Hopes and fears of an AI future

In these summer months, lunchtime walks have become a near-daily event for me. Even when it isn’t sunny it’s at least warm. While I walk I tend to listen to something stimulating – a radio 4 program or podcast – but I’m fast exhausting my familiar listening material.

This week I’m giving Sinica a go. It’s a podcast broadly about China, where the hosts invite a special guest to help them cover one specific issue each episode. Yesterday lunchtime I listened to an episode from June, Kai-Fu Lee on Artificial Intelligence in China, and I include below a ~6 minute segment from the end of the episode, which is the focus of this post.

This episode’s guest, Kai-Fu Lee is a very influential venture capitalist investor in Chinese tech startups. Like most of Silicon Valley, he is exuberantly optimistic about the social good of entrepreneurship, the power of emerging technologies and the enormous moral and ethical dilemmas fast approaching.

In fact, in addition to this naive/ignorant/self-interested optimism, throughout the episode I found each speaker much too apologetic for the Chinese government’s ever increasing censorship and collection and arbitrary use of ever more data of their citizen’s lives. Two topics stood out. First, as they broadly agreed that Google made the wrong decision to give up walking the censorship/truth tightrope and pull out of the Chinese market in 2010. Surely the incremental tightening of internet freedoms, visa restrictions and government insistence on data harvesting since then precisely justifies their decision. Each month would have brought another strained meeting, another small sellout, with it being increasingly difficult to draw the line right there and walk away from ever more accumulated progress.

Second, they again broadly praised the enthusiastic optimism of the Chinese towards new technology, contrasted with Western angst about ethical and moral dilemmas, calls for regulation and oversight and predominantly dystopian views of a sci-fi future. I would say that Chinese optimism is driven predominantly by ignorance, as the government hide from the people every accident, mistake and abuse of power they can get away with, and a lower value on individual, private human life. If a self-driving car misjudged a corner and ran over a peasant road sweeper in China, I doubt it would even make the front page of the local newspaper. Lee talks at length about a microloan app he has invested in. When a customer applies for a loan on 用钱宝 (yongqianbao), it scans every aspect of their smartphone, from its make and model, to the things they take photos of, to their contact list, and with this intimate life-snapshot, in eight seconds decides upon creditworthiness. Who from the UK would even consider such an ludicrously invasive ‘service’?

Yet having said all that, toward the end, they summed up almost perfectly my own thoughts about the coming AI revolution.

I hope that listening gives you lots of food for thought. I’m just going to pick up one point, universal basic income (UBI). Early on, Lee covers this, saying:

“The AI will make so much money that the possibility of giving a universal minimum stipend is actually quite realistic”

I would go far further than that and say that governments must insist upon its rapid implementation.

Universal Basic Income

In the future that awaits us, the basic necessities of human survival will be located, farmed, harvested, processed, manufactured and delivered without any day-to-day human involvement.

Think about that.

In 2013 a University of Oxford study estimated that “47 percent of total US employment is in the high risk category . . . jobs we expect could be automated relatively soon, perhaps over the next decade or two.”

Computer algorithms will calculate when to begin the growing season, a robot will sow the seed, nurture it and harvest it when instructed by the same AI program. Self-driving vehicles will deliver the crops for further processing by autonomous machines, and eventually directly to the door of humans, for our consumption. When we feel unwell, a robot bristling with probes, sensors and connected to a database of every medical event ever recorded in the history of modern medicine will asses us, present a diagnosis in seconds, and dispense the precise dose of drugs required. If it believes we need further treatment, it will submit our application into a central, self-learning database that immediately allocates us an appointment slot that fits precisely into our schedule and its.

The services described above might be cheap, but they aren’t going to be free, yet human employees are nowhere to be seen. How will we pay our way in this world? Or rather, who will be making the money when we are charged for them? The owners of the land, the intelligent robots and the machine-learning algorithms. Their business overheads will be low, capital vast, and their power will be monstrous. Just like human children, smart machines take years of learning before they can reliably make autonomous decisions. Their ‘minds’ will represent not only vast stores of knowledge, but also vast accumulations of capital.

In this future, the only way to avoid mass destitution and widespread social unrest will be to forcibly allocate a part of the profits from this marvellous new exploitation of capital to everyone – a universal basic income. It should be enough each month to live on, roughly equivalent to the current living wage. You can find a list of simple, logical reasons for UBI here. Even viewed in simple profit-sharing terms, each smart machine will have learned from our lived experience and the decisions we all made. The value within them was once within us and we should be recompensed for its utility.

In such an economy most of us, who currently sell our labour – our time and skills – to earn a wage to live on, lose even that method of survival. There’s no reason to give us a job when a robot can do the same role, 24/7, with the only ongoing costs being electricity and repairs. Furthermore, the main prerequisite of machine learning – enormous amounts of data on a subject – is only available to early adopters, or large, mature companies. They might license out their algorithms, but only to other businesses. If you don’t have some capital you can invest in this technology, you’re not going to have an income to live on.

UBI is also particularly suited to the work that a great many of us will end up doing. As Lee notes, “machines cannot create”. With a basic income to support us as we find our talent and interest, while we try and fail, through our writer’s block, while we accumulate first, second, third and fourth drafts, the full creative power of every human can be realised.

Finally, ~250 years since the industrial revolution turned us all into biological machines, repeating the same task over and over, with “the great increase in the quantity of work that results”, we’ll be able to shrug off those alienating chains. Work in the future will, finally, be human. Creative, empathetic and with intrinsic meaning.

Blockchain

I’ve been looking into blockchain as a technology of the future quite a bit recently, and it could offer the widespread opportunity to at least earn income and survive if governments don’t step up and equitably distribute the profits of robots to their citizens.

Without going into too much detail, blockchain is a method of transaction verification without any need for a trusted central authority. It works in a completely decentralised manner as each user adds their transaction to an ever growing historical chain – a blockchain. Hundreds of thousands of different devices all around the world power this, each adding new blocks into the chain every day. Each new transaction verifies every previous transaction, creating an indelible historical record of exactly what changed hands and when.

This is currently used almost exclusively for monetary transactions and virtual currencies, with Bitcoin being the most popular and well known. However, another, Ethereum, is by far the more interesting, as Ethereum has been designed with smart contracts in mind. A smart contract could be a way of recording anything, and specific events can trigger certain actions. It’s an impartial, transparent ledger that individuals and groups can sign up to. Imagine, for example, that title deeds to land and property were held in a blockchain. In a side chain, the buyer and seller agree to a very simple chunk of code that says;

The buyer will deposit 10,000 Bitcoins into this account before 27/07/2017. If this is completed, then on 27/07/2017, the title deed will transfer from seller to buyer and 10,000 Bitcoins will be transferred to the sellers nominated account. If this is not fulfilled, the buyer will be refunded all their Bitcoins immediately.

This replaces the arduous signing of various contracts and waiting as they go back and forth, payment for legal advice and payment for trusting law firms and banks to hold money in escrow with one, cast-iron smart contract. Humans don’t need to oversee any part of the transaction. It’s automated. The agreement is understandable in plain English and enforces itself immediately, resolving to either outcome. Currently, if trust between two parties isn’t honoured and a contract broken – say a business fails to pay a supplier on time – the only recourse for the plaintiff is to begin long, expensive legal proceedings. The time and expense of such proceedings effectively puts justice out of the reach of many.

In addition to managing currencies without any need for either central government banks or private banks (we all hold our own wallets and verify each other’s transactions), blockchains could be used to 100% faithfully record credit history, medical records or identity information.

The economic potential of these chains for us all is that they require diffuse computing power to operate. Each new block is committed to the chain through a proof of work exercise. By their nature, blockchains require a distributed base of computing power, since if you trust transaction processing with just one organisation, you might as well use a traditional static database system. When a computer connected to the network decides to pick up a transaction in the waiting queue, it will do some computing work (complex maths) and after completion receive a small fee. Transactions posted to the queue without a fee attached are highly unlikely to be processed by anyone on the network.

In the future, perhaps, in the space currently occupied by our hot water boilers, a household computing device will verify blockchain transactions, earning us enough in fees to cover our daily expenses, and use the heat generated to warm our water.

 

Leave a Reply

Your email address will not be published.