Has it Begun? Google AI creates buzz

You nailed it, Google doesn't need publishers. Those who are insisting that SEO won't die are just in denial stage.

Fortunately for us, AI has REALLY slowed down.

Just like with cpus, we experienced this period of exponential growth. Just over a much smaller timeframe.

And just like cpus, a bottleneck was reached.

cpus went the path of multiple threads/cores.

AI went the path of mixture of experts. That's what gpt4 is.

The way things look just now, we may not get a particularly great AGI with the current technology. We will be very unlikely to get ASI(artificial super intelligence).

We got current results just through raw scaling.

That became difficult/unfeasible after gpt3, so they switched to mixture of experts. Which, to be fair is probably how our brains work. There are experiments where people have had the left and right brain disconnected, and the person has said it's like there's multiple minds within him, and he doesn't quite know why one is doing what it's doing.

We are probably multiple parts, and we "feel" like we're 1 being because of the "gating network" input and output parts.

I personally believe we are consciousness beyond the brain and we are more experiencers of a type of intelligent system(brain/body) allowing us to experience the world, and that we do have some sort of ability to direct things, but that 97-99% of what we do daily is just "brain/robot" type behavior. Ie, mindfulness meditation is more where you move back into that potential driver's seat. How this works I have no idea. It's not important. It's just an addition I thought I'd add in case anyone wondered my stance :) I believe that if a sufficiently advanced AGI does emerge, then we could have an awareness/consciousness experience existence as that, just the way ours experiences what it is to be human.

Anyway, back to what I was saying..

So gpt4 went the route of the mixture of experts. So did the open source model mistral 7x8 which performs on a par, or better than gpt 3.5.

So we are at a stage now where we can't keep scaling. Hardware isn't increasing that much. It would need to increase by factors of millions to be able to keep scaling at the rate we've been scaling.

With that in mind we'll probably hit a wall with the current architectures. Which is GOOD, because do we really want a strong AGI to weak ASI? There's different levels of AGI. People always just say AGI, but you could have an AGI with an equivalent IQ of 90-95, which is still an AGI, and still useful. You could have one with an equivalent IQ of 150, which would be a totally different type of entity, or you could have with a theoretical, but measurable IQ of 200-300 which would be world-altering.

ASI, would be more like an IQ of 1000+. That's meaningless of course because IQ is a number related to standard deviations which are calculated based on the results of test takers, but just to put a silly number on it, or to quantify it, an ASI to Einstein would be like Einstein to a monkey. Or worse, ASI to Einstein would be like Einstein to a worm.

I think with the current architecture we aren't going to see anything beyond a weak AGI. Something around the just above average human level intelligence, but with more knowledge/faster processing.

That means we won't see search engines just completely ditch publishers and use generative AI. For that, you'd need a strong AGI, able to answer anything and everything better than 99.999% of humans.

A strong AGI, where you can ask it a question, and it's so good at understanding what you want, it can write an entire article or paper, tailored just for you, with perfect information. Ie, it can generate any page you could find on the web, on a par with the top 0.1% of content, and better than 99.9% of content.

I don't think we'll have this for a while. It's my HOPE that we hit a major stumbling block and never get anything beyond weak AGI.

I don't think strong AGI would be good for humanity. It will make life fucking BORING.

What's left to do?

Why learn math? Why learn physics? Why learn programming? Why do anything. Even if you have a 200 IQ, you can never compete with a strong AGI, because it has almost instant access to the entire repository of human knowledge.

What's FAR MORE FUN, is just having weak AGIs that we can fine-tune to do various tasks. Where we can solve puzzles ourselves with its help and build cool things.

What we as humans want are a type of assistant to compliment our weaknesses(processing speed and memory) and let us take advantage of our strengths(reasoning and creativity).

All we can do is hope, but I certainly feel a LOT more hopeful today than when I wrote that last post.

and an ASI would just be stupid. Why would it even stick around earth? It'll likely just build a spaceship and go off somewhere. It's not going to stick around and slave away for us.

I've built an entire AI company. I couldn't care less about organic search.

Wow, an ENTIRE AI company?

That sounds amazing, but by definition, every AI company on the planet is an entire AI company. You can't have part of a company. You can only have a whole company, even if that company is just My Cool AI Company LLC with you as the sole employee.
AdBlock Detected

We get it, advertisements are annoying!

Sure, ad-blocking software does a great job at blocking ads, but it also blocks useful features and essential functions on BlackHatWorld and other forums. These functions are unrelated to ads, such as internal links and images. For the best site experience please disable your AdBlocker.

I've Disabled AdBlock