More Frighteningly Ambitious

Continuing my discussion of Paul Graham’s frightening ambitious ideas:

The Next Steve Jobs

I don’t see how a company can set out to be the next Apple, or how an individual can set out to be the next Steve Jobs.  This isn’t the way the world works.  Apple didn’t set out to be the Apple of today.  Sure, perhaps, early on, Jobs saw the plausibility of turning computers into household appliances, but I’m guessing he wasn’t thinking of the devices we have today – because back in the early eighties, they weren’t thinkable… and Jobs was a realist – a special type of realist who knew just how far reality could be distorted in his favour at any particular point in time.  And Jobs didn’t set out to be Jobs.  Not the Jobs we knew at the end.  That Jobs was created by the successes and failures of the younger, brasher, less tidied up Jobs.

But more than anything, I dint think Jobs would have set out to be the next anything else – he would have set out to be the first Steve Jobs.

Now – there is absolutely space for people to try to bring better design to the tech industry.  and there is space fot people who want to move on the capabilities of existing technologies.  These are things we need to see.  What Jobs had was a combination of good design, a step forward in capabilities and a strong brand behind him.  The strong brand was important – the strong brand is what gave Jobs the clout to get entertainment industries and telecoms industries moving into step with him.

Getting a strong brand is hard.  But these days its easier.  Facebook might, potentially, have some of the clout we are talking about, and its still young.  But to become a strong brand quickly requires a low cost of entry for the users – and that pretty much precludes being involved in making innovative consumer electronics.

So the future of design is going to start in software.  It’ll be when one of the guys behind some particularly popular and well designed website says “screw this – I don’t want you making my site ugly” to advertisers and finds another way to make money – possibly by extending his brand into the physical world that we’ll see changes happening…

Though the other place I would look to is kickstarter and etsy.  There are more and more iphone cases and ipad covers that exude beauty.  What if one of these designers were to build a wrapper around something cheap and generic (say the Raspberry Pi) and turn it into something better.  I don’t know what that something better might be, but we are at a place where design first development of products is looking plausible.

Bring Back Moore’s Law

To be honest, i’m not hopeful that someone is going to come out and say “Look at my new compiler, it avoids all the problems with parallel processing”.  But my experience is stat you never have to solve all of the problems, just some of them.

That said, I don’t think Moore’s law is the problem that needs to be solved, when it comes to parallelism.  I think scalability is the problem.  You want a program that runs as well on 12 cores as it does on 1 core – thats Moore’s law being brought back [we all know Moore's law hasn't gone anywhere in hardware - I'm talking about getting software to take advantage of it] – but you also want a program that runs on a million cloud based servers as it does on one core.  That is a different problem.  And its a problem we’re not close to solving.  So it really is frighteningly ambitious.

Programming languages, as they have taken off in real world usage have gone from being wrappers around assembly language [C] to being more and more abstract [C++, Java], and usable [Python, Ruby] and less woried about the processors control flow and more worried about the user’s [Javescript].  Operating systems used to just cove over the complexities of the CPU, now they provide more and more abstraction – to the extent we even have hypervisors – operating systems for operating systems.  But operating systems still work like CPUs do.

There is another layer of abstraction to be jumped to.  Abstraction over the cloud.

We have various parts of this.  Hadoop is the sort of engine we need inside such an OS.  The web provides us with a user interface to it.  But we don’t have the full tools.  What should happen is this:  I right a program which handles a users request, prcreeses it andprovides a response.  A simple program.  one that doesn’t worry about what else is happening.  Perhaps I write more programs to handle background activities and the like.  And I set all these programs running on ‘my cloud’ – something which I access through a browser, develop on through a browser and which looks like one big computer to me.  The cloud takes my code, and does all the work.  It figures out what the complexities are, what the things my code requires.  Where my code needs to scale by being broken down into multiple jobs.  And it compiles the code, and runs it appropriately – probably recompiling sections of the code in response to runtim analysis of modules.  The suer doesn’t have to understand how file storage is spread across a billion disks – just like right now I don’t have to understand about my single disk’s sector sizes and rotation speeds.

And yes – if it turned out that my cloud was a single corred mobile phone, then, yeah – why shouldn’t it be able to target that too?

All of this is possible, its just a huge and frightening task.  If someone were to take it on, the world would look a very different place immediately.

Ongoing Diagnosis

The problem with healthcare monitoring is – unlike most of the other ideas – it requires hardare.  And hardware is hard to make, expensive to ship, breaks, and is generally quite big.  So the problem is making light cheap healthcare sensors.  Which is something I’m absolutely not qualified to talk about.

But two things I do know about hardware are – it is cheaper to make hardware which is dumb, and it is cheaper to make hardware which is produced en mass.

Dumb hardware simply needs to communicate with software which can do the real processing – and combine the information from lots of sensors to build a bigger picture.  It may be the market itself is not in making the sensors, but it being the best diagnosis engine combining the inputs from lots of sensors and looking them up against a database.

Getting the first sensors cheap enough is a bigger problem. Were I going into this area, I would be looking at the developing world.  Right now, parts of the word are crying out for a doctor in a box.  It doesn’t need to be small enough to fit into your mobile phone or training shoe – just into the back of a Toyota Hilux.  But if it can be made fairly cheap, the market is out there – and there are Bill and Melida Gates’s who will pay you to make your product – and to make it cheaper and smaller, and more efficient.

While we could revolutionise first world healthcare (and that is probably where the big bucks are), while we are developing this, we might accidentally make the world a far beter place.

I don’t know healthcare.  I don’t know everything that can be done.  But I know that is the sort of accident I would like my startup to have.

You must be logged in to post a comment.