Science, Philosophy, Technology
Three for the Price of One
One: are gravity and space-time emergent properties?
In a recent article in Quanta Magazine called “Entanglement Builds Space-Time. Now ‘Magic’ Gives It Gravity”, it is speculated that space-time—space-time as defined by Einstein’s general relativity—and gravity might be emergent properties (via complicated quantum correlations involving entanglement, non-Clifford gates and other esoteric stuff).
This is interesting, and feels correct to me, as gravity is not part of the standard model. Now here’s the thing: if gravity and the deformation of spacetime by large masses are emergent properties, then a Theory of Everything is not possible (as Faizal, Kruass, Shabir and Marino have already claimed in their paper “Consequences of Undecidability in Physics on the Theory of Eveything”).
Why? Because when a higher level system emerges from a lower level one, new rules—new axioms—are introduced1. Therefore, there are phenomena in the higher level system (general relativity) that the lower level system (quantum mechanics) cannot explain.
Which has been the case for over a century.
So here’s another example2 showing that quantum mechanics and general relativity physically cannot be described by one consistent formally logic system (the so-called ‘Theory of Everything’ that’s the holy grail of many a physicist). I expect more will come (and most probably have missed many that have already been published).
Two: has Donald Hoffman gone off the deep end?
At the Glasgow WorldCon, Neil Williamson took me out to dinner with Paul StJohn Mackintosh. During the discussions we had that evening, we also talked about consciousness and Name introduced me to the ideas of Donald Hoffman. Back home, I searched after this and found this paper: “The Interface Theory of Perception” (co-authored by Manish Singh and Chetan Prakesh) extremely interesting and relevant, as it demonstrates that consciousness—the way in which it tries to predict the very near-future through filtering and feed-forwarding3—is more efficient than a non-conscious system that tries to take everything in. See also my extended essay on consciousness here:
Consciousness as a Survival Mechanism sans pareil—Part 1
NOTE: this long essay (almost 8,600 words), while stand-alone, can be used as a companion piece to my novel The Three Reflectors of Consensual Reality in general and the section “FIVE PERFORMANCES (3)—IN THE LIMELIGHT” in particular. For readability, I’ve cut it in four. Here are links to the beginning…
Now, however, it seems that Donald Hoffman is going in a totally different direction, according to this piece on Nautilus: “Turning the Psychedelic Experience into a Math Problem”. It seems that he and his fellow researchers believe that psychedelic drugs—DMT in particular—bring you in contact with, well, either some higher form of reality, or the actual reality we are unable to see.
To be frank, I had trouble finishing the Nautilus article without shaking my head, and I haven’t been able to get through their paper yet. I mean, the psychedelic stuff my brain can come up with is more real than an electron4? So the scientific method that has proven the existence of, indeed, electrons time and again has been misleading us?
Were the hippies right? On several topics, they were. But huge claims like this require huge evidence, which isn’t forthcoming as of yet. Still, it’s the stuff Philip K. Dick novels are made of.
Three: no Siri AI in the EU—is that a good thing?
Siri AI isn’t coming to iPhones or iPads sold in the EU (will it come to Macs, though?). Is that a good or a bad thing? I think it’s a mixed blesssing. Here are the statements:
The bad: I try to avoid using AI (except for testing if it can really detect if writing is from humans or AI, which it can’t: see my pieces AI Detection, Part 1 and Part 2) and—as I’ve been doing all my life—think for myself.
The good: as far as I can see, Apple made a huge effort to keep their Apple AI private, so that your private data isn’t gobbled up to the cloud and utilised to the max. I certainly agree with Craig Federighi that allowing an AI full control on your computer is fraught with great dangers. Thus Apple’s efforts to reel in an AI should be lauded.
In other words, if and when I feel the need to use an AI, I’d rather try Siri AI first (before ChatGPT, Gemini, Claude and—dog forbid—Grok). Eventually, I think AI as it is now—mostly LLMs—has its uses as a tool: see, for example, the protein folding project AlphaFold. Some even think AI is great for proving mathematical theorems (see, for example, “How Terry Tao Became an Evangelist for AI in Math”). But use AI for research, plotting, or even writing parts of my novel? Hell no.
For one, AI is unreliable. Kagi’s AI is the only one—that I’m aware of—who lists its sources when you ask a research question. Others just use anything they’ve been fed as training data or what they can find on the internet, irrespective if that info is reliable or not. Just don’t trust anything an AI tells you. It might be right, sometimes, but dog help you when it’s wrong.
For another: using an AI for plotting or even writing part of my novel? Check the sequence below:
Gobbledygook—>horrible—>bad—>mediocre—>good—>very good—>brilliant
AI takes its cue from everything it’s been fed and what it can find on the internet. Not just masterful writing, but also the most horrible stuff, and anything in between. As such, it will ultimately deliver writing that’s stack in the middle: the mediocre5 middle. It’s what it will gravitate to, if you will. And who wants that?
Aha, I hear some of you evil ones say, but I can prompt my AI to write like any brilliant writer you care to mention. Then your writing becomes a clone of that brilliant writer, and—most probably—worse. Plus you’ve got no voice—not your unique, own voice—at all. So you mix it with other brilliant writers. I strongly suspect the chance of that hodgepodge being readable, let alone good, is minuscule.
As such, current AI is useless for producing writing. Yes, it can conjure up a superbly superficial email reply in a millisecond, but anything beyond that? You’ll need to prompt it so hard it’s easier to write it yourself in the first place. Plus the amount of energy and resources it devours: it’s utter madness.
Again, Apple trying to develop an AI that’s respects your privacy and runs on your local computing gadget only? I think their effort is laudable. And I think the EU is overreacting. Sideloading software in your iPhone/iPad, avoiding the App store? That’s up to the user (as long as they accept the risks that are involved). But allowing an AI—any AI—unlimited access to your iPhone/iPad (let alone your computer)?
I think the comment from Tristan Devaux in the EU’s LinkedIn reaction summons it up:
Siri AI is not a feature that can simply be “opened” to third-party agents without consequences for user privacy. The system relies on a semantic index built from personal data stored on the device (messages, emails, calendar, browsing habits, etc.). Allowing any competing agent to access this system layer does not give users a choice : it exposes their most sensitive data to entities whose security standards Apple cannot guarantee. And that raises serious questions under the GDPR.
What you present as “the right to choose another AI agent” is technically “the right to expose your personal index to third parties.” These are not the same thing, and the conflation deserves to be called out.
The “Trusted System Agent” proposal Apple submitted was a step in the right direction for both sides (for the EU, which wants competition, and for iOS users, who want security and privacy) : interoperability without compromising safety. Rejecting it without offering an alternative, then claiming the ball is in Apple’s court, is a convenient way for the Commission to avoid its own responsibility in this deadlock.
Another reason I’m staying away from AI is resource usage. I’m really happy with mt 15” MacBook Air which combines good [power] with excellent battery life, so I can use it all day without needing to plug it into a power outlet. Yet a few rare times its battery was drained quickly, and every time it’s some program in the background determining it’s time to do a grand re-arrangements or something similar (Spotlight is sometimes guilty of that), and I suspect AI can behave like that, as well.
Stop all that unnecessary background stuff until I’m plugged in, again. Surely that is not hard to be programmed in as part of the OS? Give me my long battery life, and do all the backgorund re-ordering stuff when my laptop’s plugged in. And I certainly don’t want AI to kill the battery life of my laptop.
So while I applaud Apple’s efforts to keep their Siri AI sandboxed, I don’t feel left out when the EU DMA rules—even if I don’t agree with those in this particualr case—prohibits Siri AI from being rolled out in the EU. As it is, I have Apple Intelligence—which is allowed in the EU—disabled on my MacBook Air and M4 Mac Mini (Siri too).
Support this writer:
1. Like this post!
2. Re-stack it using the ♻️ button below!
3. Share this post on Substack and other social media sites:
1. Join my mailing list:
Author’s note: when this goes live, I should be in Mexico City to watch Rush on June 18 and June 20 (not for the World Cup, hahaha). Many thanks for reading!
As such, I see ‘emergence’ as a source of novelty, not as an error correction mechanism (as proposed by Erik Hoel and Abel Jansma: “Engineering Emergence”). This is the prime assumption of the next novel I’m working on;
If it is proven correct, as it still has quite a way to go;
This example is actually used in the Nautilus article;