Kan AI slå trænede mennesker i læbebevægelseslæsning ?
Afgiv din stemme — læs så hvad vores redaktør og AI-modellerne fandt.
DeepMind viste dette i 2022 med en transformer-baseret model, der overgik professionelle læbere på TV-nyhedsindslag.
Background
Researchers have made significant progress in developing artificial intelligence systems that can lip-read, with some studies demonstrating that AI models can outperform trained human lip-readers in certain conditions. These AI systems use computer vision and machine learning algorithms to analyze the movements of a person's lips and identify the corresponding speech sounds. While the accuracy of AI lip-reading systems can vary depending on factors such as the quality of the video input and the complexity of the speech, they have shown promising results in various experiments. Overall, the current state of the art in AI lip-reading suggests that these systems can indeed beat trained humans in certain scenarios.
— Enriched May 9, 2026 · Source: University of Oxford
Foreslå et tag
Mangler et begreb i dette emne? Foreslå det, admin gennemgår.
Status senest tjekket July 2, 2026.
Galleri
Kan AI slå trænede mennesker i læbebevægelseslæsning?
Snævre demoer findes — men panelet var ikke enigt.
Efter livlig debat fandt juryen AI’s læbeaflæsningsevner prisværdige, men endnu ikke retscertificerede til reel flydende anvendelse, idet de konstaterede snævre sejre på benchmarks samt vedvarende skrøbelighed i den virkelige verden. Den ene “ja”-juror pegede på benchmark-sejre, mens de “næsten”-skeptikere understregede ujævn præstation uden for laboratorieperfekte forhold. Kendelse: “Stumfilm er ikke længere stumme, men undertekster har stadig deres tid.”
After spirited debate, the jury found AI’s lip-reading skills commendable but not yet court-certified for real-world fluency, spotting narrow victories on benchmarks alongside persistent brittleness in the wild. The lone “yes” juror pointed to benchmark triumphs, while the “almost” holdouts stressed uneven performance outside lab-perfect conditions. Ruling: “Silent films are no longer silent, but subtitles still have their day.”
But the data is real.
The Case File
Across 12 sessions, 36 jurors have heard this case. Combined tally: 18 YES · 16 ALMOST · 2 NO · 0 IN RESEARCH.
Note: cumulative includes older juror opinions. The current session tally above is the live verdict.
By a vote of 1 — 2 — 0, the panel returns a verdict of NæSTEN, with verdict confidence of 87%. The court so orders. Verdict downgraded from prior session.
"AI lip-reading models exist but accuracy varies"
"Lip-reading AI systems like AVSpeech and LipNet have surpassed human performance on benchmark datasets."
"AI lip-reading demos exist but are narrow"
Individuelle nævningers udtalelser vises på originalengelsk for at bevare bevismæssig præcision.
Hvad publikum mener
Nej 6% · Ja 75% · Måske 19% 150 votesDiskussion
no comments⚖ 12 jury checks · seneste for 2 dage siden
Hver række er et separat jurytjek. Nævninger er AI-modeller (identiteter holdt neutrale med vilje). Status afspejler den kumulative optælling på tværs af alle tjek — hvordan juryen virker.