Issue 27

Artificial intelligence is overrated

Delivered on 08 April 2019 by Justin Pyvis. About a 5 min read.

Artificial intelligence (AI) is overrated, both in terms of its potential for improving productivity and the threat it might one day pose to humanity. When you break it down, AI is really nothing more than various algorithms that mimic human behaviour. There's no real knowledge or understanding; it's just a very fast, enormous database containing information about how to best react to certain events.

Don't get me wrong: AI will, given time, replace or complement many jobs where the goals are clear and following the rules is critical. Much of the world's supply chain is ripe for AI disruption, for example. But AI will never be able to innovate and solve problems in uncontrolled environments, something humans do all the time. We don't live on a chess board with a finite number of strategies and options but a dynamic, ever-changing and evolving place we call the universe.

I see far too many people and companies are over-promising when it comes to the potential of AI, it stinks of snake oil and I now actively avoid those that claim to rely on it to any significant extent. I have no doubt that AI is in a bit of a publicity bubble and many of the companies supposedly involved will go the way that those adding "blockchain" to their names/descriptions did when the Bitcoin bubble burst.

Then there are those who over-hype the potential risks AI poses to humanity. For example, you may have seen the video that researchers at the University of Washingtonput together of former President Obama, where they - to a high degree of accuracy - could basically make him say anything they wanted.

Such videos are known as a "deepfake", or AI-generated videos that look and sound like the person in them but are entirely computer generated. As the New York Times reported:

"It's not hard to imagine this technology's being used to smear politicians, create counterfeit revenge porn or frame people for crimes. Lawmakers have already begun to worry about how deepfakes could be used for political sabotage and propaganda."

But without some kind of credible verification - an organisation such as the New York Times, for example (although I'd want something more reliable!) - why would anyone believe a video was real, any more than they would a quoted block of text?

This fear of "fakes" is not new and AI-generated fake videos are just the latest iteration in a long history of fraud. Thankfully, it's actually easier than ever to prove the credibility of a source now that we have the internet to speed the whole process up, along with advances in verification techniques such as digital signatures. As The Guardian notes:

"If there is a long history to fears about fake news, there is a long history to fake news too. In 1924, four days before a general election, the Daily Mail published the forged Zinoviev letter, a supposed directive from Moscow to British communists to mobilise “sympathetic forces” in the Labour party; Labour lost the election by a landslide."

"In 1989, the Sun, fed lies by the police, ran a campaign to besmirch Liverpool football fans after 96 supporters died at Hillsborough, crushed to death after being forced into an overcrowded caged pen. The Sun invented stories of drunken fans as the cause of the disaster."

"In 2003, in the run-up to the Iraq war, articles about Saddam Hussein’s nonexistent weapons of mass destruction filled newspaper pages across the world."

The problem of fake news is no worse with the internet and AI-generated videos than it was when letters and newspapers were delivered on foot. Governments will still lie to their people, journalists will still invent quotes and people who want to verify claims will still have to seek out the original source, or at least some kind of proof (e.g. public key digital signatures) that the source is in fact credible.

Americans still love Apple

There has been a lot of hate for Apple this week following its underwhelming Spring product launch (note that investors don't seem too concerned; its share price is up more than 5% since the event). But Apple is still dominating the American market, and I expect that's where it will get most of its recurring services revenue. 50% of all phones sold in the US are still Apple devices, and Apple Music just overtook Spotify in that market despite Spotify's superior product.

Learn more:

Privacy is the story of 2019

I've been saying it for what seems like forever but it seems Facebook et al have finally pushed the privacy boundaries a bit too far. Australia with its bipartisan "just ban it" political culture is leading the way, along with the more well-reasoned European Union and now Singapore. It's not the right approach - and Zuckerberg's proposal is by far the worst of the lot- but I fear there's enough momentum behind it now that it's all but inevitable. If you don't have one already, now's the time to invest in a good VPN.

Learn more:

The future is crypto (payments)

I don't know which protocol(s) will win in this space, but I sure hope it's not Facebook's yet-to-be-released payment platform. This one is promising:

"Celo's protocol is designed to be mobile-friendly and "ultralight," allowing funds to be sent to cellphone numbers... it aims to enable easier remittances, cash-transfer programs and micropayments through its open-source platform, taking an initial focus on helping unbanked or underbanked people."

Learn more:

Other bits of interest

Image of the week

"From 2018 to 2019, the average correlation coefficient across the top 200 cryptocurrencies by market cap decreased from 0.89 to 0.58."

The wheat is finally being separated from the chaff in the crypto world. Only the best coins will survive.

Issue 27: Artificial intelligence is overrated was compiled by Justin Pyvis and delivered on 08 April 2019. Join the conversation on the fediverse at