Meta has an algorithm problem – just not the one it thinks it has

MENLO Park must have the highest density of millionaire malcontents in the world. Months after Meta went on a talent-poaching romp through Silicon Valley for artificial intelligence (AI) hires, it now has an elite superintelligence unit that is causing status divisions and a jockeying for resources at the tech firm.

The details, reported by The Wall Street Journal (WSJ) last week, appear to be the stuff of organisational behaviour nightmares. Meta’s AI dream team, TBD Lab, is considered so essential that its members sit near chief executive officer Mark Zuckerberg in a restricted area that needs special badge access.

Unlike the rest of Meta’s payroll, the lab’s members are not even named on the internal company organisation chart, like a circle of high-tax-bracket, electric-vehicle-driving Lord Voldemorts.

Meta, for its part, has called the WSJ’s reporting on the matter “navel-gazing”, saying: “This is yet another series of false, exaggerated or mischaracterised claims.”

If you need an idea of how highly paid Meta’s new hires might be, the individual pay package numbers being bandied about in the industry are as high as US$300 million over four years. At last count, Meta is said to have recruited more than 50 people to turbocharge its AI ambitions, at least 21 of which were from rival firm OpenAI.

I’m not sure what Meta thought would happen when it decided to introduce a much-better-paid invasive species to the existing ecosystem, but longer-serving Meta employees are now lobbying for better compensation or coveted spots in its AI unit, according to WSJ. One employee was given a grant worth millions but left anyway, under the belief that the new hires were making multiples more.

You may scoff at this, but there is something to the adage that comparison is the thief of joy – the nominal sum isn’t the point of contention; the disparity is.

Even for some of Meta’s new employees, the larger pay cheque wasn’t enough to make them stay. Tech publication Wired reported in August that at least three AI researchers had resigned, two of whom rejoined their old employer, OpenAI, after less than a month at Meta. The third, a Toronto-based AI researcher who joined Meta in April, was offered a spot at its superintelligence lab, but left after being told that he would have to work on-site at the tech giant’s Menlo Park headquarters in California.

It’s impossible to tell from the cheap seats if the company knows what it’s doing. There is, however, a grim irony in the idea that the very technology supposed to make human labour cheaper has instead given us some of the most expensive humans in history.

But if Meta has calculated the long-term payoff correctly, its wage bill will seem like pocket change. The stakes for AI are astronomical – whichever firm dominates this field will shape the operating system of daily life itself.

Alas, the road to a machine-dominated future still has to be paved by humans, with all their irrationalities and complexity. If Meta wants to retain its most talented people, it will need to re-learn some fundamental management – nay, human – truths: status matters, inequity enrages people, and money isn’t everything.

Currently, it seems to believe that its biggest issue is cracking the code of machine intelligence. But to succeed at that, it has to decode its own people first.