A Talk by Dr Anupam Guha on ‘What is to be done? Critique of and alternatives to extractive imagination of AI’ was organized by the Centre for Policy Studies, IIT Bombay on 11 January 2019.
Anupam Guha is a computer scientist, working on Artificial Intelligence and Computational Linguistics based in the US east coast with a PhD from the University of Maryland (batch of 2017) and an MS from Georgia Tech (batch of 2010). He is deeply interested in and does outreach on the future Social, Political, and Economic implications of AI, especially both its oppressive and emancipatory potentials for labour.
I was fascinated by the title itself when I first read it. I have seen many people explaining artificial intelligence in past. Most of them have glorified the concept but Dr Guha had something serious to share.
He initiated discussions with ancient Chinese Room thought process which was first coined by American philosopher John Searle in 1980. It says that no matter how intelligently you code computers (or machines), they cannot replicate consciousness, mind and understanding of human being. Inline with this Dr Guha shown different feathers of the subject matter. Following are notable statements made by Dr. Guha throughout the session;
- We are all the time acting as labour to some companies than a consumer. We think that we are consumers to the tech companies but this is not the fact. Take an example of Alexa by Amazon, every time you instruct it something, your data is stored somewhere on Alexa’s server and processed to improve its future behaviour.
- Being an AI scientist, there are limitations to our understanding of mathematical modelling of transparency when dealing with machines.
- Using AI in civil surveillance by cops is objectionable as you cannot put someone randomly in lockup just because a computer algorithm is suggesting so. Whom you count accountable for initiating action? Is the computer algorithm 100% error-free?
- Earlier technologies developed during industrialization and post-industrialization improved work. Artificial intelligence is replicating work. This is the biggest threat to existing jobs.
- Information Technology shifted job flow to global south (boosting service industry). Artificial intelligence will reverse job flow to the global north.
- An output of AI machine is not a problem as far as racist words are concerned because we the human beings are giving wrong input data through our day to day instructions which involve racist words in a different flavour.
- No matter how much you improve the AI algorithm, humans are wanted because they are the one train the AI.
The statements reflect a cautious side of the implementation of artificial intelligence in today’s social structure. These powerful statements make me think if our society is mature enough to train artificial intelligence algorithms.
Note: Opinions expressed in this blog post are personal to the respective author. Dr. Guha has explicitly mentioned at the beginning of the event that his statements should not be interpreted as official viewpoints by any of the institutions to which he is affiliated to.