We acheived AGI yesterday and not a single thread about it?

You are overestimating the power of the neural network in our heads and the ability of a machine to be trained on all Stratego games ever played and store in its “head” (as weights and biases) the same or better heuristic strategies compared to what the best humans learn. For well defined search problems like board games, it is just a matter of compute power and memory, not “limitations [that] are inherent and won’t go away”. There isn’t any function — or a program that maps any input to some output — that a simple two-layer neural net with enough neurons can not learn to any desired level of accuracy.

Our cerebral cortex responsible for cognitive tasks only has ~16B neurons. The cerebellum has many more but it’s largely for coordination and motor control (like for robotics tasks).

Original Timeline (from The Terminator, 1984

The war between humans and machines is set in the 2029 timeframe, where John Connor leads the human resistance.

AI is already as good or better than the best humans at Stratego:

https://www.science.org/doi/10.1126/science.add4679

O3 solves novel theoretical math at the advanced college/grad student level, maybe even a little better in some domains. Is it going to win the fields medal tomorrow, no. But the rate of improvement is impressive.

There are the problems O3 is solving. I imagine math professors will find these trivial. I’ll let others judge.

It cost thousands of dollars of energy to solve each of these. Orders of magnitude in efficiency gains or energy costs will need to be realized. That’s certainly feasible, if not likely.

I wouldn’t say being ranked equivalent to the 175th best human is “superhuman,” exactly

How did the top 175 do on their bar exams?

I was just at a conference for a major software company that launched a proper AI agent on their website. The AI agent had an immediate 50% decrease in the number of customers requesting to speak to a human - vs the traditional bot with programmed conversations. Overnight, the humans on the other end of the chat bot had 50% of their work disappear. Now, you can guess weather they will get more high value work or some will lose their jobs. I think jobs will get chipped away over time - which is how it will effect our lives and I suppose it depends on where you work for how soon that will be. They are calling the AI agents "Digital Labor".

It doesn't have to be AGI to be transformative now. The magic happens when they train the AI on your companies data. ChatGPT learned from the internet and the model has to learn and wade it's through all of the garbage to get to the good answers. What if your company feeds it only the good data and right answers? When the AI learns from your companies data that includes 10s or 100s of the thousands of records answering problems, the data is much more targeted and accurate than the garbage used to populate the versions everyone has access too. This is early days. They won't need AGI to create dynamic change.

I thought this article was quite interesting, arguing that AI progress toward AGI seems to be plateauing.

theintrinsicperspective.com

AI progress has plateaued below GPT-5 level

Erik Hoel

Or, "Search isn't the same thing as intelligence"

Open link

Yes AI can be transformative on lots of problems, IF the data are well posed. I would say that so far for science problems it has been extremely domain-specific.

Some problems, like protein solving are generalizable, and AlphaFold completely solved long-standing problems (see Nobel prize this year). For a lot of other problems in biology the questions are much more open-ended and really data-limited rather than computation-limited.

How fast can it run a 5k

Would Chat GPT also have known how to spell "achieved" correctly, dear?

AGI is general intelligence equal to humans. Deep learning, combined with Monte Carlo tree search and reinforcement learning is already the #1 chess and Go player. We are still at ANI (narrow) intelligence. AI is not equal to humans for over 90% of tasks. For many tasks AI is not at 5th grade level. For example with Google computer vision which cost $400 million to develop, it can only recognize a cat's face 90% of the time. If the cat is sideways, the CV software is useless. That is why companies do not use it in self-driving cars.

Another argument why LLMs are not at AGI level is because humans have to continually refine them, because they aren't capable of recursive learning like humans. From my understanding, the current success of LLMs is scaling--the amount of data and parameters that are used. There is a lot of discussion about how much further they can improve because of the complexity which leads to more hallucinations.

LLMs have an advantage over humans due to brute force. Quantum computing is not really computing rather quantum processing which will further increase the speed, but does not improve reasoning. Machine intelligence reasoning is not like human understanding, it is just seeing more examples of human data and that increases the probability.

There are plenty of PHD’s that aren’t very smart. The only way artificial intelligence is to become meaningful is when it proclaims the Lord Jesus Christ is its personal savior and the only way to Salvation.

HAHAHAHAHAHAHAHAHAHAHAHAHAHA!
<takes breath>
HAHAHAHAHAHAHAHAHAHAHAHAHAHA!

Good one!

If you tell it what to do, it is not agi! It is a computerized slave good at certain tasks. You don't say that a chess computer is smarter than a chess grandmaster, just that it is better at chess.

I would be much more impressed if you asked openAI to do a basic algebra problem, and it refused to do it.

I'm using ChatGPT to line edit my current novel and it's amazing. However, I still need other humans to review my writing because AI can refine my work but it can't tell me where my blind spots are.

“I know more about AI than AI knows about AI. If AI knows about stuff, I know that AI knows that I know more than AI, and I know about myself also”

Trump has nothing do with this. He is the new president because he won the election. The opposing party with the very low approval rating, with preachy women, spoke in word salad dialect, the stench/stank of Hunter Biden, which ran a candidate that did not do well in the democratic primary were all partially responsible.

You are correct in that Republicans were better at this particular task than democrats. We'll see if liberals have recursive learning next time.

I’m a lawyer and read some legal submissions last week and at a glance they sounded ‘right’. Looking into it further though it appeared they had been written by AI, it had incorrect statute references and referred to case law which did not state what the writer/AI thought it had.

great. thanks bro