900,000 reasons for working in AI 🤑

Wanted: AI talent

While much of the commentary around AI has been about how such technologies might decimate jobs, there will, it seems, still be winners.

As The Wall Street Journal (paywall) reports, the crush for AI expertise is pushing salaries to eye-watering levels. Dating app Hinge is looking for a vice president of AI who could earn almost $400,000 a year, while the freelancer platform Upwork is offering close to $440,000 annually for a similar role.

But for those AI practitioners looking to purchase an even bigger yacht, Netflix is offering as much as $900,000 a year for a machine learning product manager.

NINE-HUNDRED-THOUSAND-DOLLARS. A YEAR.

You know what $900,000 can buy you? ChatGPT told me it’s enough for 20,000 pairs of Crocs. Perfect to go with that new boat. Of course, with such sky-high compensations sloshing around, that naturally puts pressure on other companies who simply cannot pony up almost a million dollars a year for an AI expert.

The report quotes Chris Todd, CEO at software business UKG, who notes the demand for AI talent is “smoking hot.” Meanwhile, Paul J. Groce, the head of recruitment firm Leathwaite makes a succinct point about the state of play: “We do not magically have thousands of additional AI developers, product managers, and everything else.”

So, if you’re worried about AI flattening your industry, perhaps now’s the time to learn how to drive a metaphorical AI steamroller. And then a (non-metaphorical) boat, of course.

Why it matters:

Companies don’t pay more for talent than they have to - especially those that are beholden to shareholders. So Netflix’s decision to offer $900,000 for a single AI head is perhaps a sign of how much FOMO all these big operators are feeling. But with AI salaries nearing seven digits, that will surely encourage more people to focus on the field, meaning compensations likely won’t stay so high forever.

DEF CON gets into AI red-teaming

Thousands of hackers showed up in Las Vegas this past weekend, determined to bend machines to their will. But rather than being some elaborate slot machine heist, the hackers were in town for DEF CON.

For the uninitiated, DEF CON is an annual conference that invites hackers from around the globe and tasks them with finding novel vulnerabilities across a whole array of systems and services. This includes products from the likes of Google, Microsoft, Nvidia, and more. Challenges are arranged in “villages,” with each attracting its own group of specialists.

With AI being the buzzword of the moment, it’s perhaps no surprise that this year’s DEF CON featured an AI village. Here, hackers were given specific tasks designed to test the fences of various AI tools - a process known colloquially as “red-teaming.” (Personally, I think “big scary hack attack” is a cooler phrase, but I don’t make the rules.)

The first day of DEF CON saw long lines as hackers eagerly waited their turn to bust some AI systems and make them do naughty things. But, as Axios reports, the results of those efforts were somewhat mixed.

One set of hackers were able to use generative AI to build a nefarious ecommerce site in about the time it takes to boil an egg. But another prompt engineer who got an AI platform to explain how to stalk a person found the system refused such a request the day just a day later. (I guess we just have to hope bad guys are terrible note-takers or have short memory spans.)

Why it matters:

While this was the first AI village at DEF CON, the pace and scale of AI innovation means it’s likely to become a regular feature at the event moving forward. And considering there are AI doomsayers both outside and inside the industry (see: OpenAI’s Sam Alman’s statement about AI possibly going “quite wrong”), the importance of white-hat hackers finding flaws cannot be overstated.