The Beginning of Infinity

Heard about it, have not read though. I do not expect AGI from LLM. But it's an awesome tool that helps speed-up learning, research and prototyping. And also all this hype accelerate some money to this topic, which is good. Indeed, if natural intelligence possible, why artificial -- can not? According to the roadmap, I more trust in good old-fashioned symbolic AGI and formal methods. NARS + AIXI + elegant dependent modal or substructural (maybe homotopic) typed programming language with strong meta-theoretic properties, as a carrier of observations, knowledge and judgments + a bit of game theory and evolutionary psychology = this is the way, I believe.
And maybe some tricky computational non-von Neumann architecture to have a nice computational complexity for that (not sure about that, but plausible it make sense to utilize some sort of analogous computations in addition to digital ones).

The Beginning of Infinity

Heard about it, have not read though. I do not expect AGI from LLM. But it's an awesome tool that helps speed-up learning, research and prototyping. And also all this hype accelerate some money to this topic, which is good. Indeed, if natural intelligence possible, why artificial -- can not? According to the roadmap, I more trust in good old-fashioned symbolic AGI and formal methods. NARS + AIXI + elegant dependent modal or substructural (maybe homotopic) typed language with strong meta-theoretic properties + a bit of game theory and evolutionary psychology = this is the way, I believe.
And maybe some tricky computational non-von Neumann architecture to have a nice computational complexity for that (not sure about that, but plausible it make sense to utilize some sort of analogous computations in addition to digital ones).

The Beginning of Infinity

Heard about it, have not read though. I do not expect AGI from LLM. But it's an awesome tool that helps speed-up learning, research and prototyping. And also all this hype accelerate some money to this topic, which is good. Indeed, if natural intelligence possible, why artificial -- can not? According to the roadmap, I more trust in good old-fashioned symbolic AGI and formal methods. NURPL + AIXI + elegant dependent modal or substructural (maybe homotopic) typed language with strong meta-theoretic properties + a bit of game theory and evolutionary psychology = this is the way, I believe.
And maybe some tricky computational non-von Neumann architecture to have a nice computational complexity for that (not sure about that, but plausible it make sense to utilize some sort of analogous computations in addition to digital ones).

And with LLM came to our live -- the path from the vision to [AGI] -- become notable closer…

Have you read any David Deutsch, or listened to any interviews of him? The Beginning of Infinity is very good. You might enjoy chapter 7, where he explains why chatbots don’t bring us closer to AGI.

This article of his is also good.

Let me know what you think of his stance.

I have a new Services page where you can hire me for software engineering, philosophy consulting, and more: https://dennishackethal.com/services.html

Hello, and nice to meet you. Your twit https://x.com/dchackethal/status/2031465139401093501 bring me here.
It seemed relevant to my curiosity about AGI topic, so since I believe in synergy and want to be surrounded more by such context, signed up to the website, just in case.

If you interested to discuss and share some AGI-relevant thoughts, I'm in, just let me know. Not a professional at this topic (just an average software engineer), but investigated topic for quire a while, so, I believe, have something to put on the table. And with LLM came to our live -- the path from the vision to the result -- become notable closer, so, who know, maybe we can really bring something beautiful to life.

Hello, and nice to meet you. Your twit https://x.com/dchackethal/status/2031465139401093501 bring me here.
It seemed relevant to my curiosity about AGI topic, so since I believe in synergy and want to be surrounded more by such context, signet up to the website, just in case.

If you interested to discuss and share some AGI-relevant thoughts, I'm in, just let me know. Not a professional at this topic (just an average software engineer), but investigated topic for quire a while, so, I believe, have something to put on the table. And with LLM came to our live -- the path from the vision to the result -- become notable closer, so, who know, maybe we can really bring something beautiful to life.

Welcome to Veritula, Edgar. I recommend reading this guide to learn about Veritula and rationality.

Also, one of our many discussions could be a starting point for you to join our discourse.

What brings you to Veritula?

Welcome to Veritula, @netsu. Check out this guide to understand how Veritula works and learn more about rationality. You may also find one of our discussions interesting.

What brings you to Veritula?

Welcome to Veritula, Phillip. I recommend reading this guide to understand how Veritula works.

Nice article on Popper and Deutsch. You attribute to Popper “the idea that truth is difficult to attain, and that we can only ever get closer to it.” You imply that we cannot fully reach truth.

Do you have some quote/citation where Popper says something to that effect?

How to tell a serious epistemologist from a hobby epistemologist: https://x.com/dchackethal/status/2031465139401093501

A ‘supersession’ isn’t some special flag in the system – it’s just another criticism that can be countercriticized. So I hesitate to implement special functionality for ‘special’ criticisms.

Outdated is different from superseded. The versions show at the top of the idea. That signals at least potential ‘supersession’ before you start reading.

A ‘supercession’ isn’t some special flag in the system – it’s just another criticism that can be countercriticized. So I hesitate to implement special functionality for ‘special’ criticisms.

Ayn Rand on why middle-of-the-roaders can be worse than outright opponents:

[Page 1]
August 21, 1946

Mrs. Rose Wilder Lane
Route 4, Box 42
Danbury, Connecticut

[…]

Now to your second question: “Do those almost with us do more harm than 100% enemies?” I don’t think this can be answered

[Page 2]
Page 2 Mrs. Rose Wilder Lane August 21, 1946

with a flat “yes” or “no”, because the “almost” is such a wide term and can cover so many different attitudes. I think each particular case has to be judged on his own performance, but there is one general rule to observe: those who are with us, but merely do not go far enough, yet do not serve the opposite cause in any way, are the ones who do us some good and who are worth educating. Those who agree with us in some respects, yet preach contradictory ideas at the same time, are definitely more harmful than the 100% enemies. The standard of judgement here has to be the man’s attitude toward basic principles. If he shares our basic principles, but goes off on lesser details in the application of these principles, he is worth educating and having as an ally. If his “almost” consists of sharing some of the basic principles of collectivism, then we ought to run from him faster than from an out-and-out Communist.

As an example of the kind of “almost” I would tolerate, I’d name Ludwig von Mises. His book, “Omnipotent Government”, had some bad flaws, in that he attempted to divorce economics from morality, which is impossible; but with the exception of his last chapter, which simply didn’t make sense, his book was good, and did not betray our cause. The flaws in his argument merely weakened his own effectiveness, but did not help the other side.

As an example of our most pernicious enemy, I would name Hayek.[**] That one is real poison. Yes, I think he does more harm than Stuart Chase. I think Wendell Willkie did more to destroy the Republican Party than did Roosevelt. I think Willkie and Eric Johnston have done more for the cause of Communism than Earl Browder and The Daily Worker. Observe the Communist Party technique, which asks their most effective propagandists to be what is known as “tactical non-members”. That is, they must not be Communists, but pose as “middle-of-the-roaders” in the eyes of the public. The Communists know that such propagandists are much more deadly to the cause of Capitalism in that “middle-of-the-road” pretense.

Personally, I feel sick whenever I come up against a compromising conservative. But my attitude is this: if the man compromises because of ignorance, I consider him worth enlightening. If he compromises because of moral cowardice (which is the reason in most cases), I don’t want to talk to him, I don’t want him on my side, and I don’t think he is worth converting.

As to George Peck, I don’t know enough about him to be able to tell whether he is worth educating or not. I have just received a letter from him in answer to mine. It is a very nice letter, in that he tries to answer criticism honestly, but I am appalled by his mental confusion. He maintains, for instance, that Hitler is worse than Stalin. I don’t know by what possible standard one can establish degrees of evil as between dictators representing exactly the same

[Page 3]
Page 3 Mrs. Rose Wilder Lane August 21, 1946

principle. I am afraid that George Peck means well, but has not given our cause a serious study. Perhaps, he is worth educating. But stay away from Hayek, if you want my opinion; he is worse than hopeless.

Now, am I a good correspondent?

With best regards,

Sincerely,

Ayn Rand

p.s.
I had just finished this letter to you, when, strangely enough, I received an appalling answer to the question you asked me—a final proof that our “almost” friends are our worst enemies. It was the worst shock in all my experience with political reading. I received the Economic Council Letter of August 15th. (Incidentally, I subscribed to that Letter mainly in order to get your book reviews.) And I read that Merwin K. Hart, a defender of freedom and Americanism, is advocating a death penalty for a political offense.

I am actually too numb at the moment to know what to say. I don’t have to explain to you that once such a principle is accepted, it would mean the literal, physical end of Americans; nor to ask you to guess who would be the first people executed under such a law; nor to remind you that the crucial steps on the road to dictatorship, the laws giving government totalitarian powers, were initiated by Republicans—such as the draft bill, or the attempt to pass a national serfdom act for compulsory labor.

I know that you know all that. What I wonder is: is it in your spiritual power to discuss this with Hart? If you can, if you have arguments that would reach him—please do it. I confess I’m helpless in such an instance. It’s too monstrous.

[…]

**F. A. Hayek, who shared the 1947 Nobel Prize in Economics. For AR’s marginal comments on Hayek’s best-known work, The Road to Serfdom, see Mayhew ed., Ayn Rand’s Marginalia, pp. 145–60. 

In her August 24, 1946, response, Lane wrote, “That Council Letter gave me the same shock…. I can take it up with Hart and I shall.”

Source

Sometimes postponement is impossible due to external factors, say. But maybe you can create a new option in time.

Even if you can’t create a completely new option, you could create an option saying, ‘in this situation, I have to act, and I’m running out of time to come up with new ideas, so I’m deciding to do X because Y’. And then that option, as I just phrased it, may have no pending criticisms, in which case you can still act on it. So the rule of not acting on problematic ideas remains intact because ideas are discrete and immutable.

The way out of such conundrums as Podge described them is usually (always?) to create new options (see BoI ch. 13).

Podge wrote in the FoR book club:

[A]n institution that forbids action unless unanimity is reached seems not to function coherently. There are two possibilities. If postponement is uncontroversial, then no special rule is needed, since institutions for adjudicating between competing preferences are only operative when there is disagreement. If postponement itself is contested, then it’s not clear how this rule could be applied consistently, because not acting on x is itself a choice about which we are conflicted.

Alternatively, you could make an option "allow revisions by other users" in the new idea form that is off by default.

You could make an option "allow revisions by other users" in the new idea form that is off by default.

Maybe fun = profitable thought. Not in the sense of ‘thought that leads to good monetary decisions’. I mean it in the literal sense that there’s a kind of wealth being created inside your mind.

Maybe fun = profitable thinking. Not in the sense of ‘thoughts that lead to good monetary decisions’. I mean it in the literal sense that there’s wealth being created inside your mind.

Maybe fun = profitable thinking.

Idea: ability to lock an idea to prevent edits. If you submit the first version, you get to lock the idea. Useful especially on your own profile that you might use as more of a blog and don’t necessarily want others changing your ideas.

web-haptics: Haptic Feedback Finally Comes to iOS Safari

lochie just published a package bringing haptic feedback to the web on iOS: web-haptics. I spent four years at Apple working mostly on UIs, so I pay close attention to design and UX. In terms of presentation, use case, and attention to detail, this package comes as close to perfect as I’ve seen.

The problem

On Android, haptics have been available through navigator.vibrate(), but iOS does not support this API. There is a workaround, though – the input element with the switch attribute:

1

<input type="checkbox" switch>

See this switch input in action here. On iOS, the input renders not as a regular checkbox but as a switch that can be toggled on or off:

Open the link on your iPhone, tap the switch, and you’ll feel haptic feedback. As far as I know, this is currently the only known practical workaround to trigger haptic feedback on the web in iOS.

Implementation

lochie cleverly bases his package on this single workaround and pushes it to its limits. The key lines are here, creating a switch input on the fly:

1
2
3

javascript

const hapticCheckbox = document.createElement("input");
hapticCheckbox.type = "checkbox";
hapticCheckbox.setAttribute("switch", "");

The package also creates an associated label and clicks it to trigger haptic feedback on the connected input:

1

javascript

this.hapticLabel.click();

(My understanding is that iOS will not trigger the haptic feedback when the input is programmatically clicked – so the ‘workaround within the workaround’ is to click the associated label instead.)

Custom haptics (!)

But web-haptics can do more than just trigger a single haptic ‘pulse’. The trigger function lets you pass in your own patterns with pulses of different duration and intensity, and even delays.

Want a simple success haptic? Trigger two pulses, as shown on the package homepage:

1
2
3
4

javascript

trigger([
  { duration: 30 },
  { delay: 60, duration: 40, intensity: 1 },
])

Or simply call trigger('success') as a shortcut.

Here’s an error haptic I made based on Apple’s human interface guidelines:

1
2
3
4
5
6

trigger([
  { duration: 40, intensity: 0.7 },
  { delay: 40, duration: 40, intensity: 0.7 },
  { delay: 40, duration: 40, intensity: 0.9 },
  { delay: 40, duration: 50, intensity: 0.6 },
])

lochie’s website even gives you an editor to make custom haptics. As you click and drag, both pulses and code update in real time. This experience is developer bliss:

Usage

Importing the package is easy. In Rails, for example:

1
2

# config/importmap.rb
pin "web-haptics", to: "https://esm.sh/web-haptics@0.0.6"

1
2
3
4
5
6
7
8
9

import { WebHaptics } from 'web-haptics';

// In a Stimulus controller somewhere
let haptics = new WebHaptics();
haptics.trigger('success'); // or 'error', 'warning', etc for built-in haptics
haptics.trigger([
  { duration: 40, intensity: 0.7 },
  { delay: 40, duration: 40, intensity: 0.7 },
])

Pass a debug option to the constructor to hear clicks in development/on desktop:

1

new WebHaptics({ debug: true });

Again, this package comes as close to perfect as I’ve seen. In fact, I already use it on Veritula: as you type in a text field, an error haptic plays when you exceed the max length (#4473).

I highly recommend web-haptics by lochie.

web-haptics: Haptic Feedback Finally Comes to iOS Safari

lochie just published a package bringing haptic feedback to the web: web-haptics. I spent four years at Apple working mostly on UIs, so I pay close attention to design and UX. In terms of presentation, use case, and attention to detail, this package comes as close to perfect as I’ve seen.

The problem

On Android, haptics have been available through navigator.vibrate(), but iOS does not support this API. There is a workaround, though – the input element with the switch attribute:

1

<input type="checkbox" switch>

See this switch input in action here. On iOS, the input renders not as a regular checkbox but as a switch that can be toggled on or off:

Open the link on your iPhone, tap the switch, and you’ll feel haptic feedback. As far as I know, this is currently the only known practical workaround to trigger haptic feedback on the web in iOS.

Implementation

lochie cleverly bases his package on this single workaround and pushes it to its limits. The key lines are here, creating a switch input on the fly:

1
2
3

javascript

const hapticCheckbox = document.createElement("input");
hapticCheckbox.type = "checkbox";
hapticCheckbox.setAttribute("switch", "");

The package also creates an associated label and clicks it to trigger haptic feedback on the connected input:

1

javascript

this.hapticLabel.click();

(My understanding is that iOS will not trigger the haptic feedback when the input is programmatically clicked – so the ‘workaround within the workaround’ is to click the associated label instead.)

Custom haptics (!)

But web-haptics can do more than just trigger a single haptic ‘pulse’. The trigger function lets you pass in your own patterns with pulses of different duration and intensity, and even delays.

Want a simple success haptic? Trigger two pulses, as shown on the package homepage:

1
2
3
4

javascript

trigger([
  { duration: 30 },
  { delay: 60, duration: 40, intensity: 1 },
])

Or simply call trigger('success') as a shortcut.

Here’s an error haptic I made based on Apple’s human interface guidelines:

1
2
3
4
5
6

trigger([
  { duration: 40, intensity: 0.7 },
  { delay: 40, duration: 40, intensity: 0.7 },
  { delay: 40, duration: 40, intensity: 0.9 },
  { delay: 40, duration: 50, intensity: 0.6 },
])

lochie’s website even gives you an editor to make custom haptics. As you click and drag, both pulses and code update in real time. This experience is developer bliss:

Usage

Importing the package is easy. In Rails, for example:

1
2

# config/importmap.rb
pin "web-haptics", to: "https://esm.sh/web-haptics@0.0.6"

1
2
3
4
5
6
7
8
9

import { WebHaptics } from 'web-haptics';

// In a Stimulus controller somewhere
let haptics = new WebHaptics();
haptics.trigger('success'); // or 'error', 'warning', etc for built-in haptics
haptics.trigger([
  { duration: 40, intensity: 0.7 },
  { delay: 40, duration: 40, intensity: 0.7 },
])

Pass a debug option to the constructor to hear clicks in development/on desktop:

1

new WebHaptics({ debug: true });

Again, this package comes as close to perfect as I’ve seen. In fact, I already use it on Veritula: as you type in a text field, an error haptic plays when you exceed the max length (#4473).

I highly recommend web-haptics by lochie.