Limitations of artificial intelligence

Limitations of artificial intelligence

Though they are machines, artificial intelligence (AI) can work as per instructions or by programming the data generated over time into their system. However, they have their limitations and they are more human than we can imagine. We are more than often scared of machines and robots taking over the world, nonetheless, they need to be conducted by the human mind and they often rely on their human makers to provide them with information and date to work accordingly. Aria Bracci and Emily Arntsen, two of the Northeastern juniors and reporters of Litmus podcast at the News@Northeastern speak about this matter with Northeastern University’s ground-breaking researchers to “connect what’s going on in their labs to what’s going on in your life”.

 

According to Jordan Peele, an American actor, and comedian, artificial intelligence is dangerous and can dupe us as he warns us to “stay woke”, still every machine needs a teacher and the most important threat to the world of AI, is the real life. Computers are unable to produce information spontaneously as they rely on machine learning called ‘generative adversarial network’ which gains knowledge from absorbing new data. ‘GAN’ is also used to generate fake data or “deep fake” to create malicious and super-realistic videos of people speaking or doing things that never existed in the first place.

 

Sarah Ostadabbas, assistant professor of Electrical and Computer Engineering, used such fake data in her lab to help in the medical field, specifically to prevent bed-ridden patients in the hospital from getting severe bed-sores. Her goal is to train the machine to observe a low-resolution picture of the patient and his/her position and alert the nurse at the right time to readjust the posture to get rid of bed-sores. Sarah termed the false data used in her lab as “synthetic data” and these data generated from the simulator of the situation and fed to the machine to school it. However, they still lack in producing any new information and follow the already generated data. Sarah believes that humans can teach AI to be better and smarter.

 

Rubena Bose

Avatar
Rubena Bose

boserubena14@gmail.com

Away from all of reality.

No Comments

Post a Comment