Now we are into the age where the Data Science world is incomplete without AI and ML. We are now crunching data more than before. Consumer-driven data is making a difference to every organization no matter big or small. More and more advancement is rapidly being made in the tech world, and this has led us to give an effort to paint a picture of the data science landscape in 2020. With the advancement in technology, everything can be achieved by just sitting in the comfort of your bedrooms.
AI and ML Landscape
AI is quite a vast topic and consists of many arms, such as ML, Computer Vision, etc. According to the state of AI, a site which provides present market trends for AI has made predictions for 2020 which are:-
All you have to do is to open them and search for your requirement. Your product will be delivered to your doorstep in no time at all.
- Capital funding for startups going for a breakthrough in NLP would be $ 100 million-plus.
- The self-driving concept is still confined to the R&D labs, and so far the organizations working in this field have driven 15M miles in 2019.
- Institutions are coming forward to establish an undergraduate course-curriculum and are laying the groundwork to fill in the talent void in the field of AI and ML.
- Google’s work in quantum computing has led to the formation of 5 startups working in quantum machine learning.
- Governance issues related to AI technologies are the most prominent challenging topic now, given to the situation that AI-driven technologies are becoming dominant. This has led to one or the other organizations to bring in the change to their governance model.
You can read more about this here.
Data security concerns and privacy issues owing to the several scandals unearthed in the year 2018 led the shift towards rethinking about governance model surrounding power AI tools and organizations. One such initiative could be seen at Stanford where Institute for human Centered AI has been opened which would deal with unforeseen and unexpected challenges and problems AI brings to the industry and to the lives of the people.
Data security and privacy-related issues have been at the forefront of 2019. This issue led to GDPR, which came into effect in 2018. By Jan 2020 California Consumer Privacy Act (CCPA) would go into effect. In India, New Delhi airport will start rolling out facial recognition for the passengers whereas San Francisco voted to ban facial recognition technology. These are some news which has been making headlines in the tech world. 2019 could be seen as the initiation for Privacy and data security being commoditized for the commercial market, and organizations would be heavily invested in developing technologies and products surrounding the data privacy and security.
According to Matt Truck, an investor at FirstCap and organizer for datadrivenNYC data governance, cataloging, data lineage, and data management are becoming more important than ever before. The phenomenal rise will be seen in the AI-driven Stack.
AI and ML-powered cloud is another trend that has been another hot subject. Many organizations have been working on offering AI and ML-powered cloud solutions which would help in automating, manage, and decide for cloud-related products to be used by cloud users. A glimpse into google’s beta offering could be seen here.
AIOps and MLOps pipeline would require massive AI-driven stack which would also require rare skills which at present is not sufficient thus this will force the organizations to fund more for either training or developing solutions centered around autoML and other technologies where manual effort would be the bare minimum.
According to the Allen Institute, China has been publishing more ML-based research papers as compared to the US and has been able to fill in the quality gap.
Like electricity changed our lives in every possible way, so will the AI. According to Prof Andrew Ng, AI is the new electricity. As the AI and ML trend have snowballed, the market is still void of skilled resources. This brings us to discuss what should be the AI and ML learning roadmap for 2019.
AI and ML Roadmap 2019
For people aspiring to work on AI and ML related tech-space, this roadmap should be seen as guidance to build a career in this field. One can always follow this roadmap and start to work on the skills required. To help you, there are infographics, which is nothing but the roadmap.
Google and Apple both provide their app developers with their own platform specific tools. Because of this, the app provides with an optimized and enhanced performance to the user since it is developed in accordance with the guidelines and frame the company provides.
- Knowledge of Python Programming
- Understanding of Probability and statistics
- Knowledge of Matrices and Vectors
- Knowledge of Linear algebra and calculus
AI can be divided into two parts for learning and gaining skills.
- Machine Learning
- Deep Learning
Begin with Machine Learning first followed by Deep Learning. The most popular course on ML is by Prof Andrew NG, and it is available on youtube. Machine learning is further subdivided into three parts.
- Supervised ML
- Unsupervised ML
- Reinforcement ML.
Supervised ML contains classification and Regression algorithms for classification and regression related problems. The first step begins with understanding the classification and regression problems followed by the algorithms for solving such problems. Some of the algorithm examples are logistic regression, K-nearest neighbors, decision tree, naive Bayes classifier, Bayesian network, linear regression, neural network regression, etc.
Unsupervised ML contains clustering and association problems and algorithms for solving such problems. In this case, the system is exposed to large sets of varying data so that it could learn by itself. Some of the algorithm examples are soft clustering, hard clustering, etc.
You can learn about supervised vs. unsupervised ML here.
Reinforcement learning is being seen as the hope of AI and is more focused on the system’s interaction with the environment. It deals with problems which demand ‘sequential decision making.
Deep learning is subdivided into three parts:
- Supervised DL (discriminative)
- Hybrid DL
- Unsupervised DL (Generative)
Some Deep learning algorithm examples are auto-encoder, recurrent neural network, etc.
The roles and responsibilities of ML engineers in the future may vary to a greater degree; however, some of the responsibilities are worth mentioning. ML engineer’s responsibilities could be envisaged as follows:
- The ML engineer will bear the responsibility to create and maintain ML solutions.
- He will help solve problems like privacy and data security, digital marketing campaigns, etc.,
- He will continuously be evolving himself by keeping enlightened with the developments in the field of AI and ML research and advancement.
- Will contribute to the research and development of AI and ML tech-space.
- Will look for possible solutions to scale and optimize the performance of AI and ML induced solutions running on the infrastructure.
We have seen how the AI and ML landscape is providing a glimpse at the tipping point in the industry. The industry is very volatile, and there is nothing that has been spared by AI and ML. The industry is facing skill shortages, and every day, new technologies are being built around AI and ML. The roadmap discussed in the article presents a guiding tool for the aspiring ML and AI engineers to pursue skills.