Will Performance Hit Diminishing Returns Eventually?

[Joshua the Writer]

I think at a certain point, advancing technology will only result in diminishing returns. As in, tiny increases in performance not worth the effort being put into them. The best innovation in technology right now is the increasing availability of high performance devices, especially in relation to gaming hardware and media production hardware. We may even be in the beginnings of this stage of tech right now since some companies are more so focusing on applying the very fast tech to smaller form factors that are more affordable rather than big bricks that make games run 2% faster than the previous big brick that came out, which is good because that means quality hardware is both smaller and more accessible to various types of budgets.

 

[Outdated]

I think you're right there. The business world runs on supply and demand. When supply exceeds demand prices crash. Demand doesn't necessarily just mean quantity of products, people will always try to compromise between what they need and what they can afford and when cheaply available products are ample to their needs we cross a tipping point where there's not as much need to keep investing in the development of newer technology and also not as much spare cash to keep doing so.

In my view we crossed that tipping point about a decade ago. Most people no longer need a computer, the phone does all they need and more. The result of this is that phones and tablets are dirt cheap, iphones aren't at all popular in the eastern hemisphere, it's all android. And the price of real computers has gone up because there's less of a demand to produce them, reduced manufacturing and supply.

I'm a hardcore fan of the big brick by the way. But it's not just the performance hardware that makes this decision, we can change individual components ourselves and so they're cheap to maintain, and there's enough room inside that case for decent airflow and cooling which means your hardware lasts longer.

 

[Joshua the Writer]

I think you're right there. The business world runs on supply and demand. When supply exceeds demand prices crash. Demand doesn't necessarily just mean quantity of products, people will always try to compromise between what they need and what they can afford and when cheaply available products are ample to their needs we cross a tipping point where there's not as much need to keep investing in the development of newer technology and also not as much spare cash to keep doing so.

In my view we crossed that tipping point about a decade ago. Most people no longer need a computer, the phone does all they need and more. The result of this is that phones and tablets are dirt cheap, iphones aren't at all popular in the eastern hemisphere, it's all android. And the price of real computers has gone up because there's less of a demand to produce them, reduced manufacturing and supply.

I'm a hardcore fan of the big brick by the way. But it's not just the performance hardware that makes this decision, we can change individual components ourselves and so they're cheap to maintain, and there's enough room inside that case for decent airflow and cooling which means your hardware lasts longer.

I don't think there is any demand decrease in laptops and desktops. A lot of people still use PCs so long as they can access them simply because a lot of things are easier on a PC. I wouldn't attempt to do taxes while having to suffer through mobile UI. In fact, the only convenience that phones have is being able to fit in your pocket. Computers just do a lot more. The real reason prices for PCs have gone up is because tech companies are getting more and more greedy.

 

[Neonatal RRT]

I think at a certain point, advancing technology will only result in diminishing returns. As in, tiny increases in performance not worth the effort being put into them. The best innovation in technology right now is the increasing availability of high performance devices, especially in relation to gaming hardware and media production hardware. We may even be in the beginnings of this stage of tech right now since some companies are more so focusing on applying the very fast tech to smaller form factors that are more affordable rather than big bricks that make games run 2% faster than the previous big brick that came out, which is good because that means quality hardware is both smaller and more accessible to various types of budgets.

It depends upon the topic. A.I. and machine learning, currently is moving at an exponential pace, in some cases, (Tesla's full-self driving, Optimus humanoid robot), an order of magnitude every few months.

 

[Jeff T]

Am I seeing this phenomenon in pollution reduction in vehicles?
The original smog controls on vehicles were PCV valves and later some vacuum controlled valves. Apart from original costs incurred with perhaps changing the cylinder head castings, the system itself was inexpensive.
Later on came the EGR systems with changed engine manifold castings and hard tubing etc. A bit more expensive.
Later yet were catalyst systems- the converters had expensive elements involved. They tended to fail after awhile due to leaded gas and varying engine fuel ratios. And silicone sealants! Pricey!
Then switching to computer controls with all the sensors including fuel injection, now quite pricey.
Since then, more finely tweaking the engine management system, including fore AND aft Hego sensors, etc. etc.

But I haven't found anything online that shows the effectiveness of each system. In other words, a pie chart showing
the contribution in a percentage of pollution reduced by each system over time. Perhaps an engine that has been through all these developments over the years would give insight to this? The Chevy small-block is one of the longer produced engine families, but it has increased in volume as years went by, Perhaps that could be compensated for.

Any ideas- a pie chart's location of this data on the web / cost and effectiveness comparisons / me talking out of my ass with ridiculous unsubstantiated opinions, etc?

 

[Outdated]

Any ideas- a pie chart's location of this data on the web / cost and effectiveness comparisons / me talking out of my ass with ridiculous unsubstantiated opinions, etc?

You made me curious. I didn't find anything as direct as what you're looking for but I did find this chart that shows how over the years we've been driving a lot more and burning a lot more fuel but the pollution rates haven't risen as highly. Inference that there has been some beneficial effect from all this technology.

 

[Jeff T]

Yes, driving more; partly from increased commuting time across larger cities, etc.
Also a multiple increase in the number of vehicles, partly from more cars per household. Houses a long time ago often had one car garages, now three and four car garages are pretty common. Although you can only drive so many cars at once!
Also increased population. Here especially- California in 1960 had about 16 million, now we're about 39 million.

The visible portion of pollution reduction especially considering greater exhaust output in this time frame is remarkable!
In 1975 we occasionally couldn't see the mountains we lived right next to. Nowadays they're clear as a bell.

PS- thank you for the visual data!

 

[Jumpinbare]

If the automotive industry had focused on fuel efficiency, reduced emissions would have occurred as well. A maximally efficient engine would produce CO2, water, and nitrogen oxides. The catalytic converter was created to be used with hydrogen engines, which produced water and also nitrogen oxides. The converter was to break the nitrogen oxides back down into nitrogen and oxygen. An efficient hydrocarbon engine as well as hydrogen engines burn hot enough to ignite some of the nitrogen in the air/fuel mix, thus producing the nitrogen oxides.
Instead of focusing on burning the fuel efficiently, which would have reduced petroleum sales, governments dictated that the research focus would be on cleaning up the exhaust. Clean exhaust is better than dirty exhaust, but burning more efficiently means better economy and clean, reduced exhaust.
My brother had a 4 cylinder Datsun with manual transmission in the 1970s with a carburetor system and no computer that routinely got 40+ miles per gallon. I owned a 4 cylinder Yugo, also with manual transmission, carburetor, and no computer, that also routinely got over 40 mpg.
If research had been placed on maximum efficiency, we would have doubly benefited with clean air and reduced driving expenses.

 

[FilterFreq]

Honestly, I think a lot of what's being sold right now is empty lies., indicating that we've been in 'diminishing returns' territory for quite some time now.

With "full self-driving" cars being promised in 2016, AI essentially just being plagiarized material from books and the internet at large, and enthusiastic salesman and CEOs pushing whatever the latest trend is, it's not hard to consider the idea that diminishing returns are all around us and we're too smitten to see the reality.

Every new update is kind of like, "Trust me, bro. This time it's going to be great!", and yet still nobody's seriously playing games in VR, because most of these things are still just a very expensive gimmick... that you learn about $3,000 later. Chat GPT still can't spit out two separate versions of a lot of common algorithms, etc.

I'm not saying that these things can't be refined and actually get working like we want them to (let's be honest, we're really far behind where we'd actually like to be, despite what we're being sold right now), but by the time we get there, it's going to feel like, "Wow, yeah. We finally got there, didn't we?". The masses will have already moved on to some other thing and only us nerds will realize the miracle before us .

Meanwhile, most people don't realize the power they already wield. That thing in your pocket? You can make entire programs, apps, websites, tools, and so much more with it. You can learn so many interesting things, down the architecture and OS level, giving you crazy insight into the amazing spell-casting devices we carry around with us all the time, yet about 1% of the population even uses to that degree.

Honestly, improving on what we've already got is pretty difficult. It just goes to show you how far we've already come.

 

[Neonatal RRT]

With "full self-driving" cars being promised in 2016, AI essentially just being plagiarized material from books and the internet at large, and enthusiastic salesman and CEOs pushing whatever the latest trend is, it's not hard to consider the idea that diminishing returns are all around us and we're too smitten to see the reality.

Specifically, within the context of "self-driving/autonomous" vehicles, Tesla, the current leader in this technology, learned some hard lessons. Programmers and compute were the limiting factors. In this case, Tesla's team simply had to let go of the reigns and just let the machine learning happen. With humans in charge, no matter which software version that was released prior to 2024, there was a step forward, and then a plateau. Once there was enough compute for actual machine learning, (Dojo machine learning center) which didn't happen until 2024, and the machine learning was able to have visual learning from millions of human drivers, and millions of miles per day in data points, (weeding out bad events) only then was there a huge leap forward. It's an "observe and repeat" type of learning. Versions 12.3 to 12.4 to 12.5, only months apart from each other, are demonstrating leaps forward in capabilities. Now, it's "chasing the 9's" in terms of safety and driver interventions, 99.9%, then 99.99%, then 99.999% and so on. It's already statistically safer than humans overall, but it will need a few more orders of magnitude before regulators and the general public will embrace it. Almost there. Unlike previous versions, which was a bit scary, jerky, and hesitant, it now drives very human-like and smooth. It's a whole different thing now, and for the better.

The same types of generalized A.I. learning is being developed for the Optimus robot. Show it a video. Demonstrate how to perform and operation correctly, and its visual learning system (in communication with the centralized Dojo learning center) will observe and repeat. The A.I. essentially writes its own code. Anything that involves repeatable, task-oriented behavior, the robot will be able to learn and perform within the work environment.

 

[Joshua the Writer]

Specifically, within the context of "self-driving/autonomous" vehicles, Tesla, the current leader in this technology, learned some hard lessons. Programmers and compute were the limiting factors. In this case, Tesla's team simply had to let go of the reigns and just let the machine learning happen. With humans in charge, no matter which software version that was released prior to 2024, there was a step forward, and then a plateau. Once there was enough compute for actual machine learning, (Dojo machine learning center) which didn't happen until 2024, and the machine learning was able to have visual learning from millions of human drivers, and millions of miles per day in data points, (weeding out bad events) only then was there a huge leap forward. It's an "observe and repeat" type of learning. Versions 12.3 to 12.4 to 12.5, only months apart from each other, are demonstrating leaps forward in capabilities. Now, it's "chasing the 9's" in terms of safety and driver interventions, 99.9%, then 99.99%, then 99.999% and so on. It's already statistically safer than humans overall, but it will need a few more orders of magnitude before regulators and the general public will embrace it. Almost there. Unlike previous versions, which was a bit scary, jerky, and hesitant, it now drives very human-like and smooth. It's a whole different thing now, and for the better.

The same types of generalized A.I. learning is being developed for the Optimus robot. Show it a video. Demonstrate how to perform and operation correctly, and its visual learning system (in communication with the centralized Dojo learning center) will observe and repeat. The A.I. essentially writes its own code. Anything that involves repeatable, task-oriented behavior, the robot will be able to learn and perform within the work environment.

Even though a car with self driving capabilities may be safer than humans when it becomes nearly perfected, many people would still opt out of it due to the cost.

Self driving will be really good for things like tractors and warehouse automation (for the farms and warehouses that can afford it).

 

[Neonatal RRT]

Even though a car with self driving capabilities may be safer than humans when it becomes nearly perfected, many people would still opt out of it due to the cost.

Self driving will be really good for things like tractors and warehouse automation (for the farms and warehouses that can afford it).

Scales of economy. The next generation of EVs will be significantly less expensive than ICE vehicles and safety regulators may actually require it on the next generation vehicles, much like air bags and seat belts now-a-days. So, time will tell.

Yes, definitely, large scale, corporate farming, will and already are using autonomous machines.

 

[Outdated]

Specifically, within the context of "self-driving/autonomous" vehicles, Tesla, the current leader in this technology, learned some hard lessons. Programmers and compute were the limiting factors.

We've been using autonomous vehicles in our mining sector for 20 years now. What I find interesting is that there's been a few incidents with trains but never any issues whatsoever with trucks. The trains have their own dedicated rail lines where as the trucks are operating on public roads, although restricted to a few limited routes.

The picture below is one of Rio Tinto's newest babies, hauls 425 ton at a time.

 

[Neonatal RRT]

We've been using autonomous vehicles in our mining sector for 20 years now. What I find interesting is that there's been a few incidents with trains but never any issues whatsoever with trucks. The trains have their own dedicated rail lines where as the trucks are operating on public roads, although restricted to a few limited routes.

The picture below is one of Rio Tinto's newest babies, hauls 425 ton at a time.

View attachment 132877

GPS and programmed routes? Either way, I love impressive machines like this.

 

[Outdated]

GPS and programmed routes?

Combination, uses GPS but also has AI cameras and radar so it can avoid some incidents if necessary. Let's be honest though, it weighs 425 ton, wether it's autonomous or not there's nothing that can be done if someone pulls out in front of it. There's no such thing as lucky survivors in that sort of situation.

Perhaps that's part of the reason the trucks never have any issues, people don't mind cutting you up if you're in a little Hyundai but those trucks are scary, especially when you realise no one's driving them.

 

[Joshua the Writer]

Scales of economy. The next generation of EVs will be significantly less expensive than ICE vehicles and safety regulators may actually require it on the next generation vehicles, much like air bags and seat belts now-a-days. So, time will tell.

Yes, definitely, large scale, corporate farming, will and already are using autonomous machines.

To be honest, I still feel like we need to reevalute the process on new licenses. That is probably a better way to do road safety. Because eventually, people will start using self driving as a crutch.

 

[Outdated]

To be honest, I still feel like we need to reevalute the process on new licenses. That is probably a better way to do road safety. Because eventually, people will start using self driving as a crutch.

That's one of my regular rants here. There's a lot of people on the roads that really shouldn't be and they need to devise better testing methods. I'm not whingeing about people speeding or talking on phones here, there's some people that are just completely unaware of anything that's going on around them and also struggle to operate the vehicle. They're accidents just waiting to happen.

The trouble is that everyone pretty much needs to be able to drive in this country, it's very restrictive if you can't. We're a very sprawled out place. So courts try to avoid taking people's licenses away where policy allows so as to not interfere with their employment prospects.

 

[Joshua the Writer]

That's one of my regular rants here. There's a lot of people on the roads that really shouldn't be and they need to devise better testing methods. I'm not whingeing about people speeding or talking on phones here, there's some people that are just completely unaware of anything that's going on around them and also struggle to operate the vehicle. They're accidents just waiting to happen.

The trouble is that everyone pretty much needs to be able to drive in this country, it's very restrictive if you can't. We're a very sprawled out place. So courts try to avoid taking people's licenses away where policy allows so as to not interfere with their employment prospects.

We need more walkable cities in this country. A lot of countries where people walk more in the cities (mostly European and Asian ones). Since there's more people walking around, people are effectively exercising while doing their daily tasks, so obesity and other health issues are far less common over there, since people wall around more, as many people are within walking distance of everything they need. And if it isn't, they walk to the metro to go to a different part of the city, such as what people have been doing in Tokyo for decades.

 

[Outdated]

We need more walkable cities in this country.

It's a nice idea but there's a lot of different things that need to happen for that to work. A lot of it's to do with cultural ideals, Aussies don't like living in each other's back pockets, we all like a bit of personal space, so our cities are very spread out and two story houses aren't common. No one here wants to live in a high rise.

My nearest corner shop is a kilometre away and that's pretty normal for average suburbia here. I'll walk or ride the bike for that but then I'm limited in what I can carry so I'll only buy small odds and ends that way and travel to a big shopping centre for my real shopping. Or these days I get it delivered.

A lot of people spend over 30 minutes commuting to and from work each day in a car, if they're using public transport then it's going to take over an hour each way. That's the way our cities are spread out. Because of this the large corporate supermarket chains have pretty much cornered the market and the little corner shops struggle to survive.

In order for us to start living a more European lifestyle we'd have to heavily regulate those large corporations and encourage growth of more localised economies. Personally that's something I'd like to see, more wealth remaining in local economies instead of it all being shipped overseas, but I can't see that happening in a hurry.

 

[jsilver256]

Specifically, within the context of "self-driving/autonomous" vehicles, Tesla, the current leader in this technology, learned some hard lessons. Programmers and compute were the limiting factors. In this case, Tesla's team simply had to let go of the reigns and just let the machine learning happen. With humans in charge, no matter which software version that was released prior to 2024, there was a step forward, and then a plateau. Once there was enough compute for actual machine learning, (Dojo machine learning center) which didn't happen until 2024, and the machine learning was able to have visual learning from millions of human drivers, and millions of miles per day in data points, (weeding out bad events) only then was there a huge leap forward. It's an "observe and repeat" type of learning. Versions 12.3 to 12.4 to 12.5, only months apart from each other, are demonstrating leaps forward in capabilities. Now, it's "chasing the 9's" in terms of safety and driver interventions, 99.9%, then 99.99%, then 99.999% and so on. It's already statistically safer than humans overall, but it will need a few more orders of magnitude before regulators and the general public will embrace it. Almost there. Unlike previous versions, which was a bit scary, jerky, and hesitant, it now drives very human-like and smooth. It's a whole different thing now, and for the better.

The same types of generalized A.I. learning is being developed for the Optimus robot. Show it a video. Demonstrate how to perform and operation correctly, and its visual learning system (in communication with the centralized Dojo learning center) will observe and repeat. The A.I. essentially writes its own code. Anything that involves repeatable, task-oriented behavior, the robot will be able to learn and perform within the work environment.

This is accurate, but the enormous caveat with this kind of machine learning is that it is extremely suspectible to adversial attacks.

Simplistic example -> throw a pair of green shoes over a power line and it is entirely possible for computer vision to interpret that as a green light.

I do worry about terrorists doing something like flashing a red umbrella (which is easy for computer vision to mistake for a stop sign) and causing a gigantic pileup.