IBM: Finding the right talent for AI
“IBM doesn’t do consumer AI.” It is right there in the name: International Business Machines – which means the US tech giant is focussed on use cases for enterprises over developing the next virtual assistant or chatbot.
The quote is from IBM’s chief AI officer, Seth Dobrin, who recently spoke on the topic at The Next Web conference in Amsterdam.
During a discussion with the Financial Times’ Tim Bradshaw during the conference, Dobrin used the example of large-parameter language models such as GPT-3 and DALL-E 2 to describe IBM’s approach, describing those models as “toys”.
For Dobrin, it is also key that the growing artificial intelligence market is “human-centric”, meaning finding the right talent to develop its foundational models.
“There’s two sides,” he explained. “One is the people that are hiring need to stop looking for unicorns. They need to really acknowledge that there’s also more types of skills that are needed in data science than just those of machine learning engineer, or a data scientist – you need people that understand this.”
He went on to illustrate that while many people understand software development practices, you need employees in AI to understand data visualisation and data journalism as well.
“We need to put the humans at the centre of AI,” he added. “We really need to have a more socio technical movement within the industry.”
His talk, which was dubbed “TNW talks careers in AI with Seth Dobrin,” saw the AI veteran offer aspiring AI-enthusiasts two key recommendations if they want to find jobs in this space. Firstly, learn to code – whether it is in Python or Scala or another coding language. Secondly, focus on getting involved in real projects using these skills.
“You can learn whatever you want to move, like Coursera, or Udacity, or something like that. But until you produce it in practice, you’re not going to own it. And there’s lots of ways you can do that.”
For Dobrin, coding classes are necessary – but not just for those looking to enter tech. He called for children to be taught coding at middle school or high school.
It can also be beneficial for increasing diversity within the sector – something Dobrin says is vital, especially in AI. Less than 10% of data scientists are Black or Latino, he said, while when a study was carried out looking at data scientist courses across the US in 2017, only 12% came from black or Latino backgrounds – well below the size of these communities within the US population.
Why is this the case? Dobrin said much of it stemmed from how jobs are promoted and advertised. Studies suggest white people, especially men, will apply for a role even if they are only 50% qualified for it, but this is not the case among those from other backgrounds.
“I encourage you, if you don’t feel like you’re qualified for an opportunity, either reach out to the person posting the job, or just apply for it anyway. What is the worst that can happen?”
Further to this, IBM has changed the way it recruits for its AI team, dropping any educational requirements from its job specs. Instead, Dobrin’s team are focussed on “mastery of skill” and how candidates demonstrate this. This, alongside the ability and willingness to learn new skills, is vital, the IBM exec said.
“One of the things that I really focus on when I’m interviewing people is questioning them in ways that challenges what they know now and challenges how they have learned things in the past, so that I can understand if this person can grow,” he added.
This stems in part from his own career, which has seen him switch roles and companies multiple times. Dobrin has been with IBM since 2019, but prior to this he held roles in several med tech firms, and as a scientist at Motorola.
One of the things that has kept me relevant is that I’ve changed career three times,” he said. “Every time I’ve changed career I’ve had to learn something new. You need to have that willingness to learn to survive in this world.”
What about red flags? Is there anything prospective applicants to IBM’s AI team – or other organisations like it – should avoid?
The AI chief said he was less concerned by gaps in CVs but by inaccuracies. If an applicant says they have so many numbers of years’ experience in a certain coding language, or in a product or services, at least make sure the thing has existed if their expertise has.
“Don’t overstate your qualifications,” he said. “And don’t just put down a laundry list of your achievements – what is more important is the outcomes.”
Subscribe to our Editor's weekly newsletter