Skip to main content
. 2024 Feb 21;10:e51523. doi: 10.2196/51523

Table 2.

Number of correct responses for each topic per model.

Topic GPTa-4, n (%) GPT-3.5, n (%) Bard, n (%)
Biotechnology (n=11) 7 (64)b 6 (55) 4 (36)
Evolution and health (n=9) 7 (78)b 4 (44) 2 (22)
Population and ecology (n=6) 1 (17)b 0 (0) 1 (17)b
Biomolecules (n=3) 1 (33) 1 (33) 1 (33)
Cell biology and genetics (n=16) 7 (44)b 7 (44)b 3 (19)
Ecosystem and environmental issues (n=5) 2 (40) 2 (40) 3 (60)b
Plant kingdom (n=25) 8 (32) 6 (24) 11 (44)b
Animal kingdom (n=24) 17 (71)b 8 (33) 6 (25)
Physical chemistry (n=12) 6 (50)b 3 (25) 4 (33)
Organic chemistry (n=9) 3 (33)b 1 (11) 2 (22)
Inorganic chemistry (n=15) 7 (47) 8 (53)b 6 (40)
Mechanics (n=12) 8 (67)b 6 (50) 6 (50)
Heat and thermodynamics (n=3) 1 (33) 1 (33) 1 (33)
Electrostatics and electricity (n=11) 10 (91)b 5 (45) 6 (55)
Optics (n=3) 3 (100)b 2 (67) 0 (0)
Simple harmonic motion and waves (n=1) 1 (100)b 0 (0) 1 (100)b
Magnetism (n=6) 4 (67)b 2 (33) 1 (17)
Modern physics and electronics (n=4) 2 (50)b 2 (50)b 0 (0)

a GPT: Generative Pre-trained Transformers.

bHighest accuracy within a topic.