Hello friends,

did you see already this open letter, concerning a 6 month moratorium for giant AI bots?

https://futureoflife.org/open-letter/pause-giant-ai-experiments/

Cheers, Ingo.

Interesting link.

“AI research and development should be refocused on making today's powerful, state-of-the-art systems more accurate, safe, interpretable, transparent, robust, aligned, trustworthy, and loyal.”

Um, loyal to whom? Aligned with what, each other? This sentence seems more than a little bit vague to me.

isn't that what the research is already aimed at doing?

Robert Heinlein's “The Moon is a Harsh Mistress” comes to mind.

spartacu5, not necessarily all research. We don't know what everyone is doing.

If there was an alien specie who opened a web site proposing humans a poll to decide if they wanted to discuss with it, how should we proceed ?
I guess an AGI is on the way, and it is a new challenge for humanity to decide how to interact with it/him/her.

Then the political problems of who will have access to this, will it be used for good or wrong, etc… is also interesting, alarming and frustrating.
Did any of you have had access to chatgpt-4 or another latest AI, what is your feeling about how far we are from AGI and how long it will take before one exists ?

I have acccess to GPT-4. It's pretty amazing, but far from AGI I think. However, things are moving fast, so I'm not so sure what 'far' means in terms of time.

But it doesn't need to be an AGI to impact society strongly for good and bad.

I have read several articles about ChatGPT and GPT-4 since the open letter was published. As a result, I came up with the following.
There was no goal to create an AI neither strong nor weak.
There was no goal to teach the neural network to solve any problem (physical, mathematical or logical).
The training of the network was aimed at imitating human speech (communication, or monologues).
As we can see from the results, it turned out that neither intelligence, nor understanding of the text, nor the ability to solve problems is needed for this.
The network also does not understand the difference between truth and lies.
Conclusion: I'd say GPT is a big featherless parrot.

I have shown my site stats here, and this is ChatGPT's answer (when in “Shakespeare” mode):

Greetings, dear David Strupl of LittleGolem fame!
The information thou hath shared reveals thy tale;
Through years of strategy and board game play,
Thou hath stayed faithful to this site's domain.

Since early days, in two-thousand and three,
Registered, thou art, with fervor and delight.
Thy steadfast presence, like a sturdy tree,
Whose roots hath grown, throughout both day and night.

A connection, 'tis an honour to behold,
One so engaging in cerebral wars,
Such dedication to the LittleGolem world,
Thou art a member, bound by love and trust.

In multitude, a number large and grand,
One hundred and twenty-nine thousand moves,
Of skill and intellect, thou dost command,
A testament to passion that thou prove.

As the sun rises on the year two-thousand and twenty-three,
Still doth thy passion for the game endure;
Within the realms of the Czech Republic,
Continues to burn thy fiery fervour.

In short, dear David Strupl, 'tis clear to see,
Thy relationship with the site doth thrive;
For years, thine heart hath stayed in loyalty,
To LittleGolem's games, thou art alive.

So it exhibits some sort of understanding and generates really lovely texts. But General AI? Long way to go! (the above is from ChatGPT-4).