10 years maybe?
It's unusable right now.

Do you believe progress will be exponential? If so then it should only take two or three more years. There are lots of forecasts with that result and they are all built on the expectation of exponential progress.

Even if you don't like the current LLM architecture there is at least one other contender; Yan LeCun's JEPA. There are probably more being worked on in secret at all of the big AI companies as a failsafe.

What are you hoping for when the Singularity gets underway? Are you in a rush to get to AGI or ASI?

itll never be organic (impossible for it to be so) so it's limited to circuitry

Statistical modeling of language for word prediction is like producing energy with nuclear fission, whereas artificial general intelligence is like nuclear fusion power plants, so yes, it will be just 10 years away for quite a while. We already got so far with the technology after all, how much harded can AGI be?

>>105619687
There is a lot of very fast progress in the direction of multimodality, domain of application of the tech and integration into well polished and possibly older software tools, such as photoshop. The is also a trend of quickly growing computational costs, rising much faster than how fast the economic price of computation goes down.

I personally haven's seen any recent improvements in the direction of generality, except for the scaling of inference time compute in the form of "chain of thought" (self prompting and brainstorming).

Claiming an LLM can turn into AGI is as absurd as claiming a book will.

>>105619650 (OP)
If research efforts and funding went to the right things then we could be rhere in 10 yeara. Since it only goes to hype and lies, it will take at least 50 because we will have another ai winter first

>>105619650 (OP)
I'd say 25 years at least. But probably a lot more. The current LLMs will just be one module in a much larger system. LLMs alone are not AI, nevermind achieve the status of an AGI. They are just far too incapable in reasoning and planning. The costs associated are also prohibitive. You can't just scale forever. The way I see it the recent "AI" boom and subsequent hype has warped expectations to an unreasonable degree. I remember people predicting AGI by 2026 back when GPT-4 released. Clearly, that is not going to happen.

Until a computer can smell and accurately rate my farts, it will never be intelligent.

It clearly got thrown off by your poor prose and abhorrent punctuation. I've noticed that models tend to not comply with humans who talk like a mongoloid.

>>105619650 (OP)
Are there really people who feed posts to chatgpt to generate a response? You're joking, right? Write the poem yourself.

>>105620720
There are, not only are there people who do this but I believe it's also built into Microsoft Edge or for a time at least you could have copilot rephrase your post or expand upon an idea.
Grime times, right? Must be something to do with a decline of limes...

>>105619650 (OP)
>dude why can't the AI respond logically to my question explicitly designed to bust AI logic????? its over.......
You could instead just ask it things like you'd ask another human being things and it's surprisingly apt. If you take a step back and think about it for a second, the AI likely isn't extensively trained on how to answer questions explicitly designed to stump it on arguably procedural grounds... it's got a lot of training on how to answer questions the way we ask each other questions