Industrial AI - Part 1: Lion Vision, AI vs Automation, and Why a Game of ‘Go’ Changed Everything.

What do Quality Street sweets, Vape Batteries and High Viz Jackets have in common?

They all got analysed by myself and George Hawkins at Lion Vision, in Manchester, on a cold December weekday - in search of understanding the potential of how Machine Learning can help businesses stay relevant and innovative, but also how the process of working with Artificial Intelligence, for all its promises, still clearly needs a ‘human touch’ to give it meaning and purpose.

This whole project got condensed into a 9-minute video below, (and a couple of 2-minute videos at the end), to show how Machine Learning can be trained to recognise Vape Batteries in electronic waste, helping to reduce change of chemical fires in recycling plants, like SWEEEP Kuusakoski, (which I visited in 2024 year as part of my Fight to Repair blog series).

If you’re in a hurry to get into the technical side of how we did things at Lion Vision, skip to Part 4.

However, I hope these first 3 parts are also helpful if you or anyone in your company is trying to navigate the touchy emotional aspects of introducing Machine Learning, without making people panic that ‘The Robots Are Coming!’ (usually code for ‘you’re making us all redundant’). I feel a look at the history of similar tech is worth the time spent - not just because it saves us repeating the same mistakes, but also to avoid ‘reinventing the wheel’ on certain things. There are no ‘easy answers’, but fundamentally, I believe it is about being ready for a shift in expectations about what work means to individuals, and the benefits of meeting those changing expectations on this journey of technical evolution. In almost every instance I can think of, humans and machines working in partnership is arguably the optimal goal, for a multitude of ethical and intellectual reasons, but also compelling financial reasons too, which I’ll attempt to unpack here. Each industry of course has some unique challenges and opportunities in using AI, and my discussion mainly focuses on those in Industrial Science and Engineering.

Automation vs Artificial Intelligence. Or, perhaps AI as Automation…

I’ve argued that ‘Artificial Intelligence’ is a ‘moving feast’, always becoming integrated into our lives:

When I grew up in Penrith, Cumbria, a rural town in the north of England, I can remember a new supermarket opening in the 1980s, and the entrance had ‘automatic doors’- new customers, used to doors with handles, were quite bewildered. This is easy to ridicule now, but this was of course a basic form of ‘AI’:

1. to detect a moving thing, [likely a human], in this case,
2. make an inference, [that they are moving towards the door and wish to pass through it],
3. [in response to this analysis - do a job], open a door, in this example.

We perhaps forgot that we ‘normalise’ AI so quickly, without even realising it, that unless we stop and look at all the ‘smart’ devices in our lives, we might overlook an innovation that is right under our noses.

I think this is why companies need to take a ‘creative reset’, and check if they have ‘doors that need simple AI to automatically open them’ or ‘advanced AI to perform super-human (or sub-human) tasks’. The former point most of us are likely familiar with, and although the Luddites famously resented ‘automation’ of weaving factories, the more empathetic perspective was not that they were ‘anti-tech’, they were just ‘anti-starvation’ through being made redundant overnight, in huge numbers, collapsing whole communities.

If we plan to innovate, be it Automation or Machine Learning, we need to do better than to just blindly replace human work to cut costs in the short term. There is more humanity and wealth to be generated by coming at this more holistically, and history would attest to that having been the case from the invention of the Printing Press, to abating the recent global Pandemic.

As Stanley Kubrick’s 2001: A Space Odyssey illustrates, in its opening scene with apes, utilising bones first as tools, and then later weapons, we contemplate how we have a choice. And yet this scene of an ape murdering another ape, then throwing the bone into the air in triumph, is then smash-cutting to a spaceship, which is (we presume) a non-violent thing, much as the bone is.

One message is that we have the power to use tools for good or bad. Machine Learning as a tool is no different in that regard, but yes, in that leap of 4 million years, a lot needs to go right to avoid our own demise. Technology will enrich the experience of being alive, through better food, health, communication, etc. but it will not simply ‘give us humanity’, any more than the bone or spaceship does. It is us who give technology meaning and rules to live by.

You Don’t Know What You Don’t Know // The Alpha Go ‘11 Dan Move’ Moment.

Above: Clips from the Alpha Go Documentary, where the computer makes an unorthodox move. And when world champion, Lee Sedol, realises that the computer is creating moves no human would play.

If you’ve not seen any Artificial Intelligence or Machine Learning Documentaries, I get it, they are often very nerdy and get carried away with Sci-Fi hyperbole. However, Alpha Go suffers from no such hype - it simply documents a moment where ‘Deep Learning’ computers played Chess and beat Gary Kasperov in 1996. Civilisation didn’t end. However, Chess is what you’d call a ‘brute-forcible game’, meaning you can compute all the moves in a match and just be ‘one move ahead’ every time.

The Chinese game of Go is so complex that you cannot simply ‘brute force’ it, rather you have to play the game, as humans do, by being truly tactical. To be clear, computers are not ‘sentient’, and I’m reluctant to say they are being ‘creative’, but the leap in strategic lateral thinking/computation was so advanced in 2016, that world champion Lee Sodul lost to a computer.

What is most interesting, and frankly, blew my mind, (and still does to this day), is that Lee was called a ‘9 Dan Player’ or a ‘Level 9’ player of Go - this is the ‘top’ level. The computer AlphaGo made a move (“Move 37”), which was so unfamiliar to humans, it was believed to be a ‘mistake’. The AlphaGo engineers ‘checked the code’ to see if it was a glitch, (the computer is able to flag its own mistakes interestingly), but to their amazement, the computer was steadfast in its decision as being a ‘good’ move. The computer went on to win, and when experts played the game back, they remarked [paraphrased] “perhaps we have seen a ‘10 Dan’ or ‘11 Dan’ move, and we don’t fully understand it - it’s not a typical or known human move”.

The mic-drop here is not what you think. Lee makes a spectacular comeback, not to beat AlphaGo, (he concedes he cannot easily do this and has stopped trying), but rather he invests his time in “Human + Machine” vs “AlphaGo” tournaments, which he sees as being the ‘winning combination’.

If there was ever a better metaphor for the mindset shift that we need, it is that of Lee Sodul. To realise that we may well eventually be bested by a computer in a tactical task, but that when combined with humans, and our ability to compensate and counterbalance the ‘raw logic’ of computers, we have a winning combination.

In case you hadn't spotted it, this is what Mr Spock and Captain Kirk represent in Star Trek, they are a perfect balance of Logic and Intuition, of Law and Justice, of Structure and Chaos, of ‘non-Human’ [Vulcan] and Human. It’s unnerving to engage with a ‘foreigner’ like a Vulcan, and even more so a ‘computer’, but arguably, inevitably, we will learn to intertwine our lives and thoughts with machines.

“Survival of the Most Adaptable” - and why Darwin was often misquoted.

Charles Darwin stressed the importance of ‘adaptability’ not ‘fitness’. This may sound like semantics, but in the case of human egos, this is an important point. We do not live in a world where Brawn automatically beats Brains, as anyone who’s walked through a financial district of a city can attest. For the past 50 years, since the rise of the internet, our global intellectual capital has surely caught up and arguably overtaken ‘labour’ in terms of remuneration. One could even argue that since the 1800s the power in society has been accumulated by those who can use or control ‘tools’ (Industrial Labour Savings ‘tools’ to be precise), in the most monopolistic and capitalistic ways.

However, if you can sense a contrary point coming, you’d be right…

Above: Excerpts from BBC Article about Prof. José R Penadés, and him finding an AI came to the same conclusion in his research, in 48 hours, which took him and his team 2 years to solve.

Assuming that the life of the Factory Owner is more comfortable than the Factory Worker is nothing new, (unless you’re new to the writings of Charles Dickens). However, society, since ancient Greek and Roman times, has reserved a special place for the ‘smart runt of the litter’ - or as we’d call them today ‘intellectuals’ - folks whose predecessors probably couldn’t hunt a mammoth, but who can design and build a trap for one! By ‘survival of the fittest’ rules, they’d likely be extinct, but Darwin said ‘most adaptable’ - and so showing one’s value to society was about ‘flexing’ your smarts, rather than your muscles, for such folks. For many years, this ‘adaptation’ has been revered, rewarded - and arguably it has in turn become somewhat aloof to the toil and drudgery of hard manual labour. All in all, intellectuals have so far had it pretty good recently as valued members of society.

And yet, I think even our lauded intellectuals are now feeling the pressure to ‘adapt’, when facing competition from Machine Learning. They too, perhaps for the first time in history, are feeling something close to what labourers have felt, albeit less so.

In the above article by BBC, Professor José R Penadés and his team realised that an AI was able to arrive at the same conclusion of a ‘superbug treatment’, in 48 hours, that took him and his team 2 years to crack. Now, there is some sensationalism in this article, which even BBC can’t help but ‘stoke the fires’ of AI paranoia at present, by failing to acknowledge that the professor merely gave a *short prompt* - now, to anyone in the creative industry, this is like saying “I could have come up with Nike’s ‘Just Do It’, or Apple’s ‘Think Different’, or painted like Jackson Pollack” - but a. You didn’t, and b. Those humans spent a lifetime working towards the power of asking the right question, or ‘prompt’.

The AI didn’t boot-up one morning and *decide* to figure out a superbug cure, and nor did it just ‘self-prompt’ itself to look at the right specifics of the problem. But just as Lee Sedol had to reflect on the enormity of this ‘artificial brain’ besting him at Go - intellectuals are realising computers are no longer just ‘brute force machines’ or another form of ‘Brawn’, but they are truly a ‘brain’ to be reckoned with.

What both Sedol and Penadés illustrate in their humility, is that both *adapted* to the inevitable truth that AI and Machine Learning will now most likely forge a symbiotic, or ‘co-pilot’ relationship with them.

In writing this article, for RS Group’s DesignSpark, who of course focus on industrial components for machinery both mechanical and electrical, I’m not going to get into the ‘Should AI Make Art’ debates (feel free to commission that one if you’re keen!). But there is a project with NVidia coming, so stay tuned for some novel insights and builds…

Industrial AI Blog Series Contents:

Part 1: Lion Vision, AI vs Automation, and Why a Game of ‘Go’ Changed Everything.

Part 2: “Dirty, Dangerous, Difficult & Dull” - The Case for Ethical AI Automation.

Part 3: Key ML Terminology: Are You 'Sorting Ingredients' or 'Baking Cakes'?

Part 4: ML Lessons from Lion Vision. AI Failures, and ‘Sensing Like A Robot’.

Part 5: Getting Started with Jetson Nano / Orin. And Why Octopus Brains ML Marvels.

Part 6: A *Shiny* Idea, Whilst at Lion Vision: “Hi Vis Batteries”. And Why You Need Underdog Engineers.

Are you looking for additional information regarding Lion Vision?

Request more info