pull down to refresh

The Gentle Singularityblog.samaltman.com/
431 sats \ 9 comments \ @k00b 10 Jun tech
Most of this has been said a million times already, and much of it a decade or more earlier, but it's nice to hear it from the mouth of one of the robot's horses.
write
preview
related posts
50 sats \ 1 reply \ @elvismercury 10 Jun
Haven't read it yet, but wonder if o3-pro has prompted anything. Could just be PR I guess.
reply
100 sats \ 0 replies \ @elvismercury 10 Jun
This is how the singularity goes: wonders become routine, and then table stakes.
Time will tell wrt the singularity, but truer words were never spoken on how the rest of life goes.
reply
100 sats \ 2 replies \ @optimism 10 Jun
In some big sense, ChatGPT is already more powerful than any human who has ever lived. Hundreds of millions of people rely on it every day and for increasingly important tasks; a small new capability can create a hugely positive impact; a small misalignment multiplied by hundreds of millions of people can cause a great deal of negative impact.
Thanks Sam, for confirming that we need sovereign compute.
Solve the alignment problem, meaning that we can robustly guarantee that we get AI systems to learn and act towards what we collectively really want over the long-term (social media feeds are an example of misaligned AI; the algorithms that power those are incredible at getting you to keep scrolling and clearly understand your short-term preferences, but they do so by exploiting something in your brain that overrides your long-term preference).
Speaking of dystopian shit... who's going to tell "us" what "we collectively want"? Sam? Uncle Sam? Fuck that. Thanks for confirming that we need sovereign compute, twice.
reply
100 sats \ 1 reply \ @SimpleStacker 11 Jun
who's going to tell "us" what "we collectively want"?
That's the problem. There is no such thing as "what we collectively want". It's mathematically provable that there's no such thing: #849906
reply
0 sats \ 0 replies \ @optimism 11 Jun
Yes, loved your post.
I remember finding something about the only exception being cardinal voting (ranking each option on a scale) but this gets (over proportionately) negatively influenced by sentiment, uninformed decisions or indifference and thus is only an improvement in theory, as it's impossible to execute with a large, diverse set of "voters".
Therefore imho the trick is to pre-sort limitations as little as possible and instead develop robust systems (in the widest sense of the word) that are configurable. We just had a little battle over this re: Bitcoin Core too!
With AI, which doesn't need any consensus at all, society would do better to harden itself against threats it poses, not legally, but "physically". Otherwise, you will still suffer the consequences from "criminals" but have legal recourse after the fact only.
reply
0 sats \ 3 replies \ @SimpleStacker 11 Jun
Ehhh... I kinda just think it'll be like Star Trek's computer. We'll just go "Computer, do this" and "Computer, do that", and in the end we humans are still the main characters driving the technology towards our objectives.
reply
101 sats \ 1 reply \ @ek 11 Jun
What about dark patterns in these systems?
Dark patterns are already pretty effective even without a computer that passes the Turing test.
reply
0 sats \ 0 replies \ @SimpleStacker 11 Jun
yes, I suppose that could be a possibility. The most unrealistic thing about Star Trek was that humans learned to stop fighting and swindling each other and instead work together to explore the stars
reply
0 sats \ 0 replies \ @unschooled 11 Jun
I like that. It's mostly how I'm using it now and I'm satisfied with my new "assistant."
reply