July 15, 2022
Hello, and welcome to Protocol Enterprise! Today: how Twilio is thinking about compensation and employee retention in the wake of a SaaS stock-price correction, Matt Hicks takes control of IBM’s most valuable asset, and how machine-learning exploits could all but guarantee permanent employment for cybersecurity pros.
Twilio’s path to profitability just got a lot more difficult. Recognizing the need to translate fast growth into profits, last quarter CEO Jeff Lawson told Protocol he had wanted Twilio in the green by 2023.
Then, the market downturn hit. And as a SaaS company catering to a large number of marketers, Twilio might not be as immune as other mission-critical software like an ERP or CRM system.
Employees who have their compensation tied to Twilio’s share performance might be feeling a little grumpy right now.
If you’re a Twilio employee, don’t expect the company to rescue you from the realities of the stock market.
But Twilio will have to make some adjustments if it wants to retain employees — even at the expense of profits.
They created Digital People. Now they've made celebrities available as Digital Twins:Soul Machines co-founder and CEO Greg Cross and his co-founder Mark Sagar, Ph.D., FRSNZ are leading their Auckland and San Francisco-based teams to create AI-enabled Digital People to populate the internet, at first, and soon the metaverse.
Matt Hicks, who started as a developer at Red Hat in 2006, this week succeeded Paul Cormier as CEO of the enterprise open-source software company.
“I started porting Perl applications to Java in IT, and it has been a bit surreal to be in the CEO seat now, but it's certainly been an exciting journey,” Hicks, who most recently had been Red Hat’s executive vice president of products and technologies before his promotion, said in an interview with Protocol.
“I spent many years in IT, and then coming up through the [Red Hat] engineering ranks and working hand in hand with customers … has given me an intuition I really lean on in terms of the challenges customers are dealing with and the types of technology that can help solve them,” said Hicks, who was part of the team that developed Red Hat OpenShift, the hybrid-cloud, Kubernetes application platform. “I think that intuition, combined with the market opportunity right now, it's a good combination. It'll really give me a good base of experience to fall back on in this role.”
Hicks had worked with Cormier, who will remain as chairman, on Red Hat’s strategy, including delivering on open hybrid cloud.
“The really exciting part is hybrid's going to extend to [the] edge,” Hicks said. “From our view of the market, the spend is going to increase to about $250 billion dollars by 2025. You're going to see a tremendous increase in applications that are built for it.”
When it comes to OpenShift, Red Hat will focus on serving new workloads.
“A really important area for us will be in the artificial intelligence or machine-learning space, where we're doing work with areas like OpenShift Data Science,” Hicks said. “That bridges our knowledge of hardware from accelerators like Nvidia to this new class of developers in data. Those two worlds of traditional application development and then model- and AI-based development — it's going to be a really exciting convergence of those. But then you'll also see OpenShift continue to get lightweight, and we'll do this with Linux as well, because the edge use cases require just a smaller footprint.”— Donna Goodison (email | twitter)
There’s already plenty of worrying going around over cybersecurity attacks that could be supercharged with AI. But the components of the everyday machine-learning models used by businesses — to automate filling out forms or deciding whether to show an ecommerce site visitor the pleated khakis or the Y2K-inspired denim — are themselves susceptible to all sorts of subterfuge.
Chris Anley, chief scientist at cybersecurity threat consultancy NCC Group, detailed the myriad ways ML models can fall prey to attack in a new report offering technical explanations of novel ML exploits — and ways to protect against them.
Trojanned model: This attack involves malicious models that can execute arbitrary code. These fanged models are camouflaged beside the friendlier pre-trained, off-the-shelf ones available in so-called model zoos.
Infected model: Here, malicious code is added to an existing model.
Data poisoning: The steady flows of data that ML models rely on to operate could be tainted, poisoning the well of decisions they make.
There are many more examples in the 48-page report, which also features ways to mitigate ML sabotage, such as keeping a close eye on data supply-chains and segregating infrastructure used to train models from the rest of a network.
The Department of Homeland Security believes that the threats from a failure to patch or mitigate the Log4j vulnerability could persist for years, it warned late Thursday.
DARPA plans to get more heavily involved with open-source security after realizing how broad a role that type of software plays in the modern tech ecosystem, according to MIT Tech Review.
They created Digital People. Now they've made celebrities available as Digital Twins: Soul Machines is at the cutting edge of AGI research with its unique Digital Brain, based on the latest neuroscience and developmental psychology research.
Thanks for reading — see you Monday!
This newsletter was updated to clarify a quote from new Red Hat CEO Matt Hicks.