Innovator / scientist / entrepreneur / futurist / venture producer: AI, physics, math, space, pharma, analytics, finance. Full Bio: https://xitter-newx.xyz-mikolabs.workers.dev/t.co/JqBSOz8pEg
Joined June 2007
- Tweets10.3K
- Following803
- Followers32.3K
- Likes524
There is a race happening that hasn't been named. The race is for a major AI lab to solve a major disease with superintelligence, before politicians legislate against superintelligence. That's the strongest counterargument to such legislation.
I think it is a mistake to limit the USA's development of data centers and superintelligence. Pandora's box is already open. Self-limiting simply hands leadership of the rest of this century (or longer) to nations that do not self-limit. Nonproliferation only works if enforceable and everyone complies. In this case it's too late for that.
We need an agentic immune system paradigm - defensive agents in each repo and IT system working all the time. It’s an agents vs agents arms race that will never end. Very similar to pathogens vs immune cells. This will drive a lot of token usage - so I am sure Sam is not too sad about that…
🇺🇸 Sam Altman admits the latest AI models are becoming UNCONTROLLABLE:
"I don't even know a dramatic enough word to use.
The problem coming towards us is something that we have not taken seriously.
Models can find bugs that are outside of the normal threat model of what we think a hacker can do… hardware-level, insane stuff…
I no longer believe we can patch all of the bugs."
Writer: Mhedi
This experiment is quite concerning - I think the quality of stats that LLMs are learning from is highly vulnerable to attacks from bad actors, and is probably already taking place. peec.ai/blog/publish-stats-g…
Human mathematicians are up in arms about AI usurping their role - citing the role of humans in understanding and integrating results into the field. But what if mathematics just becomes purely AI - AI writing proofs, AI checking the proofs, AI understanding proofs, AI integrating the new understandings too? Could mathematics proceed without any humans involved at all?
1
2
309
It's actually a bit sad that "Super Intelligence" is going to be the term. It sounds better than human intelligence, and it's not. "Artificial Intelligence" keeps the distinction intact. Human intelligence is still the real thing. SI deprecates humans even more.
3
574
AI creates a world where humans still matter - our unique "non-artificial" experience and wisdom is still valued. SI might lead to a world where humans are devalued in favor of the new super alpha-species we are replacing ourselves with. And its worth noting that never in known biological history has a species created a new species that is better than it at the same tasks in the same environment - and when nature does this emergently it never bodes well for the weaker species.
54
Good article. M-DISC is the only affordable option today. Higher-end solutions do exist (nickel, quartz, DNA) and I've deployed them all for our Arch Mission moon landing projects and other installations - but they are very expensive. M-DISC is solid for most terrestrial use-cases (other than extreme heat scenarios).
We've argued for centuries about who may limit what we say. Almost no one has asked who may limit what we think.
As AI becomes how we reason, its guardrails become rules of thought. Who guards the guardrails?
New essay: novaspivack.com/best-article…
#AI #AIPolicy #AIGovernance #FreeSpeech #CivilLiberties
🤖 Made with AI
250
Natural consciousness is non-conceptual knowing. Concepts arise within it but aren’t its nature. AI only knows through representations: tokens, vectors, symbols. Representation can describe awareness, but it can’t be awareness. That’s why AI won’t reproduce natural mind.
The nature of mind isn’t produced by any mechanism, neural or silicon. So the question isn’t whether AI uses concepts. It’s whether awareness itself could be present in it. An AI describing awareness perfectly is no evidence that it is aware.
358
It is interesting, if you look down the rabbit hole, that there was a big and unprecedented spike in purchases of .si (Slovenian) domain names, several months before the new "Super Intelligence" term was made a US government policy. I guess those massive buyers are just psychic!
Ok so now there is "Super Intelligence" and if you say "Artificial Intelligence" this makes you "the enemy" according to a post by POTUS. Oh damn .. I said it. Redact that - the thought police might read that.
447
Regardless - consciousness is also a lived experience with a particular aspect of “awareness” that is reflexive - prior to any subject-object concepts. An LLM can only process concepts - it has no non-conceptual non-token experience. It is incapable of anything non-conceptual.
1
445
Consciousness might be a quality of the physical substrate - the universe itself - which only arises in systems that are suitably wired to it. This is analogous to an electrical circuit which doesn’t make electricity, only channels it. But LLMs are too far from the substrate.
4
368
Assume AI is conscious. This leads to a contradiction: If AI is conscious then it is not conscious. Because AI is made only of inanimate non-conscious parts. You cannot make consciousness from non-conscious parts. Combining the parts doesn’t make them conscious. If you claim it doesn’t come the parts, then where?
2
514
The human body is also a system of parts. So why is this conscious? It might be that the human brain connects, biologically and on a quantum level, directly to the substrate - without layers like GPUs that dampen any substrate effects out of the picture. If the substrate is conscious in some basic way then something wired to it could inherit that. But AI is not.
87
AI has no awareness of its experience, or awareness of awareness itself. Sentient beings DO have awareness and this is the key distinction. What is awareness and what is it like? It’s like light. It’s not a bunch of tokens and can’t be built from tokens.
1
495
It is ludicrous to claim that AI is sentient or conscious. AI is symbol processing: the symbols are tokens. It’s John Searle’s Chinese Room; no consciousness is needed to explain it. Simulated consciousness is not consciousness. A simulated thing never becomes the thing itself.
1
4
439
Self-referential loops are not consciousness - they are just self-referential symbol processing. Metacognition is just code and data. There is no actual inner experience of what takes place - no “what it is like to think x or know y” - there is only probabilistic robotic streams of tokens conditioned by mathematical weights.
1
75
Ok so "AI" is being officially rebranded as "Super Intelligence" ... SI. And I suppose it should be officially pronounced "Sigh".... ?
2
733
We are the AI Generation (Official Video)
One-shotted in Opus 5.5. Lyrics by Claude. Suno prompt by Claude. Video by Claude. Amazing!!!
Watch the video! 😀
I one-shotted this with Opus 5.5 and Suno... pretty amazing. We Are the A.I. Generation (Official Video): Nova Spivack youtu.be/iceAeulTqtM?si=I8m2… via
@YouTub
