The only thing needed to make a universal AI is …

The ability to generate a workable plan for any situation.
A human of average intelligence could do this for any task where they could command enough resources.
If an AI can always make a plan, it means the plan could always be further improved.
That implies the AI could do anything that can be done.

You could also ask an average human to solve any problem. They would resist because it’s too annoying, but in theory you could pay absurd amounts until they pay attention. It could take a billion dollars even for someone living in the worst slum in Haiti.
But then they could still perform any task better (as defined by being able to lead) than the smartest current AI-like software. In finite terms they would still suck, but they could always get closer to the solution.

The difference is universal planning ability.

Advertisement

Cryonics-adjacent question

This question involves this blog’s major theme: my somewhat bizarre “mind backup” project, originally inspired by Rudy Rucker’s “Lifebox” idea.

Unfortunately, I made some HUGE blunders in the last years involving my life savings and the market (bonds, though I almost wish it was crypto). It’s painful to admit this involved laziness on my part (I hate hassle of any kind) as well as bad judgment. As a result, my finances are such that I can’t see a way to pay for cryonics when the time comes (or retire, but that’s a separate issue).

My question is: could I arrange to mail a hard drive (or future hyper-DVDs) to a company like Alcor after my death, containing all of my “mind backup” data. It’s basically hundreds of thousands of disjointed diary entries and diagrams and flowcharts and sketches and lists and preserved records and pics and receipts and instructions and whatnot.
They would then store this media instead of my frozen corpse (which may be disposed of in a pauper’s grave or ideally used for medical research).

Obviously, they should charge less to store a small box of disks for five hundred years at room temperature (if that’s how long it will take), than an upside-down frozen head in a canister of liquid nitrogen that needs constant topping up.

Once future software has become advanced enough, this data could be used to make a crude “copy” (https://www.lesswrong.com/posts/EKnDXLxqLAtkkix4w/personal-imitation-software) of my mind; and hey presto, back from the dead.
It’s not much, but better than the absolute certainty of oblivion, which is the alternative I would get by doing nothing.

I’m not the only individual this question applies to.
Many readers of this site already have huge collections of digital data, which will become much larger before they die.
They would not want it all to disappear after they’re gone.

A new online novel about the Singularity happening in the very near future

Singularity Soon (2022) describes the events of the last weeks leading up to a Technological Singularity that takes place in our time, and exactly how it happens.
Many human mistakes are made during this critical transition. The process is partially driven by some very bad behavior and serious personality flaws.
Unfortunately, we can’t guarantee the most critical period of history will be the best managed one.
THE INDEX:
https://www.wattpad.com/myworks/263500574-singularity-soon
CHAPTER ONE:
https://www.wattpad.com/1045211713-singularity-soon-1-meta-dream

Mind Backup should be anticipation-based

For 40 years, I have suspected that if you could describe a perception in sufficient detail, you could capture it. This has influenced my thoughts on Mind Backup research. Don’t try to map the physical patterns of your brain, only map the things that your brain loved. Make a very long list of all the things you liked. Record everything in your life of any value, at the maximum level of detail.

Then combine all that data in a virtual memory palace. That would be a virtual reality simulation of a “museum world”, dedicated to your life and mind. It could be located on an eternal island.

The point is that you would design it ,and even begin to “inhabit” it to some limited extent, while you’re still alive. Through virtual reality you can begin to experience how it will feel after your body has died, and your mind has been “recreated” as an AI. You would prepare every detail of your “arrival” and habitation there while still alive. It could be like planning and looking forward to a vacation.
Then your “continuation” would really be rendered as an AI after you die. And then you would spend “eternity” in the virtual memory palace you designed while you were still alive. This could happen even if you die before human-level AI has been invented.

You would have to do a lot of work preparing your virtual memory palace. It would be like an endlessly expandable museum world. There would be separate facilities dedicated to all the places you lived (which would be recreated there), all your schools, where you worked, neighborhoods and all other places. All these museums would be highly biased, and mostly focus on the good and meaningful things.
It’s relatively easy and fun to create a giant layout. It would be able to import some data from a shared VR model of the entire world, incorporating all human knowledge. But you would still have to spend a lot of effort sorting out and entering all the data of your life.

The only way to get people to put in the effort is if it’s an easy process. They have to feel they are making good progress right from the start. The first step in creating a virtual memory palace would require software that could probably be written right now.

What is a Mind Backup Supertest?

It’s something you could do. You provide the most essential kind of information about yourself. That information might then eventually be used to create an Artificial Intelligence or software mind that would have your personality.
It would believe that it IS you.

What I’m trying to do is said to be totally, completely, utterly impossible. We can’t even begin to create artificial human-level minds yet.
The human mind is said to be inscrutably complex. It is completely impossible to “back up”, like you would back up computer data.
But still I want to do that.

A Mind Backup Supertest is a method to extract the maximum data about your identity using the fewest effort and time.
The best idea I have so far is to write many independent short articles about specific aspects of your life.
You have maybe ten thousand ‘keystone’ memories. Each is a specific detailed setting, a kind of image of a situation from your past. It is associated with a place and a setting. Places you visited, environments you lived or worked in.
The feeling these memories trigger is real, even if the resolution is a bit low.
Each could be separately described in a short article. These will all be added to your “Lifebox”.
Anyone should be able to reconstruct part of their key memories, even if they have lived traumatic lives. Parts of the past can be made “safe”.

But what is the best Mind Backup article idea?
Probably a full description of your personality, the way that you think you are. This article might have to stay secret, as it also describes your weaknesses.
You would simply have to write down what it is like to be you.
How many different things are you thinking of at the same time and in what way? What are all your concerns at every level at one particular moment in time?

Imagining new image formats

I’ve long begged for something better than keyboards, or just a better keyboard, but that’s not all that could be improved.
It’s time to invent better ways to visualize reality, like my patented Spheroid Display (not talking about that horrible webp format that you can’t even view after saving).
One new format could be optimized to display human faces and bodies. The effect could already be simulated with those flickering gif animations that alternate photos from different angles to simulate a stereoscopic vision (also fun gifs here http://www.reddit.com/r/woahdude). To a large extent, it would obviously be used to map and represent the faces/bodies of unattainable females.
A 3-D, layered, mark-up format, using common attributes to render realistic heads and bodies, with layers for shape (skin/muscle/skull), texture, color and so on. You could rotate them through any angle without loss of resolution, every pore and wrinkle simulated with reasonable verisimilitude.
This system could combine several photos to create higher resolution composites. Call it a .3dfce format. Starting simple but steadily improved. Expressions could be defined, along with all the poses of the rest of the body.
It might be awesome for the increasing number of males unable to get female partners due to the accelerating imbalance in the “sexual marketplace“.

As I was saying . . .

It’s hard to imagine in such boring times as the ones we’re living through, but there might be increasingly interesting years ahead. With months of intense change that would feel like science fiction.
I’m not talking about the immediate but the intermediate future.
–
The future can only become so fantastic until it becomes meaningless
Before the Technological Singularity, before the Human Event Horizon, we must pass the Tipping Point, when the rollercoaster is first released into the abyss.
This may resemble the first moment of personal self awareness, experienced by some before the age of 3, or the moment someone fully wakes up and realizes they’re awake, or the Bicameral Breakdown about 3,000 years ago, or when human nature changed in December 1910.
–
Of course those last two comparisons are just made up. It should be more profound than all of them put together.
Between the Tipping Point and the Human Event Horizon lies the Accelerando, an age of economic growth that would make everyone as rich as they ever wanted. (This is the optimistic possibility, you don’t want to know about the other one.)
–
The Tipping Point may be as simple as the moment when interface technology stops getting worse. (One problem is that under the current paradigm programmers are paid to actually be evil.)
Lightweight visors and earbuds, holographic display walls, something better than keyboards, software that can interpret and anticipate user demands. To serve their users, future operating systems must become part of them.
Ending human poverty might require a method to effectively make people smarter with human mind extensions, software to convert everyone into their ideal selves.
That would be wonderful to have. Progress would actually become easier than failure.
–
The invention and development of mind interface technology would be followed by human-level artificial minds. Then artificial hyper-minds may begin to analyze and reconstruct deceased human minds from the past, using whatever incomplete records are available. Direct brain scanning methods may analyze the tiny connections that form a (likely deceased) human brain to recreate it as software. Almost a billion people die every decade. Some or most would want to live on in cyberspace.
–
It still seems like a long way off. About now, we may be entering the upslope with a jerk, the coaster just beginning to climb up the railway.
Or we may still be waiting in line. Or just dreaming of going to the amusement park.

* Timeline of the Future
* Future of Deep Learning

Prediction: real estate values MUST collapse

Eventually – not in our present epoch, of course. It’s certainly been predicted for a long time.
Meanwhile, an average house in San Francisco costs well over one million dollars. It’s a strange market, driven by status and hypocrisy. The residents are the real product, not their homes.

Home prices are rising because the area has increasing numbers of smart/skilled white/Asian people. Elites are inevitably drawn together (a single WMD would really mess things up). They also ‘subconsciously’ want to keep out less smart and skilled people, very much including non-Asian minorities, i.e. blacks, Hispanics, Muslims. If too many of the latter moved to San Francisco, the place would change in a way the current residents apparently wouldn’t like for some reason. Nothing personal you understand. The only way NAMs would be encouraged to move there would be if there was a way to control them into behaving in HIGHLY specific ways. That would require new types of surveillance, security, incentives.
Perfectly possible, in fact.

The city is a real-life communications hub, a networking center, a concentration of useful people that only exists because online communications still suck. Like most bad things, this is by choice not necessity. It has nothing to do with school zoning, as they’re hardly reproducing.
The great insight of the philosophical study field of Malism is that the world becomes evil by consensus. Most folks are at peace with other folks’ problems.

But progress can’t be stopped entirely. At least not yet. It could all be overthrown by mind-linked software.
The current computer hell will last at least another decade, but sometime in the twenty-twenties a change may come. Programmers may finally start removing embedded malware instead of adding bloatcode. Right now it’s really, really, bad. But then, their software may actually start working for a change.
One can only dream.

There’s no reason why people couldn’t collaborate in cyberspace instead of in hipster lofts. Instead of zip codes they would care more about digital reality.
“Ease of duplication / challenge of creation” means that physical objects and locations will eventually become irrelevant, the hidden base level of society.
Tomorrow’s geniuses may choose to live in out of the way places, like well-hidden and easily defended gated communities.
Those seeking to escape into digital hyperreality may consider it a badge of honor to appear uncool in real life.

Death: the darkest background

1) Awareness is fungible

Those expecting to die before 2050 may have a chance at a partial afterlife.
Any system offering even a remote escape possibility should attract adherents. The better the promise, the more adherents.
For good or bad, existing religions may have to be discredited first, as they make much better promises than any rational afterlife proposal.

Then, it comes down to the deceased Mind Backup customer’s willingness to create and leave behind an enormous trove of data to reconstruct their history and personality.
They will have filled a digital box with life stories, random data, and personality and memory tests.
This will slowly be sorted by whatever reconstruction software becomes available. Sorting this data will take more computations than the lifetime total of the brain that generated it. Only a sophisticated AI could do it.

The key insight is that mind reconstruction does not have to be fast.
There is no rush whatsoever, as long as no data is lost. The subject has all the time there is. Once all the data has been assembled, it doesn’t matter if it takes virtual centuries to make all the connections.
It doesn’t have to be an efficient or intensive process either. The awareness resulting from slow and inefficiently processed data will be just as real, even if we can’t interact with it.

In fact the reconstructed mind will emerge as a side effect of the process.
Its awareness and perception of time will be imposed externally, like a character in a novel. Different scenes will be created separately, and later fleshed out with perception details.
Single thoughts will slowly be made deeper and more detailed, linking to all kinds of memories.
What matters is the number of connections, not the order in which they are created.

Awareness may even emerge from an encyclopedic sorting process.
In fact a postmortem mind reconstruction does not have to be an exact or even approximate copy of the original to be its legitimate continuation.
The reconstruction could represent the average mind state over a lifetime. It could be true to itself even if it was mostly made up of reconstructed false memories to fill the gaps.

The most important requirement is obvious, and it suggests new testing methods:
A reconstructed mind should accurately predict how the original mind would have reacted to any situation.
Early mind reconstructions will have limited or no free will, though they won’t notice.
In fact we may only need to simulate a few ideal moments of a mind reconstruction – or even a single moment – but with a full set of memories and future plans.

2) Monumental moments

Mind-backup clients of the future will be encouraged to create tableaux of key scenes of their lives.
These virtual descriptions will be highly detailed in a few ways. Only essential perceptions and feelings need to be described, both their best and their most common days, specific and generic scenes.
The tableaux don’t have to be meaningful. This is a literary skill. The outline could be done very fast, but hundreds of them will eventually have to be completed.

It would begin with a few specific recreations. Floating cork-like in the surf off the beach, pedaling at sunset along a busy highway, lounging vegetable-like for years before entertainment boxes. Rooms and streets, fields and towns.
Then one true moment. What would it be? Maybe when operating at peak efficiency and the most connections, times of maximum influence or expectation. Or start with this moment right now.

They might even choose to compose situations that never happened but represent deep personal truths, even dreamlike or event horizon type situations. The tableaux could combine many life elements in one setting.

3) Mind extensions

It will of course take years to build such monuments.
Digital resurrection (or call it software mind continuation) should start as early as possible during the lifetime of the subject, with a self-improving mind extension.
It must be easy to begin, a simple but powerful way to record and store life data, complexity emerging as connections multiply. Early versions will mostly appeal to programmer-type personalities.
Data acquisition should become part of daily life, assembling long lists of memory factoids with associated tags (what when where why who).

So what’s the most powerful way to extract and organize your knowledge?
The memory map will start as an outline to be filled in.
Like life itself, it can never be completed, but any interesting perception and perspective could be added. It may involve the smallest units of awareness, some type of self-referential descriptions that might as well combine into a mind.
Solve this problem and every human problem may become solvable.

4) Example of a further mind scanning method

Stimulate small brain regions with focused radio waves or magnetically induced currents, and measure the resulting electrical activity in all other brain regions.
Combine this data with the best available personality tests, and the test results of many other subjects, to find correlations between measured brain activity and personality traits.
Eventually this may be done rapidly and precisely, and perhaps also reversed to create vivid interactive perceptions in the subject.

Estimated cost to develop this technique: somewhere between a mission to Mars and building the Gibraltar Dam.

The approaching dream of immortality

A human mind is immense. We’re talking millions of gigabytes. Because of bandwidth bottlenecks, mental states change rather slowly across the brain, though these states respond rapidly to the events they have evolved to handle.
A human lifetime is far too short to fully scan a mind’s contents, but not its structure. We’ll probably never download perceptions directly like in the movie ‘Strange Days’.
Fortunately, we don’t have to measure the mind’s awareness. All we have to do is measure the mind’s awareness of its awareness. Basically, the virtual brain within the real brain, a.k.a. the “soul”.

This will be the ultimate personality test. Beliefs, preferences and emotional ranges are different for everyone, a small fraction of memory-like states that control all the rest. They can predict how someone would feel and think about any situation.
The most interesting software of the not-too-distant future will make it its mission to measure these personal attributes.
Hopefully sometime before the 2030s, your first personal software assistant will become your preliminary mind extension. It will require a conceptual leap greater than a spiritual awakening, but not obviously impossible. This hypothetical software would identify, define, and track all its owners’ interests and sensibilities.

It would start as an automatic content aggregator, looking up and sorting meaningful data. First, the owner will define their lifetime interests, and all their random but meaningful events. It may seem narcissistic, but that may be the only meaning there is.
The software will also search the Net for relevant content the user would otherwise have missed. The harder part will be summarizing all this data.
Different users’ mind extensions will form online interest groups, and individual users may become experts in countless new sub-fields, like initial neurons in a World Mind.

Minds are vast, but that doesn’t have to make them complex, even if the strongest emotions are totally overwhelming.
Awareness is the deepest insight of ignorance, the vertigo of the lost past. The 1990s now seem as quaintly archaic as the 1970s did in the 1990s. The early 1980s had something called Teletext, which seemed almost as interesting as the Net does now. You could enter any of 999 page numbers on a suitable TV set, leading to short news stories displayed in phosphorescent text against a black background like an infinite resource.

Once someone starts recording the important elements of their life, an overview of the top level may emerge. The clutter can be sorted into hierarchies or added to a giant warehouse.
Human existence will have to yield to infinitely patient and methodical software. At first the subject’s life will be forced into a series of predetermined boxes like haikus. The ultimate goal of this project is very simple: to capture all the elements of nostalgia in a flowchart.

The process will change you. The final step of a mind backup attempt may be to manipulate the subject into becoming their ideal self.
It could happen in a VR environment, where an extremely intelligent AI therapist would hone in on their core personality. The only way to fully understand a human mind would be through an exponentially larger and smarter (though highly specialized) artificial mind. While incredibly complex, its creation would not necessarily be incredibly difficult.

This process will continue to its logical conclusion. Several levels above the smartest search engine, the creation of the World Mind will require prolonged exposure to the words and deeds of millions of distinct individuals, with or without their conscious assistance.