Watson and the jobs potential of growing human
- Summary:
- IBM’s augmented intelligence system is claimed to be 18 months ahead of other AI systems in being able to identify the nuances of what we really mean rather than what we write or click on in social media. Training it such arts could be one opportunity for the fabled new jobs AI is expected to create.
The world is going to change more over the next five years than it has over the last fifteen. That, at least, is the opinion of Jeremy Waites, who describes himself as just a lowly storyteller, traveling around to telling tales about cool stuff.
Officially that makes him an evangelist, and the 'cool stuff' on which he evangelises is IBM’s AI system, Watson. I met up with him at the recent Cloud Expo in London to discuss some of the wider issues about where AI fits in the business and wider world, what it can bring to those parties, and what impact it might have on the way work changes – or indeed continues to exist.
Like a growing number of people in the AI business Waites is keen to stress that the 'A' does not stand for 'Artificial', but rather for 'Augmented', on the basis that the former carries with it all those scary, sci-fi connotations. The level of augmentation, however, is already getting to be quite significant, especially in areas where AI has a well-established job to do, such as sales and marketing.
Even relatively new ideas such as monitoring social media, the press, blogs and other sources of opinion on a product or service are now being stymied in ways that big data analytics can’t manage too well. This means more subtle, nuanced approaches are now needed, he says:
Everyone is now using a private messaging app, and even then people lie. Big Data gets it wrong all the time, like they did with Brexit and Trump. The big shift that I’ve seen that’s accelerated over the last two years is this idea of dark social and dark data. We all know dark data - it’s the stuff we can’t see and the unstructured stuff. But dark social is where everyone is now on a private messaging app so nobody can see what anybody is saying.
So customer comments made on apps like Whatsapp, Snapchat, WeChat, Line, Facebook messenger can’t be seen or analysed by normal big data tools that are largely looking for clicks on Twitter and Facebook. This is an area where Waite believes Watson currently has an 18 month lead over the competition. As well as tracking the clicks in it also tracking the emotional analytics behind the language.
This is based on the natural language processing in Watson, together with tools such as the tone analyser system. It also exploits the machine learning capabilities so it can be trained to identify the emotional undercurrents of what people are really feeling:
Looking at how many people are talking about it, that’s the old style command centre. What this is doing is identifying ‘what do people feel’. If I can see the emotional attributes of an audience, I can have a totally different conversation, as opposed to the transactional model of ‘you said this, so therefore we do that’. That’s not how the world works anymore. At the core, that’s what Watson is.
Gut feel
Waites referenced some research from Price Waterhouse Coopers, published early last year which found that 75% of executives, even when they’ve got all of the data in place and are ready to make that go/no go decision with a project, still make all of their major strategic decisions on the basis of gut feel. He also observed that many also seem to work to conveniently rounded up numbers.
He told a story of a sales director who tells the sales team they have to increase sales by 10%. Using Watson to look at the state of the market, the customer perception of the company, the emotional state of the sales team and the ability of production and logistics to make and deliver such an increase, he suggested it was quite possible that the system would point to a lower figure, say 7% increase, as being achievable. This would not only advance the business but also lift morale by creating an achievable goal.
There is certainly a need to at least temper the executive’s penchant for gut-feel decision making coming down the line, not least because, as Cisco has speculated, by 2019 there will be 1 million new devices being added to the Internet every hour, and not only will there be too much information for any human to have a rational gut response, but the increasing autonomy of business processes will mean many rules will change – not least because, increasingly, the goal will be to market effectively to machines.
Waites argues:
It might be the algorithm on Netflix, it could be the algorithm on Uber, it could be something at Amazon, Echo, Geebo, Airpods, Google Home; we’re going to be marketing to that to get our service if it can get that much data ingested into it and can process it.
The thing is, it needs really smart agencies and partners to make sense of it all because it's going to be trained over a long period of time. That’s always going to be at the root of Watson. The tech is going to do really useful stuff but now you've got to train it to the nuances of sarcasm in Sweden, humour in Hungary and the way the consumers are going to act differently in China versus in Korea.
Human nuances
One of the major changes he sees coming over the next five years, and one where training will be a vital component, is voice interaction. Watson already has a Tone Analyser and a speech to text natural language processor. The IBM contention is that, by the end of 2019, 30% of all Internet browsing is going to happen without a screen. Some 20% of all Google searches on an Android device are being done by voice.
What is more, a lot of those voice searches are sentences rather than just keywords. This is moving towards an important shift of emphasis. It is no longer just the nuances of textual language,; keywords and what he now calls `old school SEO stuff’. What is coming fast down the track is the ability to understand both the language AND the tone of voice used to say it.
IBM is working with the clothing brand, North Face, in just this area, with the aim of creating voice interaction. Waites explained:
I might want to ask it: ’I'm going to Stockholm on next week, I need a jacket’. It knows the weather, and asks me what I want to do? what's your favourite colour? Do I want zips? Do I want reflective? Now, based upon my tone of voice it might understand if I’m getting impatient, so it would serve me up something quick which might be 70% accurate. But if I’m really enjoying this interaction it might be a much longer conversation in which case it ends up with a 90% accuracy.
If Watson can understand the voice and the nuances of that then it can understand what mood I'm in, it could automatically change the customer journey and totally transform the experience that I just had with that brand. No one's going ever been able to do that before.
AI and jobs
The benefits of AI to both individuals and businesses are already clear, and only really limited by the imagination of future users. But it still carries with it a potential downside of permanent significance - its impact on employment. Most companies in and around the AI marketplace currently use – and some might suggest, hide behind – the mantra that new jobs will come, they will be high value jobs, and just look at history for past examples.
But AI has the potential to be very different. Farmers could become factory workers as they required different versions of hand skills and manual labour. But does that automatically mean that miners can become data scientists? And in Waites’ view it is data scientists that will be desperately needed.
In a way Waites says the same mantra, but does point to some fundamental educational changes that might help. And they might help create a whole area of new educational jobs – teaching instances of Watson, and other machine learning systems as they come along.
He sees Watson as an assistant, and that its role is in a man and machine, man with machine partnership, not man versus machine. What this does then require are users who are not just 'experts' in one core skill or area. Instead they will need a balance between tech skills and the humanities:
That's why I love the quote that got me into the industry, ‘we ask too much of technology and not enough of ourselves’. I bet 90% of the people here [at Cloud Expo] are analytical, logical. But are they bringing any emotion into understanding the emotional side the people side, like the Watson stuff?
So whoever it is looking at jobs now and how they’re going to change – they have to learn the opposite skill. So if you’re a data guy, learn more about people and humanity, if you're a PR person and you know communities and people inside out then learn more about data algorithms and cognitive behaviour.
My take
An interesting view of where new jobs may come from if Watson, and other machine learning systems come along. There will be a need for people to teach such systems, and continue to teach them as new developments occur and new applications areas open up. The would be an `applied data science’ type of job that might just generate work out into the future, for there will always be human nuances that need accommodating – even if they are just the ones that creep into algorithms.