This is mostly still true – perhaps the part about how we reprogram computers has changed, because they are often programmed to “train” themselves through repetition.
Listening to Hamming in 1995 reminds me of something Kevin Kelly says: it feels like we’ve gone through the information revolution, but we need to be aware that computers and the internet are going to be a thing. We’re just getting going. We need to think, and think carefully about this.
[Computers] are going to dominate science and engineering. And there are reasons for this.
Economics
Computers are far far cheaper than human beings – far cheaper, and they’re getting cheaper by the year. Humans are getting more expensive by the year.
Speed
[Computers are] far far faster. Your nervous system, if you drop something on your toe, signals up to your head at about 100 meters per second. Light is 300,000 kilometers per second. You aren’t in a league. You can’t even touch electronic speeds. There’s no way you’ll come near. So speed is overwhelmingly on the side of the machine.
Accuracy
Namely the number of digits [unclear] arithmetic… they can be quite precise. They can do double precision if necessary. You would have trouble doing double precision arithmetic probably, if you tried doing it. You could work it out, but you’d have trouble.
Reliability
They’re far, far ahead of you. God or nature, however you want to do it, didn’t make you to be a reliable thing. You’ve been walking for years, and still every now and then you trip and stumble. That’s why man ended up on the top of the heap: he has the flexibility built in. But don’t ever try to get humans to do something reliable. Take for example bowling: all you have to do throw that ball down the alley exactly the same way every time and have a perfect game. Perfect games are rare, even among the most skilled experts. Precision flying and other things are very hard to do. We recognise that being very precise – drill teams and so on – are something remarkable. The human animal wasn’t really designed to do that – he was designed for something else.
Rapidity of Control
Because the machine’s got rapid control, we are now building aeroplanes which are basically unstable, and we have a computer, every millisecond, is correcting the instability. So we get better performance out of it. But the pilot couldn’t do it. If that computer goes out, the pilot’s through. The pilot is left with the large scale, the broad planning, but the millisecond to millisecond is left under a computer because a human just can’t act that fast.
Freedom from Boredom
This sounds trivial. You cannot put a human being ona job to look for something and then when it happens, respond promptly. You can put a computer on the job. You can put the computer on the job to watch for the rare event: if such-and-such happens in the atomic pile, do this. Well it hasn’t happened for four years, now the dial goes over – well the human being isn’t going to do very well. He hasn’t been looking at the thing for the last two-and-a-half years, even. You can’t get humans to be freed from boredom. Machines don’t know what the word is.
Bandwidth In and Out
In any rapidly changing situation the person in charge can only get so much information in and out, and there’s a general belief that really you can process only about fifty bits per second, maybe sixty – but you can’t process ten thousand bits per second. A machine’s got enormously more bandwidth, not only visual, auditory – put all your inputs together, they won’t match a modern computer for bandwidth. For central control the human simply cannot, in a complicated situation, compete with a machine if it is merely bandwidth in and bandwidth out. If it’s making judgements that’s another story. Thus we no longer have a crew aiming a gun at an aeroplane, we have it self-contained. The human is too slow – he just isn’t much good. We need much more rapid things than humans can cope with. The bandwidth in and out, which is really speed of getting information, is fundamental. Computers have got it all over you.
Ease of Retraining
Training to a great extent is that you learn to do something and now I change the equipment. You’ve got to unlearn the old habits and learn some new ones. And you’ve got to repeat them many many times to learn them. With the computer, I change the program and it’s done. No elaborate training, no endless hours of constant practice – big, you just put in a new program and the machine behaves in a new way. Very easy.
Hostile Environments
Outer space, underwater, high radiation fields, warfare, manufacturing situations are unhealthy and so on. I can put machines in those situations where humans are very very difficult. In space I have to keep a human being in an atmosphere somewhat like he’s used to: oxygen and so on has to be employed, high radiation will kill them and so on – how we’re going to get people to mars and back in the radiation field that’s coming from the sun I don’t know. Will we irradiate them thoroughly or maybe decide not to send human beings that far. It’s a problem.
Personal Problems
[This is] one I’m much sensitive to. There are all kinds of troubles with people. With machines there are no pensions. There are no personal squabbles – two machines don’t get squabbly with each other, but I’ve had two girls squabble together and wouldn’t even share the same room together. Unions? No. Personal leave? No. Egos? No. Death of relatives – your mother died. Machines don’t have that. Recreation: if I turn a machine off, that’s the end of it. I have a human being, I have to provide reasonable recreation.
Machines have got it all over humans.
Now all of your have probably already been saying “Oh yeah, but what about the advantages that humans have? I won’t have to list those – you’re trying to do it already. But I gave you a bunch of details that you’re going to find very hard to get around, that the machine has great advantages in many places, and because it’s economically sound you are going to see more and more machines running organisations.
Richard Hamming – Introduction to the Art of Doing Science and Engineering: Learning to Learn