Natasha Crampton: The Kiwi charged with making artificial intelligence ethical at Microsoft

AI will upend our world. Yet governments in most countries, including ours, are still at a thinking-about-maybe-regulating-it stage.

In the meantime, it falls largely on the tech firms who make AI (artificial intelligence) to set their own ethical guidelines - to ensure the technology serves humanity rather than derailing it.

In the case of Microsoft, that mission is led by New Zealander Natasha Crampton, the tech giant’s chief responsible AI officer, who leads a team of 350 people around the world charged with making sure Microsoft’s artificial intelligence tools are designed and created in a responsible way.

Microsoft is of course huge in tech, though you might not appreciate exactly how huge. With its US$2.4 trillion ($4.1t) market cap, it’s the largest company by valuation in the world, after Apple.

In January, it spent US$10 billion to lift its small stake in OpenAI - the maker of ChatGPT, the generative AI tool that had been taking the world by storm since November - from a small holding to 75 per cent control.

Microsoft then melded ChatGPT to its Bing search engine, and Office and Teams products.

Natasha Crampton, the New Zealander who serves as Microsoft’s chief responsible AI officer, at the tech giant's headquarters in Redmond, Washington, Photo / Grant Hindsley, The New York Times

How did Crampton land in her current role?

“It was that ‘a lot of hard work and a little bit of luck’ as is often the case,” said Crampton, who spoke to the Herald from Microsoft’s headquarters in Redmond, near Seattle, where she’s now based.

“I’ve always been interested in the intersection between technology, law and society, and that came pretty early on including what I studied at the University of Auckland.”

Her honours degree saw her specialise in intellectual property and technology law. She worked for a series of law firms, including Simpson Grierson, before becoming Microsoft Australia and New Zealand’s in-house counsel in 2011.

Baptism of fire

“Eventually I decided I wanted to move to Redmond and be a part of the teams that are setting the strategy here. I came over for one particular job [in 2018] that very quickly became a role focused on all things AI and ethics,” she said.

She was named Microsoft’s senior attorney for AI, research and human rights - and was immediately in the thick of events involving the Trump Administration’s efforts on the US-Mexico border.

“As it turns out, the day I arrived here in Redmond, there was a story about the alleged use of Microsoft’s facial recognition technology, at the height of the southern border crisis,” Crampton said.

“The allegation was that Microsoft’s technology was being used to separate parents from their children, which was incorrect. So in lots of ways, that particular story, which was false, nonetheless shone a light on facial recognition technology. And thatcharted my path working on these issues.

“About a year after that [in 2019], I stepped into the lead role when we decided to create the Office of Responsible AI”. In January this year, she got “vice-president” added to her title.

Who’s doing it right?

In New Zealand, there is no AI-specific regulation or legislation, and internal MBIE documents say there are “no rules or guidelines for all government agencies about staff use of AI tools”.

There have been some ad hoc moves, which lean toward throwing the baby out with the bathwater. The Ministry of Business, Innovation and Employment (MBIE) had banned staff from using artificial intelligence technology such as ChatGPT since March, citing data and privacy risks. So has Te Whatu Ora (Health NZ) - although ironically the Department of Internal Affairs, which steers a lot of government agency tech usage policy, has not.

Does Crampton think there are examples of countries that are getting regulation right?

“What’s been really interesting, especially around the last six to nine months, is how the conversation about the right guardrails for AI has really broadened. The Europeans got the ball rolling,” Crampton said.

By “broadening” she means taking the focus from individual apps, and looking at wider themes.

“Just a few weeks ago, the US Government and Microsoft and other leading AI developers announced a series of eight different commitments to make sure that AI systems are safe and secure and trustworthy. So I think that was an important initiative. And there’s lots of other activity around the world as well. The UK is going to be hosting the first global AI summit. Japan, as leader of the G7 this year, is very engaged in these issues.”

The White House commitments are broad strokes and voluntary (read them here), but the Biden Administration has also promised government procurement guidelines for AI; the President has signed an Executive Order that directs federal agencies to root out bias in the design and use of new technologies, including AI, and US$140m in federal funds has been earmarked for AI research. (This is a similar effort to the Australian Government’s A$101m critical technologies fund; there was no equivalent in NZ’s Budget.)

“I think the jury’s still out on the single best approach to AI regulation. But I think what we’ve seen is some really healthy thinking and forward progress,” Crampton said.

Kiwi AI stars

ChatGPT is “not even the most cutting-edge technology, but was the first time most people were able to see what AI models are capable of, because of its simple interface”.

“Since then [ChatGPT’s first public launch, last November], we’ve seen this broad experimentation with the rough edges of these AI models; people probing them, figuring out where they’re really good at and figuring out where they’re not so good at and their limitations,” said Crampton.

She is naturally quick to spruik the way her company has added the latest version of ChatGPT (version 4, version 3 is the freebie) into its own products, allowing you to, say, take a Word document and ask the AI to turn it into a PowerPoint.

But Microsoft has close relationships with three New Zealand companies who’ve introduced recently added AI elements to their business.

The tech giant has signed an R&D partnership with Volpara, the Wellington-based, ASX-listed firm that’s developing AI systems to enhance its software for analysing mammograms.

LawVu, the Tauranga-based startup that’s raised more than $54m in venture capital for its software for managing in-house legal teams, has developed “AI Assist” using Microsoft’s Azure OpenAI platform. The new LawVu feature is a natural language processing tool that reviews and summarises key information from contracts so that in-house legal professionals can find answers quickly without needing to read hundreds of pages. Lawyers “train” the software as they go by rating its responses.

"In 2018, Beca examined the potential of what has evolved into Generative AI and decided that if the technology developed as we predicted then it would be essential that Beca was using it," says the engineering firm's Matt Ensor. "In 2020 Beca launched an internal startup, FranklyAI, allowing the organisation to integrate the technology in a secure way and explore its commercial applications. As a result, in the last few months, the FranklyAI for Microsoft Teams app has been utilised over 100,000 times within Beca and is being used by around 2000 organisations globally. This is a great example of a New Zealand company being a global leader through proactive strategic investment." Ensor heads Frankly, and chairs the AI Forum NZ's Large Language Models (LLM) Working Group.

And Crampton says engineering firm Beca has also grabbed global attention with its FranklyAI - a tool that adds AI smarts to Microsoft Teams.

It was created by an in-house team led by Matt Ensor - previously head of Beca’s smart cities unit - using natural language (via ChatGPT4) to replace questions and surveys as the firm looked for a way to streamline internal comms and public consultation over various construction projects.

“They added FranklyAI to Microsoft’s store, and all of a sudden 800 companies around the world (and counting) are actually using it,” Crampton says.

She sees a lot of local firms experimenting with what apps they can build on top of Microsoft’s AI tools.

What fears that AI software will replace human workers?

Beca’s Ensor told the Herald: “I always say AI won’t take your job. Unless you’re not using it.”

Natasha Crampton

Age: 41

Born: Auckland, New Zealand.

Family: Married to Murray Crampton (who works for Vulcan, the philanthropic venture founded by the late Microsoft co-founder Paul Allen) with two sons aged 9 and 6.

Education: Karaka Primary School and St Cuthbert’s College; University of Auckland (BCom/LLB Hons) with one year on exchange to the University of New South Wales to study e-commerce law.

Career: Law firms in New Zealand (Hudson Gavin Martin, Simpson Grierson) and Australia (King & Wood Mallesons) before joining Microsoft 12 years ago. “I’ve been in Redmond for five years; my first seven years were working in the New Zealand and Australian subsidiaries of Microsoft.”

Lives: In Kirkland, Washington, just down the road from Microsoft’s headquarters in Redmond.

Now reading: Yellowface by RF Kuang.

Last holiday: Auckland-based adventures with family over the Christmas/New Year break.

Chris Keall is an Auckland-based member of the Herald’s business team. He joined the Herald in 2018 and is the technology editor and a senior business writer.