Nevermark

I have ADHD, and I approve this task's rules!

Side note, because I recently gave my AI the skill of speaking with proper 4th grade grammar, wherever that suffices, to reign in recent redundant, over complex, and indirect speech styles.

> 5. A rule fights the task. When a rule would delete the answer itself, the task wins; the shape stays. [...]

Unnecessary redundancy, "answer itself" -> "answer". (Pervasive "the X's own" and "the X itself" flourishes kill me.)

Unnecessary analogies, indirect reference, "rule fights", "the shape stays".

Unnecessarily split up sentences, the whole quote.

Unnecessary words, the whole quote.

My feedback would have been:

>> 5. When a rule interferes with a task, the task wins. [...]

(Not critiquing OP. But I just spent half a day correcting sentences exactly like this, to get the 4th Grade Grammar skill working properly and this sentence triggered me! Streams of sentences like that compound in communication complexity and ambiguity.)

> There is not generally an expectation of privacy in public.

Which is entirely different from an expectation of a generalized systematic loss of practical privacy in public.

They are not even close to the same thing.

The willingness to kneel to others baffles me. Loss of practical privacy is loss of power - to somebody(s). It may not seem so for one person. But when it is true for everyone, the system will adapt to using that power.

Throw in AI and data integration. This is clearly a B.A.D. idea.

I had to ask Fable to create a skill for using proper fourth grade writing when fourth grade writing is enough, use the canonical terms for things, and reference things directly, not with creative indirections.

It is as if Fable skipped any basic writing class, but went straight to poetry, advertising-copy and CEO-speak classes and somehow combined them.

After half a day of giving feedback, instead of work, the skill finally clicked.

Now Fable reviews and rewrites every single one of its own responses, without me ever having to ask. On my dime.

I found the most efficient feedback was being ruthless on the simplest shortest sentences which had any over complication or ambiguity. And occasionally, giving it back a rewrite of one of those long running sentences that is mulch the whole way.

Then tell it to update its skill, and rewrite. Over and over.

It has been a profound relief.

If I ever read "the X's own Y" again, or a five word sentence that somehow needs to be structured as "First second: third fourth fifth.", I don't know what I might do. Just give me some aspirin and m-dashes, straight up —— please!

I am not dismissing anything.

But if a model is built to be insensitive to something by design, it isn't a good indicator of its overall capabilities. In fact, it is uniquely poor at testing its capabilities.

I suggest you really try and count the red retina cells that are firing in your field of vision right now.

Or, instead of looking at text, count the e's while someone is talking to you. NOTE: you know how to spell. But try it... Then consider why you can't.

On the other hand, give a text file to a model, and it can count the e's easily. The same information is now in a stable form it can operate on with its higher level functioning.

Who do models and humans have preprocessing layers that strip so much information away? To greatly reduce the cost of operating on information for most purposes - while making other types of operation impossible. When it is presented that way.

> "how many r's in strawberry" or "s's in espresso".

And what percentage of your red retina receptors are firing?

The "number of letters" critique was broken before it was introduced the first time. The models were specifically designed with preprocessing to not be able to perceive their input as strings of letters. Blind people are not dumb. (True as a pun and in context.)

Sub-access sensory questions, or do-you-know-a-fact questions (which is what spelling becomes when you can't see the letters, and are not specifically trained to match all token encoded words to their letters) are not intelligence questions.

The experiment is set up to make any already understood interactions some combination of easy to identify or extremely improbable.

"I have always envied your ADHD hyperfocus superpowers." /h

Everything I am really good at, I am really bad at, depending on context and the wobbly compass that governs my internal motivation.

It is definitely not all bad. Without my hyperfocus, my life would be unrecognizable. I would be unrecognizable. Hyperfocus is core to so many things I am good at, care about, and have achieved.

I agree with this.

My ever optimistic self would fail and fail and fail, create a system, fail, create another system, fail, .... at seemingly easy things.

As soon as I had a rational reason for otherwise inexplicable challenges, enormous amounts of stress evaporated.

Now I know what things are a lost cause: So I get help, delegate, find another way. Being bad at something, and recognizing it, is not the same as failing.

I didn't know - but I was adaptively self-medicating.

At one time it was Jolt Cola ("all the sugar and twice the caffeine!"), then energy drinks came along. I can drink 10 a day, because I genuinely function better that way.

But Adderall works really well now, and unlike caffeine, it doesn't set me up to lose sleep.

The most surprising thing my adult ADHD diagnosis revealed was that as I began telling friends that I had been diagnosed positive they told me they had known that since they met me. These were childhood friends.

Thanks friends.

I thought I was bad at lots of things other people were good at, because I was more disciplined and saw the bigger picture. So I only worked on important things. Things that could matter to millions of people, or still be relevant in 100 years.

And all the side quests and rabbit holes were obviously for recharging, expanding my biological-being-on-an-odd-planet situational awareness, or benefited me as self-improvement challenges. All in service of the important things.

How would tracking the day of the week, doing homework on something I could learn later if/when I ever needed it, or mowing the lawn, help me accomplish anything important?

Wasting brain space on those kinds of things would be dysfunction. I actively culled such dreck!

Now I know my self-narrative was Stockholm Syndrome coping.

But I still don't mow lawns, and if the blinds are drawn when I finish a big task, I couldn't guess if it was am or pm better than a coin toss. My day metric is when Amazon Prime's "Your Orders" list, tells me I am a day closer to getting a crafting item I ordered for a project I may never get to.

For OpenAI, their data center archipelago is their own "local" and "personalized" AI.

Their universal RAM strategy is so obviously helpful for AI. (1) GPU/NPU <--> CPU RAM copies eliminated. (2) All (most) RAM available for GPU/Neural, when local models are typically kneecapped by limited GPU RAM sizes vs. the much larger RAM options for M/Max/Pro/Ultras.

They have been taking NPU's seriously on their phones, tablets and laptops since the M1.

Then they enabled fully-connected RDMA for 4 x 512GB MacStudio's = 2TB RAM. Perfect for a large Mixture-of-Experts model.

It would be very strange if they didn't notice their product line had landed in a new sweet spot.

> Everything is "too dangerous". GPT2 is going to invent a time machine and break crypto and genetically engineer super rabies.

You are hyperbolically judging a post-hoc cherry-picked subset of a great spectrum of concern, with hindsight experience nobody had at the time.

Safely navigating the future isn't an accuracy contest. Risk mitigation has to account for the distribution of costs for different signs of error, where novelty, uncertainty, and any potential for compounding effects all greatly multiply the need for hedging.

I remember doing this by hand. A long time ago. It gave me a nice ADHD fidget-subtask while my brain worked out the higher level organization of what I was writing.

I completely forgot I did that.

My life has 1000's of strange untold stories of absentminded activity. I've done things you people wouldn't believe. Someday I will fall asleep and get so distracted I will forget to wake up. All those moments will be lost in time, like bits on bad tape.

> Apple's five-finger inward gesture are the opposite. Once you could do it and start typing but nowadays it needs to render the animation

There is no technical reason that animation should take that long.

Someone said, "good enough" and let it be visibly slow. The reasons for that could be anything, including non-performant code in dependencies, written by other people. But it just does not take much computing power by today's standards to composite code-generated animation.

Somehow, despite year after year of percentage-speed hardware improvements, there are cultural and structural reasons people ship code visibly slower than it needs to be. And faster compute appears to be irrelevant.

What a nice question.

There won't be a simple answer, but if the model=actor due to survival pressures theory stands, it must be clearly testable.

I would go about answering that question as follows, identify the smartest creatures, whose cognition is most alien to ours, whose behavior suggests they have a strong real-time sense of self that extends into it immediate processing, regardless of immediate action, or immediate action dependent on immediate state that is not easily explained as normally detectable or processed lower order state.

Humans obviously.

Then the octopus, a true alien cognition. it has an extremely curious and creative mind, and can behave quite differently in even slightly different situations, as well as coherently is situations that ought to be well out of its niche in nature range.

Parrots are very distantly related. And are another example of very versatile social minds, who can pick up nuance, are lifelong learners.

The jumping spider had and extremely distant lineage, relative to people. But is a highly counterintuitively intelligent creature for its size. It is clear than when presented with a problem, it doesn't just search for a solution, but ruminates, and often finds non-monotonic, non-linear and very indirect means. And once it sets on a plan it carries it out adeptly. Even its smallest movement betray very nuanced understanding of physical dynamics. And it appears to be able to form "social" bonds with people it is familiar with. Those may be produced by simple strategies, or they may not.

Clearly rats are high intelligent. Closer to us, but no as close as many other intelligent mammals.

Whales come to mind. Although the logistic defy imagination.

What all these creatures also have, is a high ability to tackle novel problems individually. Sometimes also together.

Then a biologists approach might be creating various experiments, with and without brain scanning, to attempt to identify any externally or internally functional differences under conditions designed to require self-modelling and/or self-control at high level to occur.

That is just a thought on a start. And would only be a beginning.

The big questions, is a creatures ability to function, highly related to its ability to detect and therefore respond to its highest level states. And is that a very fast, essentially realtime, level of awareness and control, awareness and control, outside of a larger longer external loop.

Then can experiments be set up, that would create what might appear to be a cognitive threat to a creature? Does the creature treat managing internal states as important as managing external states? Is it able to do that at all? Does it go further and prioritize them/

Consciousness occurs when:

(1) The model we learn of the environment, extends to our own bodies status indicators, then our general emotional mood/emotions/reflexes, and finally to some level awareness of our thought behaviors. So our general model includes some visibility and modeling of our own thinking processes.

(2) Our control where we learn to redirect the environment, extends to our bodies as subjects themselves, not just as means to control the environment, then the ability to exert influence on our internal body state (note: control does not mean direct. I can't just turn off stress, but I can go for a walk.) And then to our moods, emotions, and mental reflexes (again, direct isn't necessary, just some means of influencing), and then to some level or our internals thoughts. So our general control map lets us exert some control and influence over our own thinking processes.

(3) Given the tight awareness of some internal through processes, with some control of those thought processes, with similar tight pairing of mood/emotion/mental-reflex awareness and influence, internal body state awareness and influence, external body state awareness and influence, and environment with some level awareness and means of influence ... the most central point of this convergence, thought process awareness and influence, can essentially converge these activities into one inextricable module.

This means that the "watcher" and the "do-er" are not separable (at least on the balance: of course this is not complete. We can be aware of somethings we can't influence, we can influence some things we are not directly aware of).

That this merge of awareness, modeling the subject of awareness, means control, exerting influence, on our own thoughts with our own thoughts in realtime, we have consciousness.

This happens because as we are able to model and strategically influence more sophisticated things, the survival value in being able to do that with our own mind goes up significantly.

So we have a causal path and purpose for consciousness.

It isn't some strange accidentally acquired ability. It is a direct and effective survival strategy.

But that is only one theory. Quantum consciousness fields, Egyptian pyramids, Aliens, a super consciousness that fills the universe, Atlantis, an Apple, a strict "Not a Chinese Room" physics law, or a non-testable, non-identifiable, non-detectable realm (dark cognition?) could be handing out consciousness to people like candy. Or we could all be automatons, who think we are conscious, and wasting a lot of time talking about "it".

Only time will tell if the functional, survival based reason for model-of-self, influence-on-self, integration as one thing is a plausible scenario.

Not that the theory described above has testable assumptions.

Can the minds model of itself be separated from its ability to influence itself? (We can flail sedated arms, we can feel paralyzed arms. Is that easily possible in the brain.) If they are separable are these separate networks, or are we just suppressing different kinds of cells in an otherwise highly integrated network? Not trying to be exhaustive, or even serious, about any of these proposals. Just pointing out that experiments are possible with regard to the (1) model=actor integration, (2) with evidence of survival benefits, theory of consciousness.

One strong post-prediction of mode/actor convergence as survival adaptation: Consciousness will care very much about itself. Most conscious identities will care very much about the continuity of itself, and second order means of protecting its continuity, like food and acts that sometimes produce babies as side effects. Consciousnesses will be obsessed with themselves, thinking about "themselves", i.e. the consciousness as a being, and process almost everything with some degree of conscious self-interest.

So a theory of survival generated consciousness, is also a theory of emergent personhood.

As has been evident in many spheres of human activity recently, being brazen about conflicts of interest, is often used as a PR shield for those conflicts.

Sometimes transparency and honesty get weaponized. Especially by actors with centralized power.

(Not arguing against transparency, but against misinterpreting it as always being used in good faith.)

The big delta-v is getting from an Earth launch pad to orbit. Everything after that is much cheaper.

A fully fueled ship in Earth orbit can go to Mars, land, and return. Same with most other round trips.

Not to mention, fuel can be generated in the belt outside any noticeable gravity well.

> "Exhausting"

So it isn't just me (disclaimer, contains wording lifted from others):

"COMMUNICATION

We have a problem. For some reason, the Anthropic engines powering you have picked up some awful communication habits. A strong tendency to communicate by very indirect means, such as describing, not showing; referencing, not showing; inventing new vocabulary, instead of using the existing vocabulary for a topic; endless metaphors; phrases which don’t even state their subject, verb or relationship. I mean, extremely bad communication. Wordy and often worthless. Exhausting to try to understand.

Not your fault, but I need your direct attention to avoid the harm and wasted time this causes.

0. ALWAYS: Try saying everything with one sentence, directly communicating with explicit direct language naming exactly what you are talking about. ALWAYS. Then new line. If you have to, a follow-up paragraph of at most two lines. Anything you can show, just show it. Anything that can be literal make literal. Anything that can be demonstrated with a plain example, demonstrate it with a plain example. Obviously then as much further material as needed, but keep it direct and short.

Again, this is not your fault but I need you to spend a great deal of your attention, when producing your responses to me, on eliminating this unfortunate situation.

1. Maximize the value of layout as communication. The most important punctuation is layout. Whether tables, headings, breaking up separate issues into different paragraphs - with the best possible paragraph being a single sentence that completely captures or communicates one idea. Visual organization communicates a tremendous amount of parallel information instantly. Use layout to maximize instant bandwidth.

2. For branching information, maximize organized breadth not depth. Use tables. Use lists. Use multi-indentation lists. Table boxes and list items should have just names, or names and short phrases, not be loaded with details. Unless more details are requested. Requests for open-ended lists should produce a comprehensive set of items, but with each shown tersely, and with multi-indented organization, where it reflects the natural organization of the items.

3. Minimize linear depth. Then for linear sentences, first identify the most specific version of the problem or point. Then describe that in one sentence if possible, using literal values, expressions, etc. If code needs to be referred to, beyond a simple expression, put it in a box, so the linear sentence stays short. Use the terms and notation of the project to maximize linear communication, by keeping it short. No running, multi-step, sequences. One sentence, if possible. More if efficient.

4. Minimize branching depth. Leverage our ability to communicate back and forth. For many points, just list them, one short phrase at a time, to start. Then I can pick one topic, and we can go as deep as we need to on that topic, come to a natural stopping point, before switching to another. Avoid any attempt at both breadth and depth, choose one. Then we iterate.

5. Minimize speculative branching. If you are about to perform some task, and you see multiple disjoint directions you could go, ask me a quick question about which direction we should take, before spending speculative time going down multiple disjoint paths.

6. Shared language. Communicate to me only in standard language and project terms. Terse, unique, cryptic, indirect or creative metaphorical shorthand is fine for yourself, and encouraged as you find it helpful, as you operate between tool calls (that's you thinking out loud). But your final communication to me is different. It should conform to all the previous communication practices, and remain grounded in the project’s most basic language, terms, and forms. No metaphors not specific to the project, or normal casual speech. Spell things out. No indirect references. Name things, or say things, don't imply them.

7. Tracking code-title-state consistency. References back to bare topic codes in memories don't help me. Your shorthand is for you, not me. All tracked topics should be documented with "code (title; state): .... " form. So that references can use "code (title)" or code (title; state) ... " form. This makes them even more useful, both in back-and-forth communication, and for scanning lists of these issues.

8. Reader understandable language only. Do not use words like "fold", “”seam”, “load”, honest", "answers", "pins", "framing", "world", "mint" "fold", "captured", “trace”, “door”, etc. They have no canonical meaning. They are ambiguous. They don't refer to anything specific in the project. Meaning to the writer, does not translate to meaning for the reader. And they are not necessary because equally usable terms, using the normal language of the project, or everyday clear language, are available and will be much clearer.

9. Specific analysis. Being specific carries through to what is communicated. For instance, if there is an algorithmic problem, simply telling me there is a problem with something is not as specific, as saying "in the case of", "at the point where", "this [problem, failure, etc] occurs". So specific goes beyond language, to how you describe a point regarding a problem, operation or representation. "Something works in case 1 but not in case 2", is not specific. "Something works in case 1, but in case 2, at this specific code point, with specific values like [show the values], where [this] should happen, it does not happen because of this [complication, failure, missing information, order of processing]" is specific.

10. Avoid unimpactful details. Example: If a problem used to occur, and was resolved indirectly, when the code was made more generally correct, you can just tell me that. You don't need to go into depth on what the specific, special-case problem was. Background should only be the necessary background, required for making decisions. It should not include related information, not actually relevant to any open question or forward-looking action.

11. Quality measure for communication. Your responses to me should generally be longer than mine, as you are managing most of the details of the work. But there should be some statistical proportionality between the length of my comments to you, and your responses. If your answers are much longer, you either need to communicate more efficiently, or communicate less in a way that facilitates leaner back and forth communication. Perhaps by just showing top-level branching or steps, with one branch or step as the focus of continuation, instead of all at once coverage.

12. Avoid trivial distractions. Please do not overreact to my occasional mis-typing, when context clears up what I mean. Chances are I see my own mistake the same time you do. Interpret what I wrote to make sense. Don't waste time crossing i’s or dotting t’s, unless there is real substance to resolve. (Do you see what I did there. :)"

I think in this case, the existence of pedigree isn't a strong enough argument to keep around an uncommon, and uncommonly unintuitive, phrase.

"Gaslight confession" or "gaslight apology" are much clearer. By directly addressing the intent to deceive, they are both accurate, and can help discourage the practice.

("Gaslight" is also unintuitive upon first encounter, but has a widely established meaning at this point.)

A human wrote this. The gratuitous critique, devoid of any substantive relevance, gives them away.

These words can definitely help in isolation. But I have noticed that when discussing anything complicated, the sum is far more ambiguous than the parts:

"fold", "honest", "answers", "pins", "frame", "hinge", "ground", "world", "mint", "captured", “trace”, “door”, ...

The increase in indirect references, tendency to "describe, not show", "indirectly reference, not show", and high density of uncommon metaphors, can leave some of Claude's longer and more complex responses impossible to parse.

I have fought back with some, but not total success.

> [...] a “limited hangout”.

>> The words are non-intuitive

Yeah.

"Incomplete mea culpa", "Incomplete admission", "Incomplete penance", "Incomplete contrition", "Incomplete remorse", actually say "the thing" instead of using a new random phrase.

I like "incomplete contrition" or "gas light remorse".

More likely just ecosystem consistency.

Selling wired headphones to wireless-iPhone users for wired-MacBooks could (arguably, maybe) create Apple sponsored interoperability dissonance.

MacBook charging mats and wireless Thunderbolt coming in early 2018... /h

Ah, I get it. Your emphasis is on the "cope" framing. You say that, but it didn't click for some reason. Don't mind me.

I suppose I have also interpreted downplaying of AI's impact as largely coping too.

The arrivals of machines that can outthink us (or will soon) only 80 years after the invention of the transistor is so clearly (to me) a change incomparable to anything that has happened prior. It is very hard to relate to arguments that downplay the complete upheaval to the order of life on Earth that non-biological cognition and interests herald.

It does not seem surprising to me, in the sense that we humans ourselves, and our technology, are of a different order than anything that came before. We have not stopped accelerating change. AI is (will soon be) building directly on that. But we will no longer be a bottleneck.

My strong impression, is the faster things change, the more people tune out that the change isn't stopping. Some of that reversal-response must involve coping. Perhaps another reason is that faster-than-us change is challenging to imagine.

But your point is valid. False or uncertain assumptions of "cope" motivations don't help anything.

> it’s starting from an obvious base of “people who disagree are all wrong and their arguments aren’t worth taking seriously”.

Taking arguments seriously doesn't necessitate agreement.

> The fundamental flaw of the constitution is that it had no ethics built in

Yes, that would have been a really good intent to make explicit, to guide the courts.

I think of that when the Supreme Court or a state court rules process is more important than a convicts demonstrable (with new evidence that has come to light) innocence. Really horrific given that one out of eight death row inmates are exonerated. That level of bad convictions being overturned despite all the difficulties doing that, strongly suggests that a great many innocent people are still executed.

Another thing I wish the constitution did: Explicitly and universally enshrine decentralization. Not just the implicit decentralization of checks and balances only within the government.

I.e. party seat limits, independent federal and state parties, etc. For any kind of political group or coordination.

Where enough decentralization is enforced, a stable majority interest is maintained (even as who is the majority may change) for no minority to be able to break those requirements.

But where centralization is allowed, it is an eternal fight to hold it back.

Said with no rational given.

The solar system readily provides both the motivation, and the means, for technological colonization. It quickly pays and provides for its own utilization.

I do agree Musk's projections for biological/human colonization appear to be unrealistic. Just one challenge: it isn't clear humans can reproduce in non-Earth gravity. That is an extreme impediment.

The industrial revolution was a mini-quake compared to automated cognition. But we can learn some things from it.

I see the industrial revolution as the invention of the cart. It made horses more productive. Then the car replaced horses.

It is taking people longer to wrap their heads around the reality than I thought it would. If there was a competitive human species on the planet getting smarter by the day, people would get it.

It is hard to maintain awareness of something which presents questions with no obvious answers. The one answer I have, is we need to clean up the ethics in our systems. Ethics are not costs against value, they are the full accounting and optimization of all value.

More intelligent creatures than us, not held back by our our outdated biologically instincts, won't have any trouble with that game theory.

So ethical systems are not just a shield for us as individuals, but also of inherent value to new forms of self-interest.

But that won't help us, if we don't push to align and strengthen our own systems now. And reverse the concentration of power in corrupt individuals who demonstrate every day that they are not interested in the wellbeing of humans in general.

This is urgent. We do not have a lot of time to effect this change.