The three questions you need to ask of technology to understand its impact.
As the Director of the Accenture Labs in Bangalore, Shubhashis Sengupta is used to transforming powerful ideas into real-world solutions, from research, ventures, labs, and studios to innovation centers and delivery centers. His experience working closely with AI is helping reimagine and reinvent the future for our clients.
With twelve years of experience, Shubhashis has always been a fan of AI within Accenture Labs, But the new trend strikes differently, as new models have brought about a new paradigm shift to creativity.
“What we’ve seen happening in the last three to four years in the field of AI is that we have developed the technology that enabled AI to predict what comes next,” he said. “These are essentially called generative models because they generate predictions. They are able to generate the most appropriate words or phrases that come under a passage. And of course, then came the photos and the images, and we will see the intuitive use of AI to help you to do many things. It can draw a picture for you, it can summarize this article into 100 words, and can turn it into a song. But I wonder, as a possible next step, once we turn this technology into common day-to-day applications, what could it mean for society?”
As an expert, Shubhashis thinks that, with new technology, we have to ask ourselves three questions: Is it going to do things better, faster, or cheaper? Is it going to do new things? And, is it going to create a disruption while doing so?
“We can already see AI models passing examinations, answering questions on jeopardy, and giving lectures,” he said. “These are knowledge presentations and AI models can do this better than I can, but the key question to address people’s concerns is: can it go from being a productivity tool to being a creative tool? I can instruct AI models to paint like Van Gogh, but it can’t be Van Gogh — it can’t be an artist because it can’t use imagination. One thing is deductive reasoning, and another completely different is abstract reasoning. Combining them would result in symbiotic programming, which combines the data part with the sixth sense to make sense of it, but we are not there yet!”
Harm alongside good
Shubashshi believes that for all the positives it can have, harm can be done too. For example, a lot is being used for creating medical image analysis. For example, in cancer and cancer cells when you don’t have enough data baseline to work with, people can get a few images and ask AI to create integration so that they can potentially know all types of cell formulations in that particular operative case. But they can also use this technology to make up someone who says any type of weird and extreme thing, put it on Twitter and the damage is done.
One reason for this is that people still don’t have regulations addressing some of the most basic questions about the Internet. The things lying around on the Internet don’t necessarily mean that people could go and grab and use them because when Creative Commons licences were created, they were actually intended for something else. Because of this, you can have a look at how AI models are actually being developed and decide whether to use them in a good way or a bad way.
Another emerging issue with generative AI is that it can never be 100% factual. “Imagine if I’m creating reports for my manager,” he said. “I cannot use an open-source generative tool because it might actually create something which makes no sense. The challenge is that most languages and concepts have layers of meaning and interpretations that are really hard to fully understand for a generative AI model, and I don’t think AI will ever come to understand them. It’s almost like reaching a higher level of consciousness when the creators will get to connect words with meaning and understanding of context and intention, which is how humans have developed such a complex language. Right now, we are working on AI trying to mimic something without understanding what it is.”
All of this ties back to the central point: Is it going to do things better, faster, and cheaper? “Yes,” he exclaimed. “With some set of applications and some sort of use cases, I think we can do better, faster, and cheaper, but not immediately.”
Is it going to do new things? Things we thought were not possible. It might in the future, but it will be a much more advanced version of what we know today. And, is it going to create a disruption while doing so? Once we manage to develop symbiotic programming, it most likely will.
Follow Shubhashis on Linkedin.
