40 Comments

well, if others are building this and sharing it openly, then I might as well share my AI market place solution to poverty that makes a lot of profit too. My solution is public governed and public domain so that's probably good? the solution I propose is open source, open data and extensible. It is also highly profitable and incentive positive in a free market. It produces a profit in any economy by exploiting any form of inequality and ultimately reduces inequality as a side effect of operations and wealth extraction. https://sites.google.com/vi...

Expand full comment

@Lump1,

you wrote "Your argument for why this is unlikely is that we live in a world where a single actor simply cannot develop a significant AI lead."

I disagree. I believe a single form of AI developed by a single or multiple actors can develop a significant AI Lead. Here is a candidate for proof by contradiction at https://sites.google.com/vi...

p.s. If you object to my comment for any reason(such as TMI), just ask and I'll delete it from the record at your request.

Expand full comment

I think the article and the update miss the key point that almost nothing Amazon, Google, Facebook, or Microsoft do is a secret. You could build a new infrastructure to rival any one of those easily with enough resources. All regularly blog or publish papers about their advances in Machine Learning. Much of the backend is either open source or has well documented APIs. They use open server center designs. And as was pointed out they sell sufficiently powerful and affordable infrastructure to all comers who can grow to be their next competitors.

Dropbox was using Amazon for storage until they grew to be large enough to move to their own infrastructure. The only problem was cost, time and people. NOT software. They even blogged about it. I know several start ups who do scaling magic using Google kubernetes engine to create products that directly compete with Google.

The real problem is people and data. Look at how Apple is struggling with machine learning. They could literally read Google's and Microsoft's ML blogs and copy everything But there aren't enough engineers to go around - the best people want publish everything they do. And not enough data.

Even Amazon's warehousing and fulfilment infrastructure probably wouldn't be that hard to replicate. JIT and robotics are pretty well known. And it's not like this hasn't been done.

To build and scale a new Facebook, Instagram, WhatsApp or Twitter would be relatively trivial. But the problem is mindshare. There is no shortage of technologically viable competitors who almost nobody uses.

And even if you had the people and money you need to do organizational learning that takes years.

I'm not opposed to technological regulation but in this case it should be regarding behavior in the marketplace and tax and labor law compliance. I don't see how compulsory licensing would solve any of the gap. Or even what would be the target of the licensing.

Expand full comment

Is Microsoft able to change Windows without breaking all that legacy software people have written for old versions?

It's already happened at least once, and some people's Win 7 software just doesn't work on Windows 10. Microsoft tends to leave support in for the most-recent past version of Windows, but even then not always.

Apple gets away with this almost every OS upgrade, but their developer community is smaller, and they support small developers better.

Expand full comment

"Table stakes" in any natural monopoly are insuperable barriers to entry. The idea is simple, but the investment required is too large for competition to be viable.

Expand full comment

Terms used by experts to communicate complex ideas "should" have simple definitions?

Expand full comment

(Statement of bias: I work for a company in this field)The technology that might disrupt this is Salesforce and its ilk. Good systems of this sort make it much, much easier to create good, custom applications. It's still hard, but it's doable by a bunch of average engineers rather than a bunch of great engineers. The big issue here is that stuff like Salesforce doesn't seem to have hit medium or small companies that much, so they're with off-the-shelf software that doesn't fit their needs, or hiring consultants to build something custom, and then being screwed when they need to update it or support it. Still, these things barely existed fifteen years ago, so the trend is promising at least.

Expand full comment

I agree, I think we'll see waves of open source as it becomes economically nonviable to pay the costs associated with the tech giants. There was one a few years back though sometimes the in house costs to implement these is greater than buying a product that was made less "democratically". With that said though these tech giants also have control over the internet and search results. It's extremely easy for them to make the competition invisible and we're already starting to see some of that.

Expand full comment

It happened to Microsoft. Part of the issue is the world is getting more complex, and moving faster while government is getting slower and their hiring practices and work environments aren't attracting smart young people with new ideas and solutions to these problems. Though young people want to change the world, they also want the opportunity to do so without the bureaucracy. The tech industry realizes how long regulations take and uses it to their advantage. Startups don't need a long life, they can work outside of regulations, and sell for a huge profit long before the government has time to react. Sometimes the length of time it takes results in them giving in because the industry has become so well established.

Expand full comment

Agreed. These cloud services could increase rates dramatically and still retain customers once they're locked in. I have no doubt they will change rates and terms to the point of it being far worse than having in house solutions.

Microsoft will need to break legacy at some point. They're held back by this and the security of their OS is held back by this. Virtualization is definitely possible and we're almost at that point where the OS shouldn't matter with so much turning to cloud services. I'm a much bigger fan of MacOS or even ChromeOS from a user standpoint. Windows has too much legacy holding back the UI, and requires far too much support from IT departments.

Expand full comment

Oligopoly would be the correct term. There are few instances where there are no competitors due to the antitrust laws. That said the few in power tend to work together on price fixing, or "taking out" competitors so they get some of the benefits of a monopoly. This happens in tech, cable/internet/cell service providers, media, grain or dairy industries, junk food, pharma etc. The anti-trust laws tend to work but business always finds ways to stay ahead of government regulation.

Expand full comment

The other issue with internally built systems is being able to retain employees that know how to use or manage the system. We're losing our database admin who built a Student Information System from the ground up. It handles far more than typical competing products do but he is the only person who knows how it works. He is currently training someone to look after it until we find an off the shelf replacement. People tend to not like the system we have but no doubt they'll probably like the 3 or 4 systems it takes to replace it much less. Eventually the budgets or efficiency will suffer. I agree in house can be far better but finding an organization willing to take that risk is becoming much harder.

I'm finding more and more institutions want replaceable people using standardized software and the standardized work IT workers are forced to do leads them to move on from an organization far sooner. There's a certain intelligence barrier for entry to manage IT at the highest level but much of the work is often too low a level to remain engaging. Contracting out has become the norm for a lot of bigger projects that require time devoted away from the day to day operations.

Expand full comment

On a related not why hasn't the Tron meme "I fight for the users" not been adapted as a rallying cry for NORML? these things seem made to fit together.

Expand full comment

The people, the code the regulations... They are one. Separation is not possible.

The more interesting avenue then standardizing the code is to add a phase to every project where the engineers on the development team get to talk with the operators. Not the process people not the accountants not the lawyers, the operators.

unfortunately this phase is almost universally "cost engineered" out. But if you want things to work sit me down with the pore bastard who actually has to live with my terrible decisions for the next five to twenty years; He'll care he'll be invested. I'm in development if you couldn't tell.

Expand full comment

Integrating two large IT systems is like trying to fit two skyscrapers one into another. With their miles of electrical wires, plumbing, elevators, ducting, stairwells, it simply cannot be done. I don't understand why people keep trying. In particular, governments keep sending truckloads of money down this hole after being misled by the likes of Oracle, SAP or IBM. See for example: http://www.mcleodgovernance...https://www.nytimes.com/201...

If stories 10 through 20 of your skyscraper are too old, they need to be custom redesigned and renovated one room at a time. You also need to keep the blueprints (source code) available so you can fix and upgrade later. You can not buy another building and just shove it in there.

Expand full comment

I've worked for one of the majors and I think it would be hard to dis-embed the software from the people and culture that surround it. Even if you required licensing of the source code I don't think it would be much use without a lot of the implicit knowledge that went into building, deploying and operating it.

One possibility would be to allow employees to license parts of the source code and let them spin off the work into a separate company. One of my friends is doing something like this, taking an internal product that one of the majors built and turning it into a company. Obviously he doesn't have access to the source, so he has to write everything from scratch.

From a regulatory perspective, I'm not really sure how this would work and would probably require regulators with a high degree of subject matter knowledge, so I'm not sure how feasible it would actually be. Eliminating non-competes and providing some kind of legal protection for former employees might help.

Expand full comment