We need to talk about data science skills. Real talk about AI, deep and machine learning

Saw this story after it was tweeted by Puni Rajah  – Bosch to invest 300m euros in AI, employ 100 experts from India, USA, Germany. As ever India has done great job building its native skills base. But the fact Bosch needs to look outside Germany is telling. We’re seeing a lot of pressure owing to skills shortages, and companies, countries and cities everywhere are going to need to up their game to avoid brain drain. Silicon Valley is still the main place data scientists, in particular machine learning and AI specialists, are ending up.

Pierre Etienne Bardin of Société Générale expressed the need to be more active in hiring and training succinctly in a recent post on data transformation as the new digital transformation.

“You can’t wait for experts to come to you. They won’t come unless you’re Google or Facebook.”

Yesterday James Kisner, an equity analyst at global investment banking firm Jefferies, said that “IBM Appears Outgunned in the Talent War.” Part of his argument rests on IBM having fewer open positions for ML than competitors on Monster.com. Not the most scientific methodology. To Bardin’s point Silicon Valley firms have a halo, and the fact they’re held to very different standards by equities analysts means they can spend a lot more on compensation. Frankly if you had the chance to work with Dr Fei Fei Li at Google you’d be crazy to turn it down. Unless of course, you didn’t want to move to the USA, or even to the West Coast. Carnegie Mellon in Pittsburgh is another centre of excellence.

That said, let’s get real. The enterprise is hardly ready for big data, let alone AI/ML. The data transformation is in its very early days. This isn’t the end, or even the beginning of the end, but it may just be the end of the beginning. Google and Facebook are making money with machine learning, not for machine learning. So far the model is kind of similar to open source in that respect – Web companies are making money with open source.

It seems unfair (though of course them’s the breaks) to mark IBM down for consulting engagements with its customers to help them understand and realise the value of data science, training models and so on, given its competitors aren’t selling services to do that. Maybe Kisner knows something I don’t about someone selling AI/ML, in terms of revenues.

Before going back to the skills shortage a couple of conversations with IBM and Google recently are worth noting, with respect to the AI/ML market.

A few weeks back I met with Rob Thomas, General Manager, IBM Analytics. He said something kind of powerful. He said he wasn’t so interested in talking about AI/ML/Cognitive with Watson clients, but rather “decision systems”, because that was where the value was. Using IT to drive better decision-making. AI as augmented intelligence.

It would be easy to scoff that this was IBM’s problem, but we had almost exactly the same conversation with Google a couple of days ago, at the launch of their London data centre.

An analyst from another firm asked what they felt Google’s key competitive differentiator was – something they could explain to “Joe Bloggs on the street”? He said he thought the Google advantage was machine learning. But neither Tariq Shaukat, who heads up the Google Cloud Platform customer team, or Ben Treynor Sloss, VP of Engineering, agreed. They said they were cautious about trying to explain that as an advantage. Google was good at using ML, but it might not be the best idea to lead with it as a message, because frankly it’s hard to explain.

In the long term, yes. We’re at a significant inflection point. As Sloss put it:

“Cloud is going to change how things are built. AI is going to change what things are built.”

But for now this stuff is hard to explain and the customer isn’t ready.

After all, If we can’t hire people to use the tools, what’s the point of leading with a technology story? The UK has done pretty well lately in being seen as an AI/ML leader. The country’s universities, notably Cambridge, have been doing a sterling job of seeding the market, leading to acquisitions. Google acquired DeepMind, which in turn acquired Dark Blue Labs and Vision Factory. It’s not clear whether the Brexit shitshow will leads people elsewhere, but computer science at least for now is in pretty good shape in the UK.

Tel Aviv is another market rich in AI talent. Apple acquired RealFace there.

If AI/ML is going to a be source of competitive advantage, then the skills to deliver it are going to be at a significant premium. We need better online learning, more university courses, better funding, and a clear understanding that skills will be the difference. The irony of machine learning is that we need better human learning and teaching to take advantage of it.

Finally I’d like to introduce Ivan Beckley. He’s smart, driven, studying medicine, and wants to study data science at UCL for a year. He’s also black, looking for sponsors for the course, and I am looking to help him. If you’d like to improve the skills pipeline in groups that are under-represented in tech, Ivan would love to hear from you. He’s prepared to collaborate on his research agenda, and work as an intern with you in Summer 2018 if you’d like to sponsor his course. Let me know. You won’t find anyone brighter or more hard-working to collaborate with. This is his post about joining us at Thingmonk as a diversity scholar.

We all need to invest in data science skills. Let’s start with Ivan.

(Read this and other great posts  @ RedMonk)

LinkedIn Twitter
James, aka @Monkchips is co-founder of RedMonk, the open source analyst firm, which specialises in developer advocacy and analytics.