Has Meta done “almost nothing” in AI? The numbers tell a different story

Futurism published an article making a bold claim: Meta allegedly has “almost nothing” to show for its actual AI achievements. The headline is sensational, but if you check it against public metrics, the picture looks different—and this discrepancy is more significant than the journalists’ spat with the corporation itself.

The issue isn’t whether Futurism or Meta is right. The issue is that “AI results” are measured differently depending on who’s looking. For the media, the marker of success is a consumer “wow” product that you can touch and describe in a single paragraph. For engineers and investors, the markers are different: the market share of open-source models, the scale of infrastructure, the number of downloads, and the speed of releases.

Meta is making the Llama family openly available, investing in GPU clusters, and actively publishing research. One can debate the quality of the models and Zuckerberg’s strategy, but to say “almost nothing” is either a deliberate provocation or a misrepresentation of the criteria. When an assessment is reduced to a single consumer product, all the infrastructure and research work falls out of view.

The practical takeaway for those following the industry: don’t trust a single narrative—whether from the media or a corporation. Cross-check bold claims with measurable data: releases, benchmarks, licenses, and real-world usage. That’s where you can see the gap between the hype and what actually changes a developer’s stack.

Source: Hacker News - Newest: ""AI" "LLM""