Five big mistakes that people make about AI

A few weeks ago, I was interviewed by Tim Hughes, of DLA Ignite, about the five biggest mistakes that people make about AI and its impact on the workplace. This article is based on the full interview, which you can find here.

1. Automation is the same thing as intelligence

Automation is not intelligence. It is the use of machinery to replicate a unique human activity, from punch cards that operated sophisticated weaving looms during the industrial revolution, to mid-twentieth century business computers that calculated bills and operated the payroll. Very often these activities are repetitive, error-prone, and in some cases life-threatening.

The principle of automation is nearly always the same. A machine does one thing faster and more accurately than a human operator. But it does not add intelligence, in the sense that it cannot learn and adapt to changing conditions or data.

This is the fundamental principle of artificial intelligence, where technology to some extent replicates human reasoning to ascertain outcomes and decisions – often using machine learning. If a punch card operated loom could learn to design and add new patterns, it would manifest artificial intelligence. But of course, it can’t.

Another important distinction is the sheer number of AI applications that impact a single industry. Returning to design and fashion, AI and machine learning are reshaping the industry from detecting popular colours and trends to designing clothes and accessories. Putting it another way, AI can be applied in almost any business situation where there is enough data from which to learn and evolve.

2. AI can write books and compose songs

There’s a growing tendency, especially with headline writers, to take an AI breakthrough in a given field and then predict the death of a profession. So, despite what you might have read, AI cannot write books or even blogs without human intervention. The same goes for music. AI can compose tunes and manage orchestration, but it still requires a human being to stitch together this output into a coherent piece of music.

You can input a paragraph (or the opening bars of song) and the software does pretty well for a short while before deviating from common sense or anything related to the original song. You can hear this during an experiment where the software is fed the opening seconds of Hotel California.

See how far you can get before you tear off your headphones. AI commentators politely refer to the software “hallucinating”, which is a polite way of saying that without intervention the software spews out nonsense.

Instead, composers and journalists are making use of AI’s creative and analytical potential. Here’s an example of a song based on the later works of the Beatles. AI software generated many different themes and tunes, but it required an expert composer, Benoit Carré, to curate and combine these elements into a complete song.

The biggest risk posed by AI is not the destruction of jobs, but the dissemination of fake news and ‘deep fake’ video and audio. This recital of Hamlet, by Jay-Z is based on audio copied from an interview with the musician. The software only needed a few seconds of speech to be able to replicate his voice convincingly.

There’s more to this than just the odd fake news story or video. The real danger occurs when software churns out massive volumes of content that confuses or overwhelms social media algorithms put in place to filter out inaccurate and misleading narratives.

3. AI is a once in a lifetime upgrade

Let me let you into one of AI’s biggest secrets. It’s incredibly boring. I exaggerate slightly, but there’s the belief, propagated by blockbuster films, entrepreneurs and certain parts of the media, that AI is the once in a lifetime fix that will transform your business – and humanity.

Not so. Right now, the most successful AI applications are quick, pragmatic fixes to specific problems that run behind the scenes without the end-user ever knowing.

Take video conferencing, something we've all grown to live with over the past six months. During the chat I often move the computer to make space for a document or to type a message to other participants. I often fail to notice that I’m slipping out of view, especially when I minimize the view of my own camera. Not a good look.

Microsoft fixed this problem in the latest generation of their Surface laptops by training the camera to follow the eyes of the user. The software ensures that you always appear to be looking directly at the camera, even when you aren’t. The beauty of this unspectacular solution is that it solves a widespread problem, that potentially improves the call experience for millions of people.

AI has the biggest impact when it runs in the background unobtrusively. Rather than looking for expensive, big-fix solutions from third-parties, businesses need to look more closely at their internal systems and their data. Which brings us neatly to mistake number four…

4. AI is difficult and expensive

In recent years, the cost of AI technology has plummeted. In the same way that Amazon, Google and Microsoft made their datacenters available to customers in the form of cloud computing, they now offer machine learning platforms based on many years’ experience of training software with massive amounts of data.

So if your business is sitting on large amounts of historical data, or processes data in real time, you can use an AI platform to improve the performance of your business from sales strategies to supply chain decision making and marketing. Companies, such as Spotify, use Google’s Cloud platform to refine home screen recommendations based on end user data.

But you don’t have to be a billion-dollar business with millions of customers to benefit. These cloud AI companies can help you ‘start small’ and guide you through the complexities of building a permanent AI capability into the business. In the past I’ve called this ‘cold-case AI’. It’s tech that enables you to go back and solve problems from years or even decades earlier.

Earlier this year, scientists from Warwick University in the UK, used machine learning to study datasets collected by NASA telescopes and identify 50 new planets. The information was always there, but it required modern, lightning-fast machine learning to separate false positives from the planets themselves. Every business is sitting on top of a gold-mine of data. You just need the right technology to dig it out.

5. It’s too late for me to find an AI career

By the mid-2030s one-third of existing jobs could be replaced by AI and automation, according to a report from PwC. Is your job in danger? If it’s focused on one activity and highly repetitive then it could be at risk. But in the majority of cases, AI is rewriting the job description, creating more roles than it destroys.

Journalism is an excellent example. Breakthrough news stories, such as the Panama Papers scandal, required international teams of journalists, using machine learning software, to sift through thousands of documents to uncover money laundering and tax evasion.

The Radar News Service, in the UK, is another good case study. Its journalists use AI technology, to draft articles based on data sets released by the UK government. Using their investigative skills, the journalists identify data from which they can derive a story and then build a template into which the data and standard phrases can be assembled.

For journalists to flourish in this environment, being able to work with automation templates is essential. Some simple programming knowledge is also required. Skills such as these can also help copywriters and content marketers remain relevant and rewarded.

The trick is to look at your profession and understand its direction of travel shaped by artificial intelligence. What new skills do you need to learn? Which sectors or businesses are the ones most likely to hire?

If you want to get into the AI industry itself, as a data scientist, machine learning researcher or an AI engineer, there are dozens of online courses that can help you get started. Some companies, including Google, now require job candidates to have passed through their own curriculum to apply for a role.

Here are courses from three of the biggest names in AI:

IBM’s K-12 course (aimed at schools) is also an excellent introduction for beginners.

For daily roundups and breaking AI news, follow me on Twitter and LinkedIn.

An archive of artificial intelligence stories and more can be found on my Flipboard magazine.

Feel the connection?

© 2020 Peter Springett.