Has Human Officiating in the NBA jumped the shark?

[QUOTE=ralph_i_el;15045745]Line judge in tennis is incredibly simple compared to reffing basketball, but I get what you're saying.[/QUOTE]

I was reading today what detection tech the Waymo robotaxi has: lidar, radar, ultrasonic sensors (like a bat's echolocation), thermal imaging. It costs $110K per car. You can 'see' in every sense- heat, sound, reflection. Using cameras, radar, laser.

In my earlier post, I point out how the NBA is already using computer vision to evaluate human refs and their mistakes.

According to AI, the tech for AI refs is 5-10 years away. The hawkeye tech from tennis is already there.

"[B]These could be automated right now with near-perfect accuracy:[/B]

Out-of-bounds calls (computer vision is excellent at line detection)

Goaltending / basket interference

Shot clock & game clock violations

Foot on the line (2 vs 3 points)

Defensive 3-seconds

Backcourt violations

Whether a ball was blocked or tipped

[B]2. What’s close but not perfect yet (~3–5 years)?[/B]

These require better body-tracking, collision mapping, and intent estimation:

Off-ball contact (pushes, grabs)

Charge/block decisions where players collide at angles

Travel detection

Illegal screens

AI can track bodies extremely well today, but the NBA standard would require:

millisecond-level joint tracking

reliable force estimation

context understanding (e.g., “was this enough to be a foul?”)

Readiness level: ~70%.
Expected maturity: 3–5 years.

[B]✅ 3. What’s the hardest and furthest away (~5–10+ years)?[/B]

Calls requiring intent, degree, and context, like:

Whether contact “affected the play”

Flopping vs legitimate reaction

Whether a swipe was on the ball or on the arm

“Marginal vs illegal” contact

Continuation calls ("and-one" judgment)

Flagrant-hood (reckless vs unnecessary vs excessive)

These require semantic judgment, not just detection.

Today’s AI is improving fast at physical reasoning, but to reach NBA consistency:

needs highly accurate force estimation from visual data

needs deep behavior modeling of players

needs to understand basketball context, not just physics

Readiness level: ~40%.

Expected maturity: 5–10 years for raw technology to be good enough."