2045 predictions

[wjfox]

I'm keen to update our 2045 predictions, which appeared on the timeline back in 2009, and were heavily influenced by Kurzweil's book. IMO they need tweaking to better reflect some of the recent trends/developments we've seen.

So, please give me your predictions for 2045 – particularly with regard to the following topics:

- AI, automation, and robotics, and how these will affect everyday life/society
- Brain-computer interfaces (and other implantable devices), and how these might shift from medical to consumer uses
– Supercomputing (likely somewhere between zettascale and yottascale?) and its potential applications by then
- General consumer tech, gadgets, VR, home computing, Internet, etc.

Any other predictions are welcome, but those are the main areas I want to focus on.

Basically, I want to describe a world and society that is clearly approaching a Singularity-type transition - but which isn't quite there yet (Ray Kurzweil tends to overestimate by a decade or so). I'll try to incorporate some of the best ideas that are posted here.

Any links/references you can find will be particularly helpful.

Thanks!

[TrueAnimationFan]

Here are two guesses of mine based on what exists now.
I'd say that by 2045 we should see at least one supercomputer pass 100 zettaflops. I also think that by the end of the 2040s, we should see the first humanoid robots that are effectively indistinguishable from biological humans (though maybe not quite as intelligent) - and not just in appearance, but in their movements as well. Just look at this footage from 25 years before then, for example:

[Nero]

There is a zero percent chance that any machine with 100 zettaflops of computing power would be less intelligent than a human being. Such computational ability is several orders of magnitude above anything the human brain can compete with.

As for a prediction, it would be likely that in 2045 The Singularity will have already occurred. Making specific predictions impossible.

[funkervogt]

By 2045, robot servants will be common in upper-income and middle-class households across the developing world. Some will be function-specific, like autonomous lawn mowers, while others will be multifunctional, like robot butlers. They will work more slowly than humans and will make mistakes more often, but nevertheless, they will save their human owners many hours of work each week.

A high-quality multifunction robot servant will, in 2045, cost about $10,000 in today's money. In other words, cheaper than a new car, but still a significant investment of money.

[weatheriscool]

America either will be an a.i ruled police state that uses robots to control the chuttle that is the human population or will rip itsself apart in massive civil wars.

[Tadasuke]

I agree about robots servants prediction by funkervogt. I think it is possible to predict 2045.

Kurzweil's Law says that each technological paradigm shift accelerates subsequent innovation logarithmically, or at an exponential rate of change. Think ripples in a pool around a heavy object.

About supercomputers, I think that in general-purpose double-precision the TOP1 machine will be 30-50 zettaflops (10x every 5 years). That will mean very accurate weather predictions, insights to how chemistry and biology work, human-level AI (which I predict for ~2040 in a supercomputer), exoplanets discovery and understanding, fusion working commercially, material science advances resulting in better materials, batteries, gene editing. Medicine will be not found by trial and error, but researched, understood and designed with supercomputers.

Vehicles will be mostly electric, lightweight and autonomous. You will be able to drive more than a thousand km on a single charge and the charging process will be very quick. Supersonic passenger airplanes will be back, this time quiet and efficient. Patrol boats will patrol for pirates, maintenance and sea wildlife inspection. Robots will clean rivers and oceans (also boats underwater). Robots will clean cities, warehouses and houses. All lights will be LED and they will be any color someone wants.

Cashiers, taxi and truck drivers will be gone completely. Automation replacing workers will be in high gear. Retail space per person will decrease in countries where it was high in 2021.

Telescopes will see exoplanets and we will know some things about them. No alien contact. Moon base and Mars base. Space station around the Moon. Space hotel around Earth. Cost to orbit in tens of dollars per kg per launch.

VR will have 256x (2^8) higher pixel count than in 2021. Most of that will go into resolution and some into refresh rate. So expect crisp 210° high-refresh image. No full dive yet. Very convincing 3D sound. Haptics, tactile feedback built in gloves. Vests giving more feedback. Some people will use neural connections and other BCI to control VR. Mind uploading will be still science-fiction.

Computer chips will be all 3D. 3D will be the standard everywhere. Consumer mechanical drives will be history. Consumers will use only SSDs. Full sized desktop PCs will be only for enthusiasts, professional users and very hardcore gamers. Streaming will not replace gaming on your device like it did with movies and music. Many people will still prefer gaming on devices of choice. Just that 3D architectures will mean more compact computers, but with elaborate cooling systems.

Virtual entertainment will be divided into 3 categories:
- games which are only games (Tetris for example)
- games which are also virtual worlds (Assassin's Creed for example)
- virtual worlds which are not really games, but virtual 3D worlds nonetheless (for non-gamers and gamers alike, like Second Life), used for many purposes

Health will be monitored via wearables. The most common personal computer will be AR head-mounted glasses or VR head-mounted headset like Oculus Quest. There will still be laptops and smartphones (which are going to be foldable into 3 or more parts). Although games will be rendered on-device, some things in virtual worlds may be done on servers (AI, physics).

By 2045, virtual youtubers will be super big. And there will be already actual AIs doing it. Virtual porn will be more popular than real porn. Also interactive.

In 2045 you will be able to ask AI to do things for you: find something, write something, draw something, fake a video, fake voice.

[andmar74]

Basically, I want to describe a world and society that is clearly approaching a Singularity-type transition - but which isn't quite there yet (Ray Kurzweil tends to overestimate by a decade or so). I'll try to incorporate some of the best ideas that are posted here.

Kurzweil based the estimate of the year 2045 for the Singularity on the following: In 2029 a computer passes the Turing test for the first time, and in the next 16 years the computer is maturing much like a child growing up. I think there's a good chance of success in the Turing test in 2029, but it's obvious now that it will not need 16 years for maturing, so I would say the Singularity can happen a short time after 2029.

[funkervogt]

Today, there are many AI experts who say that we still have no clue how cognition works, wonder algorithms like GPT-3 are not actually "intelligent" and instead use many bells and whistles to "fake it" for laypeople, and there's a very long road ahead to building a real AI.

By 2045, those kinds of people will be nearly nonexistent. I think an AGI is slightly unlikely to exist by then, but we will be close. Even skeptics will admit we've made major progress understanding, at an algorithmic level, how intelligent thinking works.

[Tadasuke]

Today's AI works like video games graphics, which only fake everything by "smoke and mirrors", it fakes intelligence. Raster graphics uses techniques like baked lighting to fake lighting and shadows, cube maps to fake reflections. Games use scripts to fake events and NPC AI. GPT-3 is much much more advanced, but it's still not generally intelligent like a human. It doesn't really understand stuff, it only follows the algorithm.

But I still firmly believe that by 2040 machines will achieve human capabilities, no matter how they do it. Faking or not faking, they will have all human-level capabilities in both mental and physical tasks. By 2040 human-level android will be 100% possible. It may cost $2000 000, but will be real. Human likeness in looks will be possible as well, not necessarily desirable however. By 2045 wealthy families will indeed probably own androids to do chores for them.

[wjfox]

Detroit: Beyond Human was IMO a very accurate portrayal of what to expect. I really recommend that game to everyone here.

Basically, the androids start out passive and compliant, but gradually become more aware of – and resistant to – their slave-like status.

So I believe these robot servants will initially be like highly intelligent Roombas, completely docile. But over time, as they develop into more advanced and capable models, more "emergent" behaviours and subjective experiences will trigger a kind of breakthrough point in their programming. This will eventually culminate in them being recognised as fully sentient individuals with rights and freedoms.

I agree with 2040-45 as the inflection point when we'll start to see very human-like robots in home settings. They will also be popular as sex workers, and it might even be possible to get legally married to one, in some jurisdictions at least. But probably a few more decades before they are truly self-aware and independently making a mark on society.

I wonder how they'll be constructed? Presumably some sort of 3D printing process, perhaps replicating the 206 bones in the human body, with an outer polymer layer of skin. Tweaks available for different build types, gender, skin tones, etc. I imagine the face and mouth movements will be especially challenging to replicate. Maybe this could be solved with a kind of GAN-like process, where the movements are mapped in 3D and then optimised to match the most human-like subtleties. New materials might also be needed for more realistic skin, as well as the "twinkle" effect of eyes.

[funkervogt]

Detroit: Beyond Human was IMO a very accurate portrayal of what to expect. I really recommend that game to everyone here.

Basically, the androids start out passive and compliant, but gradually become more aware of – and resistant to – their slave-like status.

So I believe these robot servants will initially be like highly intelligent Roombas, completely docile. But over time, as they develop into more advanced and capable models, more "emergent" behaviours and subjective experiences will trigger a kind of breakthrough point in their programming. This will eventually culminate in them being recognised as fully sentient individuals with rights and freedoms.

As an important caveat, I'd like to add that, even after AGI has been created, it won't be infused into every robot. For example, a Roomba whose purpose it is to vacuum floors would not benefit from having human-level intelligence. The benefits would probably top out at cockroach-level.

It would also be more trouble than it was worth to make more intelligent robots than was absolutely necessary, as such machines would be smart enough to behave unpredictably, and to demand more rights.

Even if Skynet ran the world, it wouldn't make every machine intelligent; it would see the value in keeping most of them dumb and incapable of performing more than narrow ranges of tasks.

I agree with 2040-45 as the inflection point when we'll start to see very human-like robots in home settings. They will also be popular as sex workers, and it might even be possible to get legally married to one, in some jurisdictions at least. But probably a few more decades before they are truly self-aware and independently making a mark on society.

I wonder how they'll be constructed? Presumably some sort of 3D printing process, perhaps replicating the 206 bones in the human body, with an outer polymer layer of skin. Tweaks available for different build types, gender, skin tones, etc. I imagine the face and mouth movements will be especially challenging to replicate. Maybe this could be solved with a kind of GAN-like process, where the movements are mapped in 3D and then optimised to match the most human-like subtleties. New materials might also be needed for more realistic skin, as well as the "twinkle" effect of eyes.

I doubt sex androids will replicate every aspect of internal human anatomy. It's possible that incorporating only 103 bones into their skeletons will make them outwardly indistinguishable from humans. Additionally, since sex with them will mostly happen in dark or low-light conditions, deficiencies in the humanness of their faces and expressions will be masked.

[Nero]

A lot of the practical applications of machines will not require AI to function much better than it does now in 2021. Generally most of a specific task can be replicated and you really do not need to achieve anything beyond that. It is more likely that machines remain stupid but grouped via some form of IoT that is within the boundaries of their intelligence, this will probably lead to more machines but smaller doing more and more specified tasks.

[funkervogt]

A lot of the practical applications of machines will not require AI to function much better than it does now in 2021. Generally most of a specific task can be replicated and you really do not need to achieve anything beyond that. It is more likely that machines remain stupid but grouped via some form of IoT that is within the boundaries of their intelligence, this will probably lead to more machines but smaller doing more and more specified tasks.

Right. Nature itself provides a picture of what the optimal setup looks like. There are many thousands of animal species on Earth, but only one of them has intelligence. By far, the most common animals are small, simple, and only capable of doing one or two functions that support the broader ecosystem. This is an insect-dominated planet, and each species has a little niche.

This also touches on why I don't think the discovery of AGI will be necessary to cause widespread technological unemployment. By building thousands of types of robots that only have cockroach-level intelligence and training them to do one task apiece (e.g. - mow lawn, drive car, wash windows, sell insurance), we could automate most human jobs.

[wjfox]

I need a new title for this 2045 prediction. Suggestions welcome.

Currently thinking of "An inflection point for human-AI interactions" or something similar.

[raklian]

I need a new title for this 2045 prediction. Suggestions welcome.

Currently thinking of "An inflection point for human-AI interactions" or something similar.

New era for human-AI relationship?

[TrueAnimationFan]

2045 - 2050
Machine intelligence is rivalling that of biological humans

[wjfox]

Thanks for your comments and suggestions so far, guys. I now have a better idea of what to write about – mainly the spread of humanoid robotics in home and workplace settings, the progress of generalised AI, integration of BCIs and other implants, ongoing improvements in computing, etc.

I just need some links/references to make this prediction more credible.

Anything you can find that relates to 2045, or includes mentions of "25 years from now", "the 2040s", "mid-century", etc. would be helpful.

Ideally what I'm looking for is graphs/projections and long-term reports from governments, market analysts, and research institutes, recent books by experts in their field, etc. Alternatively, non-sensationalist articles from reputable news sites, journals, science magazines, and whatnot.

In fact, it doesn't even have to mention 2045 specifically – even a partial trend or projection to, say, 2030, would be very helpful, as I can extrapolate further into the future. So, any useful data relating to AI and robotics would be helpful, such as trends in language/image processing algorithms, and how the speed and abilities of robots are changing over time, plus the general trends in computing technology, device sizes, market growth, costs, etc.

I may briefly mention The Singularity is Near, but since that book is now 16 years old, and increasingly outdated, it's the previously-mentioned sources that are really preferred. Thanks!

[raklian]

Thanks for your comments and suggestions so far, guys. I now have a better idea of what to write about – mainly the spread of humanoid robotics in home and workplace settings, the progress of generalised AI, integration of BCIs and other implants, ongoing improvements in computing, etc.

I just need some links/references to make this prediction more credible.

Anything you can find that relates to 2045, or includes mentions of "25 years from now", "the 2040s", "mid-century", etc. would be helpful.

Ideally what I'm looking for is graphs/projections and long-term reports from governments, market analysts, and research institutes, recent books by experts in their field, etc. Alternatively, non-sensationalist articles from reputable news sites, journals, science magazines, and whatnot.

In fact, it doesn't even have to mention 2045 specifically – even a partial trend or projection to, say, 2030, would be very helpful, as I can extrapolate further into the future. So, any useful data relating to AI and robotics would be helpful, such as trends in language/image processing algorithms, and how the speed and abilities of robots are changing over time, plus the general trends in computing technology, device sizes, market growth, costs, etc.

I may briefly mention The Singularity is Near, but since that book is now 16 years old, and increasingly outdated, it's the previously-mentioned sources that are really preferred. Thanks!

The best I could find is this: https://ai100.stanford.edu/sites/g/file ... 831fnl.pdf

[Doozer]

I predict that both graphene and nuclear fusion will be fully commercialized at this point. There are only very few fusion reactors and graphene is still mostly reserved for large-scale industrial use. Fusion is however slowly reducing the need to burning coal for electricity and graphene is applied to everything including construction, electronics and filtration.

The world's population however has reached crisis levels and plans are underway to establish space colonies. Efforts are made in undoing as much environmental destruction as possible thanks to new technologies and sciences. As for health, many common deadly diseases such as cancer and AIDS have been reduced greatly with ever advancing medicine.

[funkervogt]

By 2045, hardware and its costs will be no barrier to building an AGI. A computer with as much computation and memory storage as a human brain will cost only about $1 million. If AGI doesn't exist by 2045, it will be because we lack the software.