Not quite. Trading horizontal speed for altitude gets you quicker out of the atmosphere, where there's less drag. There's an optimal curve, and it's not a matter of going purely horizontal until the Earth "falls away". Quoting Elon Musk:
When you drop...the rocket, you have the slight problem that you're not going the right direction. If you look at what Orbital Sciences did with Pegasus, they have a delta wing to do the turn maneuver but then you've got this big wing that's added a bunch of mass and you've able to mostly, but not entirely, convert your horizontal velocity into vertical velocity, or mostly vertical velocity, and the net is really not great.
Quoting Wikipedia:
After five seconds of free-fall, the first stage ignites and the vehicle pitches up. The 45-degree delta wing (of carbon composite construction and double-wedge airfoil) aids pitch-up and provides some lift
If modern computers can't even manage to do something that simple, consistently without constant human intervention, do you really think they're anywhere near ready to handle the 1000x more complicated task of driving a car through poorly-marked city and rural streets without a human there to keep them from potentially causing major carnage?
There's a big difference in approach to these problems. The first is a simple rule based system, the second is a neural net system, which is much better at capturing complicated patterns.
All through my professional career I've bought eval boards for various new products so I can play with them in my weekends. Several of my coworkers thought I was crazy to pay stuff from my own money, and to work in my own time.
A rocket's payload has to be accelerated to 17,700 mph in the right direction to achieve orbit. Allen's Stratolaunch will provide the first 3% of that speed
One of the problems is that the rocket still needs to gain a lot of altitude at the point of release. The Pegasus uses a small wing to turn the rocket, and convert the horizontal speed into vertical speed. The wing adds extra mass, so it's not free.
And of course jet fuel is much less expensive than rocket fuel, especially when you consider the special handling rocket fuels require
According to figures on Wikipedia, a pegasus launch costs $56 million, and it can take 997 pounds to LEO. A Falcon 9 launch costs $62 million, and it can take a entire Pegasus rocket (40000 pounds) to LEO, and then land the booster again, using only $200k worth of fuel.
For example, a scanner could be developed that could allow police to ask a victim to remember what the criminal looked like and produce a near photographic image of him.
That's assuming that the neural patterns are still present in the brain, which is a big assumption
What we need is a spring dynamo generator in the shoes to recharge our wearables. I walk 4 times a week to work out - if there was a company that made a shoe charger for wearables, I would definitely buy stock, and would suggest everyone here to the same. It is an untapped goldmine.
You need to provide a power source capable of supplying 6MW. You need a cable capable of carrying 6MW to the car, which is going to require industrial equipment to lift because a person won't be strong enough, and wouldn't be safe anyway.
As others have explained, this battery can be charged simply by replacing the liquid inside. But even if you wanted to charge it electrically, you don't need industrial equipment. You could make a cable that's not harder to lift as a traditional gas pump hose, rated for 20000 volts and 300 Amps to get your 6 MW. And if you could wait 2 minutes, and fill the tank half way, that could be reduced to 10kV and 150A. A well constructed cable would be much safer than the heavy gasoline spraying system we have now.
Most charging is done while the owner is doing something else, like shopping or sleeping or working.
If you launch from a plane you can use a much smaller rocket - hence the cost savings.
If you can reuse the rocket, it doesn't matter much if it's bigger. And the difference apparently isn't that great. According the Musk, you only save about 5% by launching from a plane[1]. Add the cost of the plane itself, the more complicated launch, plus the fact that you'll be limited to small rockets and payloads, and it's clear why they didn't choose this path.
What part of playing Go was really an advance in what we already knew about neural networks
The playing strength was suddenly far beyond what people were expecting. They were not the first to try out neural networks for Go, but they certainly created the biggest jump.
Have they demonstrated it getting better and better as it plays and self-adapting (meaning no code changes) . or do they need to make code changes each time?
It acquired most of its knowledge from self-play at rapid time controls, using a few weeks of training time. Once it is set up, no more code changes or human input is necessary to improve playing strength. They bootstrapped it by letting it watch human games first, but I've read that people are working on a new version that starts completely from scratch. Similar techniques can be used for driving cars. Google must have endless footage from current vehicles that can be used to train new networks in the lab. By just recording millions of hours of pedestrians walking on/near the roads, and using those as input, you can probably build a good model of a pedestrian, and their likely movements, without any human input.
What percentage of playing the game of Go is the same ruleset applied over and over and over, and what percentage is an actual unique idea?
Every game and every position is unique, once you get past the initial opening moves. The computer has learned to recognize patterns of patterns, and apply these patterns in new situations.
Nothing about Go involves understanding the world or people. Everything about driving involves understanding the world and people.
My point was not that Go is equal to driving. The point was that progress comes in leaps and bounds, not gradual.
Playing Go means you have to understand vague concepts like "influence" that are not defined by rules. Driving means understanding vague concepts like "pedestrians". Even though these are different concepts, they are comparably fuzzy in their definition, so it's likely that progress in autonomous driving will also happen in leaps and bounds.
Slashdot Mirror
In this launch video you can clearly see it curves upward. https://www.youtube.com/watch?...
That would be a silly waste of velocity.
Not quite. Trading horizontal speed for altitude gets you quicker out of the atmosphere, where there's less drag. There's an optimal curve, and it's not a matter of going purely horizontal until the Earth "falls away". Quoting Elon Musk:
When you drop...the rocket, you have the slight problem that you're not going the right direction. If you look at what Orbital Sciences did with Pegasus, they have a delta wing to do the turn maneuver but then you've got this big wing that's added a bunch of mass and you've able to mostly, but not entirely, convert your horizontal velocity into vertical velocity, or mostly vertical velocity, and the net is really not great.
Quoting Wikipedia:
After five seconds of free-fall, the first stage ignites and the vehicle pitches up. The 45-degree delta wing (of carbon composite construction and double-wedge airfoil) aids pitch-up and provides some lift
It helps a lot that you can launch from California instead of Florida.
If modern computers can't even manage to do something that simple, consistently without constant human intervention, do you really think they're anywhere near ready to handle the 1000x more complicated task of driving a car through poorly-marked city and rural streets without a human there to keep them from potentially causing major carnage?
There's a big difference in approach to these problems. The first is a simple rule based system, the second is a neural net system, which is much better at capturing complicated patterns.
All through my professional career I've bought eval boards for various new products so I can play with them in my weekends. Several of my coworkers thought I was crazy to pay stuff from my own money, and to work in my own time.
A rocket's payload has to be accelerated to 17,700 mph in the right direction to achieve orbit. Allen's Stratolaunch will provide the first 3% of that speed
One of the problems is that the rocket still needs to gain a lot of altitude at the point of release. The Pegasus uses a small wing to turn the rocket, and convert the horizontal speed into vertical speed. The wing adds extra mass, so it's not free.
And of course jet fuel is much less expensive than rocket fuel, especially when you consider the special handling rocket fuels require
According to figures on Wikipedia, a pegasus launch costs $56 million, and it can take 997 pounds to LEO. A Falcon 9 launch costs $62 million, and it can take a entire Pegasus rocket (40000 pounds) to LEO, and then land the booster again, using only $200k worth of fuel.
For example, a scanner could be developed that could allow police to ask a victim to remember what the criminal looked like and produce a near photographic image of him.
That's assuming that the neural patterns are still present in the brain, which is a big assumption
Another labor attorney even suggests tech firms are hiring younger workers because they ask for lower salaries and less time off
As an older person, you can also ask for lower salary and less time off.
What we need is a spring dynamo generator in the shoes to recharge our wearables. I walk 4 times a week to work out - if there was a company that made a shoe charger for wearables, I would definitely buy stock, and would suggest everyone here to the same. It is an untapped goldmine.
https://www.youtube.com/watch?...
Replacing goal with gas is good, but when the gas runs out, will you switch back to coal ?
You need to provide a power source capable of supplying 6MW. You need a cable capable of carrying 6MW to the car, which is going to require industrial equipment to lift because a person won't be strong enough, and wouldn't be safe anyway.
As others have explained, this battery can be charged simply by replacing the liquid inside. But even if you wanted to charge it electrically, you don't need industrial equipment. You could make a cable that's not harder to lift as a traditional gas pump hose, rated for 20000 volts and 300 Amps to get your 6 MW. And if you could wait 2 minutes, and fill the tank half way, that could be reduced to 10kV and 150A. A well constructed cable would be much safer than the heavy gasoline spraying system we have now.
Most charging is done while the owner is doing something else, like shopping or sleeping or working.
Except when you want to make a longer trip.
You could vent the LOX without too much trouble, I suppose.
If you launch from a plane you can use a much smaller rocket - hence the cost savings.
If you can reuse the rocket, it doesn't matter much if it's bigger. And the difference apparently isn't that great. According the Musk, you only save about 5% by launching from a plane[1]. Add the cost of the plane itself, the more complicated launch, plus the fact that you'll be limited to small rockets and payloads, and it's clear why they didn't choose this path.
[1] http://shitelonsays.com/transc...
Reducing the price of the launch comes with reuse of the rocket. Launching it from a plane isn't going to help much.
Another advantage is that you can launch in bad weather, and that you can use an engine configuration optimized for lower air density.
space is up, not sideways
Actually space is mostly sideways. https://what-if.xkcd.com/58/
But I agree that the extra complexity from a plane launch is probably not worth the trouble.
What part of playing Go was really an advance in what we already knew about neural networks
The playing strength was suddenly far beyond what people were expecting. They were not the first to try out neural networks for Go, but they certainly created the biggest jump.
Have they demonstrated it getting better and better as it plays and self-adapting (meaning no code changes) . or do they need to make code changes each time?
It acquired most of its knowledge from self-play at rapid time controls, using a few weeks of training time. Once it is set up, no more code changes or human input is necessary to improve playing strength. They bootstrapped it by letting it watch human games first, but I've read that people are working on a new version that starts completely from scratch. Similar techniques can be used for driving cars. Google must have endless footage from current vehicles that can be used to train new networks in the lab. By just recording millions of hours of pedestrians walking on/near the roads, and using those as input, you can probably build a good model of a pedestrian, and their likely movements, without any human input.
What percentage of playing the game of Go is the same ruleset applied over and over and over, and what percentage is an actual unique idea?
Every game and every position is unique, once you get past the initial opening moves. The computer has learned to recognize patterns of patterns, and apply these patterns in new situations.
Nothing about Go involves understanding the world or people. Everything about driving involves understanding the world and people.
My point was not that Go is equal to driving. The point was that progress comes in leaps and bounds, not gradual.
Playing Go means you have to understand vague concepts like "influence" that are not defined by rules. Driving means understanding vague concepts like "pedestrians". Even though these are different concepts, they are comparably fuzzy in their definition, so it's likely that progress in autonomous driving will also happen in leaps and bounds.
You shouldn't eat so much granola bars.
Do you also insist on eating from hand plowed fields, and goods carried to the store by hand ?
Many companies in the retail sector work on very slim margins and try to make it up in volume
Even if the margin is very slim, you can still tax a percentage without hurting them too much. Or you can make up a progressive tax structure.
That's a 5% reduction. For an outlay that starts at $800 billion a year and scales to $7 trillion.
That's the value of the "passenger economy", not the cost. We save money, and we save lives.
How about spending a fraction of that money on real driver education, training, and enforcement?
We've been doing that for decades. Why do you think that we could do a lot better ?
we can ONLY create the false illusion of them, a parlor trick, fakery
That's good enough. My brain also creates false illusions, parlor tricks, and fakery, and they allow me to drive a car.
Scientists who study the human brain agree with me
citation required.
Seriously, income tax is the best form of tax.
Except when nobody has income.
some large multinationals pay less taxes than me and my spouse together.
Taxing robots won't fix that, though.