Now Reading
Interview - Roger Schank, Leading Artificial Intelligence Theorist


Interview - Roger Schank, Leading Artificial Intelligence Theorist


Roger SchankRoger Schank, the name is synonymous with Artificial Intelligence. Dr. Schank is one of the world’s leading researchers in the space of Artificial Intelligence. He has spent nearly 50 years working in the field of artificial intelligence and his contributions to this field have been immense. An artificial intelligence theorist, a cognitive psychologist, a scientist, an educational reformer, and an entrepreneur is what portrays Dr. Schank.

He is a fellow of the AAAI, the founder of the Cognitive Science Society, and co-founder of the Journal of Cognitive Science. Beginning in the late 1960s, he pioneered conceptual dependency theory and case-based reasoning, both of which challenged cognitivist views of memory and reasoning.



Analytics India Magazine (AIM) recently did an Interview with Roger Schank to understand his views on Artificial Intelligence. Here is the excerpt from the interview.

[dropcap size="2"]AIM[/dropcap]Analytics India Magazine: Tell us something about your current work and your achievements in the AI space

[dropcap size="2"]RS[/dropcap]Roger Schank: I have been in the field of AI since 1965. Over these years I have watched AI go through various phases. Today all kinds of things that have nothing to do the modelling of human intelligence are called AI. I like to give an example to my audiences. I ask my audience to name 50 states of US. Most Americans would say they know the 50 states. I ask them to write the names down and the chance is usually 1 out of 100 can name all 50 states.

This shows us something about how human memory works. People have to visualize a map and go down the east coast and up the west coast in order to list all states. They know the names of all them, but there is a big difference between recognition and recall memory. Modern day “intelligent” machines don’t worry about that distinction. If you ask any AI machine it will easily list all 50 states (in alphabetical order) because their algorithms are all about words and their memories do not try to recall or visualize or imagine the way people do.

People do not memorize data and then look that data up when asked. AI, to me at least, is about modelling human intelligence, not effective data processing.

AIM: Though there is rise in AI, we are yet to reach a stage where a truly intelligent machine exists. How far are we from that?

RS: AI has lot of subfields under it. For instance: face recognition. Facebook uses face recognition to recognize your picture. This is an example of Artificial Intelligence. Robotics is another area in AI which is growing. But the part of AI that interests me is building a machine that behaved very similarly to humans, i.e. one that engage in a meaningful conversation.

For instance, if I am travelling to Denmark, I will think about what my last visit there was like. I will go down memory lane, think about what food I ate, who I met, what I liked, and then, based on my past experiences make decisions for my current trip. Humans use past experiences to make decisions about the present. But machines cannot look up for past experiences because they haven’t had any and because AI people have not studied well enough how prior memories are indexed and retrieved. Modern AI’s are fed with data and they can only look up that data for answers. They do not have experiences or memories to think about or to draw inferences and new plans from.

AIM: Your views on machine learning being used in AI

RS: Machine Learning and Deep Learning are words that are very fancy but really about human learning. As humans we learn from experience. For instance, I just gave a speech in Spain and I learned lot from that experience, I learned from the interaction with the audience. “Deep Learners” can’t do that kind of learning.

See Also
Sudharshan Ravichandiran, Data Scientist at Param.ai

Machines can do lot of things but are the efforts going in the right direction? There has always been an argument in AI between those who want machines to do things fast and beat humans at games, and those who care about the replication of human abilities and processes.

AIM: Do we see AI winter anytime soon now?

RS: AI winter will come soon, maybe next year. The companies investing in AI now are all venture capitals and with high expectations about AI projects, and they will be disappointed for the most part.

AIM: How do you see this era of AI as compared to the earlier times?

[pullquote align="right"]Real AI machines would come up with novel opinions and be prepared to discuss and debate. When we see that happening we will be seeing AI. [/pullquote]RS: There were always people in AI who were concerned with building experiential memories that changed with each new experience. For instance, if I ask you which school is good to attend you will rely upon your past experience in order to make suggestions. Machines today could search college rankings. They are fed with set of data and rules and they search and give answer. Humans don’t search for answers. Either they have a point of view or they try to use their past experiences to come to some conclusions. As humans we imagine things and based on our imagination we make decisions. But today’s machines search text from the data they are fed and really have no idea what that data is even about.  Humans have opinions of their own. Real AI machines would come up with novel opinions and be prepared to discuss and debate. When we see that happening we will be seeing AI. Right now we are seeing search on massive amounts of text, which is something cannot do and wouldn’t try to do.



Register for our upcoming events:


Enjoyed this story? Join our Telegram group. And be part of an engaging community.

Provide your comments below

comments

What's Your Reaction?
Excited
0
Happy
0
In Love
0
Not Sure
0
Silly
0
View Comment (1)
  • Absolutely agree with Dr. Schank’s viewpoints. Most AI startups today are merely providing Chatbot type products or purchase assistants. These are more Predictive analytics than AI. Agreed that the analytics gets better with usage as the number of data points and use cases increase and the search trees get better weighing factors to reach better predictions faster but these are not representative of Artificial Intelligence. Having evaluated at least 3 such startups for funding over the past 3 months, they are all more similar than different and none i can say is AI. You need to first understand how the brain makes decisions and these are not just happy or sad as in most sentiment analytics programs. The brain makes 2 broad types of decisions, quick/ instinctive or researched/ deliberated. Most learning machines try to understand or mimic the 2nd type. In my opinion, the ones that are more difficult to mimic are the ones that are of the 1st type and we have very few use cases being made by these startups.

Leave a Reply

Scroll To Top