Hollywood’s AI Hype Train Rolls On—With One Big Skeptic

The Runway AI Summit last week played like a revival meeting for Hollywood’s AI faithful. Speakers framed generative tools as civilization-level shifts—fire, the printing press, the usual suspects—while glossing over the fact that most studios still treat AI as a glorified storyboarding assistant. The only discordant note came from Star Wars producer Kathleen Kennedy, who publicly questioned whether the tech was ready for prime time, let alone mythmaking.

That skepticism arrived at an awkward moment. The summit unfolded just days after the death of Sora, OpenAI’s high-profile text-to-video model, which had been positioned as the next leap forward before its abrupt demise. (Runway, notably, didn’t mention the elephant in the room.) The timing underscored a pattern: AI’s boosters rush to anoint each iteration as transformative, while its critics—like Kennedy—point to the reality gap between flashy demos and what actually ships.

Early signals suggest the summit’s hype cycle followed a familiar script. Runway’s latest tools, per leaked benchmarks, still struggle with consistent character rigging and lighting coherence—problems that VFX supervisors say make them ‘useless for 90% of shots.’ Yet the marketing machine chugs on, repackaging incremental gains as paradigm shifts.

The real story isn’t the tech—it’s the industry map. Runway, backed by Google Ventures and NVIDIA, is positioning itself as the ‘creative OS’ for studios, a play that pressures legacy VFX houses like ILM and Weta Digital. Their pitch: Why pay for artists when you can license our model? But as Kennedy’s comments hint, the math doesn’t add up—yet. GitHub activity around Runway’s tools remains flat, and indie devs note that ‘it’s faster to do it manually’ for complex scenes.

Meanwhile, the community’s irony meters are pegged. The summit’s closing tagline—‘Thank You for Generating With Us!’—became an instant meme among VFX artists, who’ve spent years watching AI demos promise ‘one-click cinematic quality.’ The Blender and Unreal Engine forums are filled with the same refrain: Show us the pipeline, not the sizzle reel.

If there’s a shift here, it’s not in the tech—it’s in the marketing’s growing desperation. After Sora’s collapse, the industry’s AI acolytes are leaning harder on metaphor (‘fire! printing press!’) and softer on deployment data. That’s not how revolutions start. That’s how bubbles inflate.