Hand-drawn data viz news roundupfrom @itsmarceliina

Transcript

Look at the data behind this week's headlines. First, Anthropic became the most valuable AI startup in the world. It raised $65,000,000,000 in its series age round, hitting a $965,000,000,000 valuation. That edges out OpenAI, which was last valued at 852,000,000,000. The two companies have been increasingly compared on both valuation and revenue metrics. And on both, Anthropic has recently been closing the gap fast. After this round, it's reported that Anthropic's seven co founders, who each own less than 1% of the company, got about an $8,000,000,000 net worth bump each. That's the insane scale of this AI good rush. Last week, space related stocks have been surging on the back of SpaceX's upcoming IPO. But on May 28, a Blue Origin rocket exploded during a pre launch test at Cape Canaveral. The Jeff Bezos owned company is working to launch Amazon Internet satellites into orbit as a direct competitor to Elon Musk's SpaceX. The next morning, those space stock gains started unwinding. For example, Rocket Lab, which had run up significantly in the week prior, hit an intraday low of down nine percent. Similarly, AST SpaceMobile dropped 20% at its worse, Intuitive Machines tier, and Voyager 12%. And finally, a Stanford study of 4,000,000 job applications found adverse impact in AI hiring. Focusing on the Pymetrics hiring system, they found 15% of Asian applicants and 26 of Black applicants routed to positions where the algorithm discriminates against them under federal law. But a more novel finding is about algorithmic monoculture. When applicants apply to multiple employers using the same model, they expect independent evaluations. But their numerical score is identical everywhere, meaning a rejection at one triggers rejection across all of them. Them. This chart shows exactly that. The observed systemic rejection rank, shown in dark blue, consistently exceeds what chance would predict, shown in light blue. For example, ten percent of people who applied to four jobs were rejected from all four of them, a much higher rate than what would be expected if it were up to chance. The data is from 2018 to 2022, and a lot of today's algorithms work differently. But the concern holds. If a major vendor's algorithm has a flaw, every company using it gets hit, sometimes without even knowing it. And that's all for today. See you next week.