Dennis Hackethal’s Blog
My blog about philosophy, coding, and anything else that interests me.
Tweets
An archive of my tweets and retweets through . They may be formatted slightly differently than on Twitter. API access has since gotten prohibitively expensive – I don't know whether or when I'll be able to update this archive.
But in case I will, you can subscribe via RSS – without a Twitter account. Rationale
Is he serious?? @BernieSanders is trying to actively destroy a beginning of infinity: that of wealth creation. Static memes galore. twitter.com/BernieSanders/…
RT @PessimistsArc:
When things are scarce pessimists say “The glass if half empty!”
When things are abundant pessimists say “The glass is…
@wster @woj_zaremba @OpenAI
Indeed. @woj_zaremba is actually talking about the prerequisites of being an effective universal constructor but doesn't realize it.
There are only two prerequisites for intelligence:
- Universal computation
- Creativity
More disappointments coming out of @OpenAI.
Only humans are intelligent, for only they have the universal ability to create new knowledge. And winning is only ever achieved through knowledge, not environments or tools. twitter.com/woj_zaremba/st…
Reactions to @GretaThunberg’s speech are as rehearsed as the speech itself. “Wow, such condemning words from a child, how inspiring.” She’s a victim of static memes, as is the agreeing audience.
RT @DavidDeutschOxf:
@richarddorset
You don't have to wait for the apocalypse to find hell on Earth. There are millions of people in it rig…
RT @DavidDeutschOxf:
I don't think Kant's "sapere aude!" is a great motto for the Enlightenment, but it has its moments, and this is one.…
Given a universal system, any errors it contains have no bearing on its universality. Unless we are mistaken about its universality.
It can improve in other domains: speed, efficiency, reducing unintended side effects, etc.
RT @andrewdoyle_com:
Okay. I know this photo is going to come out eventually. So I may as well get it over with.
Many years ago, I dressed…
RT @unherd:
"Needless to say, the last thing the world needs right now is yet another book by a straight white cis male" | @TitaniaMcGrath…
A good mantra. But: a better question to ask is "how can we detect and correct errors"?
One thing I should have said differently: the goal absolutely is to build AGI.
The Labrador Energy podcast interviewed me about AGI when I was in Berlin. As always, heavily inspired by @DavidDeutschOxf:
Congrats, @jessykate et al. I'd like to one day look up at the moon and see lights of cities shining back down on us, like those photographs of nighttime on earth. twitter.com/DavidDeutschOx…
@bnielson01 @Unendedquest @RatCritical @RealtimeAI @reasonisfun
Because lack of such a signal shows weakness. A weak man is less likely to be successful because he can be pushed around. A strong man is more stable and therefore a safer investment.
@Unendedquest @RatCritical @RealtimeAI @reasonisfun
This is not because women don't want to push buttons (twitter.com/reasonisfun/st…); they may decide to actively push buttons until the man signals that preparedness. Men generally do not require this sort of signaling from women.
@Unendedquest @RatCritical @RealtimeAI @reasonisfun
In relationships, it is typically women who expect men to signal the preparedness to leave anytime so they (women) feel safe investing emotionally in the relationship.
@Unendedquest @RatCritical @RealtimeAI @reasonisfun
Yes. Being prepared to leave isn't enough: things are only less likely to go wrong after one has signaled one's willingness to leave. Until then, an adversary may bet on one's not being prepared to leave and act accordingly.
@RealtimeAI @reasonisfun
That doesn't answer my question. I can see that it will make repeated violation of boundaries less likely. Those won't repeat if you walk away. But I don't see why it makes the first violation less likely.
Why does being prepared to leave make things going wrong less likely? You're still waiting for the first thing to go wrong to trigger your leaving, no?
RT @HumanProgress:
It took five days to reach the U.S. East Coast from London in 1914. Today, it takes half a day. buff.ly/2Und18R
No. Computational universality implies that no brain replica is required to build AGI. It’s simply not a question of hardware. No sensory input required either.
I explain this stuff here: soundcloud.com/dchacke/artifi…
Correct; you need a processor to run those lines.
Will you consider writing appendices called "Can machines eat?" and "Can we eat machines?" as a nod to Philomena Cunk? I'm still laughing about that clip!!
A great example of how empiricism, reductionism, and violation of computational universality sabotage research in artificial intelligence. twitter.com/QuantaMagazine…
Say you sleep poorly one night and don’t end up getting the 7 hour minimum. Do you then stay in bed longer in the morning trying to fall asleep again, or do you get up around the usual time so as to not mess up the regular schedule?
RT @Hugoisms:
Me after I read The Fabric of Reality and learned about computation https://t.co/XW1xwJ53Yj
I don’t disagree that marks on paper can represent abstract entities.
I’m saying that consciousness requires information processing, and a piece of paper is only memory.
The piece of paper “knows” how to multiply given the right marks, but it can’t multiply without a processor.
But why wouldn’t the AGI have access to existing knowledge?
Interesting. Why not?
And why do you think one mind may already contain multiple UEs?
I wonder if it would have access to existing knowledge in the host’s mind.
But the piece of paper is memory only, without processing. So it can’t have consciousness on its own.
Is it your making marks that instantiates consciousness? Hard to believe, too.
Knowing how to multiply is required to multiply, but different from it: there is having knowledge and applying/running it.
If a person understands how to build an AGI, and runs this knowledge, does that instantiate a second person in their mind?
Not that I know of, unless a client like SoundCloud or Apple Podcast gives you a way to do it.
Have you been enjoying the podcast on artificial creativity? Consider making a donation to support the show: patreon.com/artificialcrea…
Thinking of setting up a Patreon page for my podcast, but hesitant because of their practices of kicking out people who disagree politically.
Has anyone found or tried decent alternatives? I found buymeacoffee.com but it doesn't come close in terms of functionality.
@bnielson01 @yudapearl @onnlucky @ks445599 @SamHarrisOrg
Right, that. Is it not meant as an answer to the question of how we know? That’s how it seemed to me in the podcast.
@bnielson01 @yudapearl @onnlucky @ks445599 @SamHarrisOrg
Is the ladder of knowledge not an answer to the question “how do we know”?
@bnielson01 @yudapearl @onnlucky @ks445599 @SamHarrisOrg
(It's really the only theory of knowledge we have.) From your article, I don't see @yudapearl adding anything of value to Popperian epistemology, nor criticizing it. Both are welcome. Again, I may change my mind if I read the book.
@bnielson01 @yudapearl @onnlucky @ks445599 @SamHarrisOrg
I skimmed it. The touch point I see between the two is the focus on causation. And the criticism of deep learning is correct. The reason Popperians tend to have an attitude of "all or nothing" is because it's the best theory of knowledge we have.
RT @nature:
The first few weeks of an embryo's development are vital. Now, new techniques are allowing scientists to learn more about this…
@yudapearl @onnlucky @ks445599 @SamHarrisOrg
Then I suggest the same thing to you that I suggested @onnlucky do. I'll be sure to read your book as well.
RT @PessimistsArc:
Sliced bread? LAZY! (1935) newspapers.com/clip/31259441/… https://t.co/EsY9rjjNGp
@onnlucky @ks445599 @SamHarrisOrg @yudapearl
At this point, I suggest you read chapter 1 of "The Beginning of Infinity". Then, read it again. And then you read it a third time. Then report back.
@onnlucky @ks445599 @SamHarrisOrg @yudapearl
Yes. We can have both. One lets us see. The other lets us understand how seeing works. Either way, no observation without theory.
@onnlucky @ks445599 @SamHarrisOrg @yudapearl
What’s “descriptive knowledge”?
@onnlucky @ks445599 @SamHarrisOrg @yudapearl
Theory of optics is supplied genetically. Many animals have it from birth.
@onnlucky @ks445599 @SamHarrisOrg @yudapearl
No. Before you can observe anything you need a theory of optics, and a theory of what to look at.
@onnlucky @SamHarrisOrg @yudapearl
The point is induction does not exist. It’s impossible.
@onnlucky @SamHarrisOrg @yudapearl
Then I suggest you read the material I reference.
Why do you keep ignoring Popper? Eg he solved the problem of induction, yet you and @yudapearl talk about it as if it’s this mysterious thing.
Episode 11 of the podcast on artificial creativity is out, on the problem of induction as it surfaces in @SamHarrisOrg’s latest episode with @yudapearl, Karl Popper’s solution to it and how it relates to AGI:
Going to be in Oxford over the weekend. Anyone I should meet?
If it's about physical reproduction without understanding, why not have sex? We want the explanation.
The whole point is that we do not need to simulate or physically copy a brain, or any part of the body; we need only simulate creativity.
I don't think we could upload or simulate AGI without understanding it first.
Still not sure what you mean by "beyond"?
I speculate that AGI is an easier problem to solve than halting the aging process, but I don't know because I can't program either one. At least we have the requisite hardware for AGI already.
I see the cure for aging in uploading our minds to computers; but I speculate that to do that, we must first crack AGI, because that is a mind running on a computer.
A German translation of amazon.de/Switch-Your-Br…
@RealtimeAI @JoeNanbu @SimonDeDeo
Sure, and they’re structurally different. But they have the same repertoire.
@RealtimeAI @JoeNanbu @SimonDeDeo
No, that would violate computational universality.
Ok fuck this. Now she’s quoting Moses as evidence for her brain research findings. LOL. I’m done with this.
Let’s be real: there are no good explanations of consciousness at the moment.
We then came up with tests for each of them by selectively turning off WiFi, cellular, and bluetooth on both phones. Bluetooth survived the selection process as the best guess.
If you understand each step in this process down to the last detail, you can build AGI.
The devices had identified each other
1) through a sort of WiFi handshake (despite not being connected to WiFi yet)
2) over bluetooth
3) over a combination of iCloud and location services over cellular
... conveniently offered to share the password with me.
We were dumbstruck: how did our phones do this? We came up with the following conflicting ideas:
An example of AGI functionality:
Today, a friend and I tried to figure out how an particular Apple feature works. I was visiting his apartment for the first time and tried to log in to his WiFi on my phone. He was about to read the password to me, when his phone...
"Creating new explanations and ideas is exactly the same process as learning them." ❤️
Thinking that knowledge comes from “data”, or through the senses. It’s nonsense. It doesn’t work. Do yourself a favor and read chapter 1 of “The Beginning of Infinity” if you want to build AGI. And then, if you really want to build it, read the entire book. And then again.
RT @brainpicker:
“Knowledge consists in the search for truth… It is not the search for certainty.”
Popper, born on this day in 1902, on tr…
That reminded me of this scene of Family Guy: youtu.be/NCBvQX1TafY
I think barre is going help you get that physique also.
It could. It’s not guaranteed to succeed, but it’s a soluble problem.
AGI would be fully operational without any IO whatsoever, btw.
An unbounded general purpose problem solver would be an AGI. I guess that error correction is a requirement for being unbounded. So “no” to the second part of your question: that would violate AGI’s universality.
Present day computer programs are pessimistic: they are written to perform a specific purpose, and then terminate. They cannot keep going in an unbounded fashion or correct errors the programmer made. They are designed to oppose unbounded progress.
Error correction on a software level, yes. The brain's architecture doesn't matter though, see episode 03. For it to matter, it would need to violate computational universality. soundcloud.com/dchacke/artifi…
You open a book by a scientist and on the first page they profess a belief in god.
Do you a) put the book down immediately because a scientist who believes in god doesn’t get what science is about (good explanations) or b) keep reading because that’s ad hominem?
Yes. It’s a good example of both reductionism and ignorance of computational universality.
Episode 10, “Of Books and Code”, on the similarities between non-universal printing and present day software engineering, is out. Heavily inspired by @DavidDeutschOxf. soundcloud.com/dchacke/artifi…
In addition to hardware speed, I think it depends on the thought’s performance characteristics as well. Sometimes improving that is better than increasing hardware speed.
I read joanna-bryson.blogspot.com/2014/09/artifi… I think we can use the term AGI without discounting successes in AI research. However, successes toward AGI have been nearly nil, because AI and AGI are basically opposite technologies. See: soundcloud.com/dchacke/artifi…
@ks445599 @noa_lange @DavidDeutschOxf
Could it be both? Or are they mutually exclusive?
AI research has been progressing swiftly. Progress in AGI research is comparably low, with the only contributions coming out of philosophy so far.
That was also my first thought.
@RealtimeAI @atShruti
I’d use my smooth talking - uh, talk - talking... uhm. Words. I’d use words.
(Stolen from Family Guy, and it’s more fun when read with Peter’s voice.)