to believe the hype blinded them. Patience is the key.

    The effect of fake AGI and ASI being released is very likely. Because the language itself is a good gaslighting tool. Current AI mimick human data so, it creates a strong bias towards people who review them. It is a powerful illusion that makes people think something is closer than it really is. This happened in the past.

    Unironically, Sam Altman itself said AGI will change the world less than we think. Maybe, well, because all we will got will be a semi AGI (that is only AGI on the perception of the first uses)

    https://x.com/_rockt/status/1799006089260007549?s=46

    Share.

    1 Comment

    1. AnonDarkIntel on

      That’s because true AGI requires much more data than the internet. It need constant FDVR passthrough data, neuralink data, and even more neurological limbic data. Good luck getting that from humans.