India
Local engineers are part of projects that help avoid food wastage and aid in increasing crop yields
At a time when IBM is going through its most remarkable transformation, the Big Blue’s India research labs are playing a ‘significant role’ for its massive bets on artificial intelligence, blockchain and quantum computing, said Dr. Michael Karasick, VP, global labs, IBM Research. Recognised as ‘IBM Master Inventor’, he said the applications of innovation coming out of its India labs range from helping farmers to increase crop yields and solving food spoilage to even detecting gender biases in the Bollywood industry. Excerpts from an interview with Dr. Karasick and Sriram Raghavan, VP, IBM Research and CTO, IBM India:
What kind of bets is the company making?
Karasick: We bet the company on artificial intelligence, but in an interesting way. We had the same realisation we had many years ago when people started using the Internet… that it wasn’t a consumer thing, but a business tool. [It’s the] same [with] artificial intelligence, it’s a business tool and the value of AI is seen when it is applied in specific domains. One of the things that we’re seeing as [with] the Internet, it doesn’t matter what the industry is, people are reinventing how they work using AI.. it’s a massive reboot, redo of the entire company.
Raghavan: [Another bet is Blockchain]… what [blockchain] promises is trust in business transaction and trust is a big deal. So, very much like AI, we believe blockchain is not limited to one industry or use case.Here is a technology that’s going to change how you bring trust into business network transactions, it applies across many geographies. So our focus has been…on recognizing that the core of this technology has to be built in the open because, you have to have commonality at that level for this to scale and the innovation to continue. And then based on what is… being built in the community, bring it back and ask how do we help our clients absorb this technology and consume it in the most effective way possible. You’re seeing the work we’re doing with Maersk through a joint venture around transforming shipping or the work we are doing in food [industry] by bringing in companies [Cargill, Walmart] across the food supply chain to solve an industry-wide problem of food spoilage.
What role is India playing in these bets?
Karasick: From a research and innovation point of view, the lab in India has got a significant role in the blockchain work worldwide and as well probably one of the larger contributors to the AI effort too. The team here has been focused quite a lot on new technologies [like] blockchain technologies. We talked about machine learning quite a lot in AI, the team here is also focused on something we call ‘machine-reasoning.’ So we are focusing on the kind of AI that businesses need in order to make decisions and a big part of the innovations are coming from here.
Raghavan: There are new elements about AI systems that… the lab is working on. What are the technologies and tools that allow us build these secure, bias-free ethical AI systems? Could I [also] use technology to understand social bias? We intuitively understood [growing up in India] that there was gender bias in Bollywood. But I can actually quantify the bias to you, I can tell you that in X-percentage of movies, while the poster had the female role, that the actual role was only the 25th most important role in the movie. The bigger thought is there is work to be done to build secure bias-free AI systems. The technology could also help you understand social and systemic bias and the Bollywood is a great example of the type of work that was done. There is no way you as a human can pore through the last 30 years of Bollywood scripts and every poster and draw a quantitative conclusion.
How are you applying AI in areas such as agriculture?
Raghavan: Technologies which require deep instrumentation in the soil… to put sensors on the ground to deliver value, the cost, the scaling model… doesn’t work. So could we use AI by combining information from multiple satellites and is there a way to get a level of insight that is almost comparable or as good as putting sensors. It’s actually possible. Combining data from… different satellites, we can predict with 95%-plus accuracy the moisture content half a metre below the soil at the level at which the plant gets its water from. You can measure vegetation health [from the sky] but the trick is to use multiple data sources and you combine them with AI. Now I can give insights for a better reduction of raw materials like water.
I can provide disease prediction models… all without depending on deep instrumentation in the soil which is not possible in a country like India. We have [now] completed three pilots in India [for like farmer advisory, pest and disease prediction and yield estimation].
Researchers at MIT-IBM Watson AI Lab want to make computers more humans. Could you explain how?
Karasick: One of the areas that they’re (MIT-IBM researchers) interested is in how to make the user interface ‘more human.’ It means doing speech recognition and understanding [emotions:] stress, happy, sad, frustrated… I can look at your facial expressions [and] analyse them in the same way that I’ve analysed tone.
We can analyse the language people use, the way in which it is used, facial expressions and give the systems the same visual cues that you and I use to interact.
There is concern that AI could replicate human prejudices on gender and race.
Karasick: We’re engaged with a lot of the large [companies like] Google, Microsoft, Amazon, universities and non-profits [through partnerships] to investigate and understand ethical use of AI systems. There is no such thing as a biased or non-biased AI system.
There are biased and unbiased datasets. The other part is about deploying machine models with guardrails, to constrain what they’re allowed to learn and what they’re not allowed to learn. Because the systems learn as they’re deployed.
There is a whole research area that we’re very involved in called ‘adversarial machine learning’ — meaning, these machine models can be fooled, can be hacked. So, like you have ethical hackers, we have a team that is learning how to fool machine models — both ours and others — and understanding from a scientific point of view, what the defences against [it] are.
What are your views about the impact of AI on jobs?
Mr. Karasick: Anytime there is a technology change, there is a disruption. And the important thing is to understand that there would be disruption and plan forward to decide what to do. Henry Ford (founder of Ford Motor Company) was quoted as saying if ‘I’d ask my customers what they wanted me to build it would have been a faster horse,’ only that he build cars. The apocryphal story is buggy whip manufacturers, probably went out of business, but they did something else. So, there is always disruption that’s the nature of the beast. Some jobs will be lost many, many more jobs will be gained. Going back to Henry Ford, it was the beginning of mass production and an entirely new segment of the economy was created by somebody who wasn’t afraid of the problem. I think the same thing will happen here.
