Easy-Bake Ovens for AI?
Image: Boris SV/Moment/Getty Images

Easy-Bake Ovens for AI?

Protocol Enterprise

Hello and welcome to Protocol Enterprise! Today: the risks inherent in giving untrained employees access to AI, the decentralized web suffers from a central point of failure, and the dirt underneath the sheen of Google’s vaunted SRE organization.


Spin up

The ongoing worldwide chip shortage, unsurprisingly, is good business for chipmakers. An astonishing $583.5 billion worth of chips were sold during 2021, according to Gartner, a 25.1% increase compared to the previous year.

The risks of democratizing AI with low- and no-code tools

A growing group of tech companies including C3 AI, Dataiku and DataRobot want to “democratize” AI. They say with software that requires little to no coding at all, business executives can become “citizen data scientists.” But even people who see the value in these tools worry that gifting amateurs with AI powers can put data security, business intelligence accuracy and AI transparency at risk.

Demand is clearly there: Not only is it tough to hire data scientists, but even lowly marketing associates have uses for Easy-Bake AI, according to those vendors.

  • Low- and no-code AI tools rely on visual interfaces with drag-and-drop functions for building machine-learning models.
  • They can serve a variety of everyday business needs, from performing repetitive data-input tasks and generating invoices to watching out for equipment failure.
  • Sometimes these AI tools simply automate processes that in the past would have required more manual labor using spreadsheets.

But AI and data practitioners are on guard. Despite the promise of democratizing AI for everyone, machine-learning experts and people who manage data security are advising caution.

  • No-code AI business users “need some form of guardrails without slowing them down,” said Anthony Seraphim, vice president of Data Governance at Texas Mutual.
  • Ecommerce giant eBay is also wary. “We have to be very careful as we do this, because we need to understand what’s being put into production in front of our customers,” said Nitzan Mekel-Bobrov, eBay’s chief artificial intelligence officer.

Control over data security is one risk. So is making bad business decisions.

  • Data is at the heart of AI, of course. But if people without proper training feed no-code models with inaccurate data or information that’s not appropriate for the model’s purpose, there could be problems.
  • And if business users connect data sources without proper IT and governance oversight, they might inadvertently use low-quality data, creating inaccurate machine-learning models that produce bad advice.
  • Companies including C3 AI and Google allow users to define who can access low- and no-code AI models or specific data flowing through them. “As I log in as a citizen data scientist, there are objects and services that I can use and see, and there are others that I can’t, because I shouldn’t,” said C3 AI’s president and Chief Technology Officer Ed Abbo.

Businesses want transparent AI, but it’s not clear they’ll get that with low- and no-code AI tools.

  • Some low- and no-code AI systems automatically generate the corresponding code that’s produced from what non-coders see, showing the steps taken to build the model and produce results.
  • But some AI practitioners question whether simply archiving the code and training data is enough to explain why a model does what it does.
  • “Remember, the training code of the model is not the model code,” said Krishna Gade, founder and CEO of Fiddler, which provides an AI monitoring platform. “It is hard to know how the model will make a prediction, and that creates mistrust in how to use it and how to assure customers the AI products are making the right decisions.”

— Kate Kaye (email | twitter)

A MESSAGE FROM DATAIKU

At Dataiku, we know data isn’t a destination, so we build tools for what comes after. Dataiku is the only AI platform connecting data and doers, enabling anyone across organizations to transform data into real business impact. Because there is no soul in the machine, only in front of it.

Learn more

Non-functioning tech

A database used by NFT marketplace OpenSea failed Thursday, which prevented new NFT owners from having their prizes delivered to many of their crypto wallets. The world somehow managed to soldier on despite this disaster, which is both kind of funny and illustrative of how hard it will be to build truly decentralized tech infrastructure.

MetaMask is a crypto wallet that stores NFTs, and it relies on OpenSea’s API to let it know when one of its users has purchased a new NFT. However, as reported by Vice, that API went down Thursday morning and remains down as of this afternoon, which threw a wrench into Twitter’s announcement that NFTs were now available for purchase as avatars.

Whatever your take on the wacky world of Web3, decentralized tech infrastructure is an interesting opportunity that a lot of smart people and deep-pocketed companies are working to advance. It’s also clear, however, that the earliest champions of the Web3 universe very much rely on the same centralized protocols, services and infrastructure techniques they bemoan.

And as any veteran enterprise tech infrastructure person will tell you: Once you’ve built something, it will be harder to swap it out down the road than you think.

— Tom Krazit (email | twitter)

Enterprise moves

Mustafa Suleyman joined Greylock as venture partner. Suleyman was formerly the VP of AI Product Management and AI Policy at Google and the co-founder of DeepMind.

Lori Castillo Martinez was appointed chief equality officer at Salesforce. Martinez was formerly the interim chief and head of Global Employee Relations and Equality.

Carla Piñeyro Sublett passed the torch as IBM CMO. Sublett is still with the company, but Jonathan Adashek will be the new marketing head.

SAP promoted Alex Klaeger to president of Middle and Eastern Europe. Klaeger was previously a managing director for SAP Deutschland and the head of Cloud Business for the MEE region.

Simon McDougall was hired as chief compliance officer at ZoomInfo. McDougall was previously deputy commissioner for the Information Commissioner’s Office in the U.K.

Lynn Martin joined Google Cloud as VP of the North American Public Sector. Martin is a former VMware executive who focused on government, education and health care clients.

Eamon McErlean joined ServiceNow as VP and global head of Accessibility. McErlean previously led digital accessibility initiatives at Nike and will lead inclusion for the Now Platform.

Adriana Gil Miner joined Iterable as CMO. The marketing platform hired Miner after leading marketing at Qumulo and Tableau.

Data consulting firm Artefact hired Wayne Campbell as partner. Campbell has previously worked for Accenture, BP and Deutsche Bank.

— Aisha Counts (email | twitter)

A MESSAGE FROM DATAIKU

At Dataiku, we know data isn’t a destination, so we build tools for what comes after. Dataiku is the only AI platform connecting data and doers, enabling anyone across organizations to transform data into real business impact. Because there is no soul in the machine, only in front of it.

Learn more

Around the enterprise

Google’s SRE department was once a blueprint for operations teams around the world. As Bloomberg reported Thursday, that’s no longer the case amid allegations of a toxic work environment complete with racism and ageism.

Tonga’s internet infrastructure is starting to recover a week after a devastating volcano eruption and tsunami, but remains heavily reliant on satellite connectivity.

Thanks for reading — see you tomorrow!

Recent Issues

Veni, vidi, Vendia?

Slack has a Quip handy