Presenter: Gargi Dasgupta

Moderator: Jibu Elias

How Enterprise AI Can equip business for future success

First let’s explore what enterprise AI really means. We all have heard about AI taking over business, change the way we live, interact and more. AI is an important tool for enterprises to succeed and there are different technologies that can help enterprises succeed. IBM research is primarily tasked with innovated on technologies that matter. In India, we are working on the transformation of cloud and AI; in IBM Research, we tackle the hard problems worth solving now. AI has applications across supply chains, conversational aspects of business.

With the cloud, there is security and is a modernised offering. This is pivoted on helping our clients move to the cloud. Aggressive adoption of cloud technology will help navigate tough times like COVID19. Whatever your infrastructure may be, it has to be modernised, managed and focused on security. One of the major drivers of AI is data. ML and AI techniques that can establish trust and deliver explainability via data is core. Through various language media, made possible through documents, emails, webinars like this one, a far better, stronger understanding AI can be developed and brought ahead. For better reasoning capabilities, state of the art tools can carry out better research. In addition to working on providing leading infrastructure for the cloud, IBM is focused on technologies to understand weather patterns, agriculture with the goal of augmenting the food ecosystem. We have strong partnerships with a range of academic institutions like IITs and IISc, in addition all our business units keep us grounded on client needs.

Sitting at home, and working from home now, the whole world is responding to a global pandemic. While these are grim times, there are some positive changes in businesses like healthcare, shopping and social interactions. Moreover, there is a definite change in the way enterprises are adapting now, with a direct impact on their agility and growth. There is statistical proof that hybrid/multi cloud and AI will help enterprises navigate more efficiently than ones who do not. Two dominant forces driving transformation is the hybrid cloud and AI. Right now, companies want to understand how they can use AI to accelerate the journey to the cloud.

The utility of cloud is largely centered around operational efficiency but there’s more. Unless companies adopt the latest in cloud, they cannot innovate. The trend has to be less about outsourcing to a technology company but partnering with a tech company so that they can drive revenue growth. For example - a retail outlet would now want to partner with technology companies to push their sales, understand their customers better and create a better instore experience. There is definitely a need for such solutions and there are more digital leaders thinking this way across industries.

Hybrid Multi Cloud Is The Ultimate Power of Choice In The Playbook of Digital Transformation

Right now cloud adoption is largely is three categories – public, private and hybrid. Most customers ask if they are ready for public cloud? This involves one’s health, personal data existing on a public platform. On the other hand, there’s the private cloud which exists is a secure environment. However, 94% customers around the world are exploring hybrid cloud solutions – this is a combination of data spread across private, public, and the rest on data centers. About 60-70% of vendors say they manage multiple clouds now and that they will coexist. Be it a Google, IBM or Microsoft, certain services will be better on one platform over the other. Hybrid multi cloud is a power of choice. The commonly asked question is the challenge of migration but if applications have been developed with the right sanitation measures, it can be deployed anywhere. This is a key trend on the cloud. In addition, it can can also be managed securely being an open structure.

Initially, we were good at solving narrow AI problems like image recognition and bringing down speech error rates. While we proved a lot of our tech in the narrow AI domain, we are now moving into broad AI applications. This is very important for enterprises. As a community, if this move from narrow to broad isn’t made, enterprises cannot jump on this bandwagon of AI as they have very less labelled data. On the contrary, they have requirements to be as good in AI as a human is. By moving broader with tech like transfer learning, learning from limited data and infusing AI into business processes, the model needs to be trusted and explainable. For instance, in the case of banking, a machine could very well be looking into my loan application, gauging my worth and making a decision like a human. Hence, the model needs to be fair, robust, trusted and explainable. This brings in new use cases. If I want to use AI, its less likely that a sophisticated RNN model would be developed and deployed and then never followed up with. Developers have to continuously learn from the model, drift of the data, and build on the insights to be able to detect bias, fairness, and robustness. The entire lifecycle of AI has to be studied. This is where we are at in 2020 – there are a range of broad problems with more experts focusing on it. While general AI capabilities (which is human intelligence) and being able to do human enabled neural tasks as AI is still a longer shot, broader AI is a definite focus area now.

'Code is becoming the ubiquitous language of machines'

Code refers to all deployment artifacts, configuration files as well as deployment tools like CICD, logs to manage applications and the broader view of code. We are doing really well on speech performance, voice recognition, sentence similarity/contradiction, conversational AI, document understanding. Now, we want to apply the same metric of success to code similarity, code completion, code translation between languages and this is where it all comes together with cloud. There are 26mn developers around the world. India has 5.5 mn - around 1/4th of the world developers are in India. We need to grow them, enable them with AI tech. In 2011, Mark Anderson in Wall Street Journal wrote a piece on why software is eating the world – this article touches upon the point that every company needs to be a software company. IT is not restricted only to technology companies. It alludes the success of companies like Netflix over Blockbuster, Amazon over Barnes & Noble. Software development will be unprecedented, developers will most important persona and we need to accelerate AI for them to stay ahead and continue making the best software.

There are parallels between human and machine languages. While human languages are English, German, Spanish etc; for machines its Java, Python, COBOL. Vocabulary includes nouns, pronouns and variables, while for machines, its functions and classes. Both have their own syntax and language construction models as well. The tech used to master human languages will be used to master code as well. It could mean configurations, deployment files, log files; it could also mean slack conversations. NLP for software artifacts will be a dataset of laws, metrics, looking through hundreds and thousands of configuration. It will involve applying automated reasoning and decision making, and explain how this helps business.

Acceleration IT transformation via AI

Advise - Move - Build -Manage – this is under AI for Hybrid Cloud analysing code and a very high demand area for scientific research

• AI for application modernisation
• AI for Developer Experience
• AI for Security
• AI for Compliance
• AI for IT Operations

AI is changing how businesses operate today and has a wide and far-reaching impact on customer service, risk compliance, IT operations, business operations, and financial management. AI acts as a team member at every step of the way of incident management, and help flag issue detection with predictive abilities.

Currently, the areas that enterprises seek expertise include:

Artifact gathering to narrow down critical alerts and address main problem areas

Diagnosis and triaging of an issue and provide recommendations

Resolution & summarisation of the problem and the takeaways from these challenges to be conveyed to the team. This is a goldmine for conversational AI enthusiasts