Introduction to Past:
Thinking is an important piece of the puzzle, time has solved for the evolution of human intelligence. Over centuries of primitive development, we, human’s have mastered the capacity to think, understand and respond better than other competitive species and sub-species, which played a very vital role for our domination as a race. This power of ‘ being able to think & act’ and having the knowledge of our own thought networks helped us to strengthen the roots woven around communication. Indeed, our analytical skills in Learning and Getting optimised, Co-creation and contribution towards a common goal and a smarter world are fruits bore from the effectiveness of our thought networks. A thorough observation of human intelligence provides interesting insights, about the formation of ‘Human Thought Networks’ and ‘Judgement Heuristics and Biases’, ‘Critical Thinking’, ‘Reflexive and Recursive Actions’.
Even after the decades of research on biological aspects of human brain, with very minimal success, there were lesser than few concrete conclusions about the inner workings and functions. However, an exterior understanding of evolution framework illuminates, ‘Trial-Learn-Fail-Communicate-Adapt’, a cycle of actions that is reminiscent to the Elementary Principles of Nature.
In the last decade or so, researchers were able to closely replicate a few of fundamental abilities of our brain.
- Collecting and storing information — Data Storage.
- Calculating that preparedness to trigger recalling ability of such information and — using Sophisticated Algorithmic Calculations.
- Improved efficiency to adapt and change over time in the environment around us — developing Synthetically Simulated, Trainable AI Models.
The extensive use of connected devices and quick mobility has given a foundation for researchers and an opportunity for enterprises, to reap the benefits of Data harvesting. Companies with capable resources at hand, had fast-tracked the process of digital augmentation into territory of technically better and intelligent world. This opened new areas of opportunity for design research and interaction. To illustrate this case, let us sneak into today’s tech world.
Leveraging Data- Uber:
In late 2017, Uber’s engineering department, published a blog post which was a subject of particular interest for me. Authors, Lingxue Zhu and Nikolay Laptev had technically given a rundown, on how Uber leverages behaviour sciences to create a better environment under uncertainty in Our day-to-day short traveling experiences. The primary aspect of this article is to distill the concept of training an artificial neural network model for classifying behavioural data using Bayesian Inference and other theories. This didn’t come down as a surprise to me, but, a development of a production ready model, which is working at scale and defining the billions of rides over 700+ cities globally was something fascinating.
In short, Uber have adopted a neural networking model to forecast uncertainty at scale, based on all external factors in a given environment. With synthetic and real time data, after several stages of a model development, their AI team had come-up with a ‘Future-proof anticipatory services based business model’, which has been serving as a backbone for their revenue stream. No surprise that the current market valuation of Uber is at ~$91billion, while their direct rival lift is comparably standing very far at ~$17billion market value. The trend of ‘anticipatory behaviour’ based Information Architecture was in use for a very long time and with the help of artificial neural networks Uber pushed the horizons of behaviour sciences to a brand new level. Accurate Location Estimates, Advanced matching and Dynamic Pricing Algorithms are few of many production ready models they have developed, which are some of the best models in use. A novelty about the fact that they named their ML-AI platform after Michelangelo, is like a story that keeps on weaving for a possible future AI-ML Renaissance.
Neural Network Systems (NN) are wide areas for discussion and this explosive growth dates back to 1980’s. In 2020, this machine and automated intelligence had widened into many branches (ML, NLP etc.) other than just being confined to Neural networks (NN). This only means there is a path being laid down for being more realistic and simplified towards the development of AI. And the influence of AI on relative subject matters like Behavioural Sciences in everyday life, as in case of Weber using for daily commute is absolutely significant. This reveals how efficiently a variety of data set analyses can help us in shaping products that people actually use, effectively. Call that, Impact!
Influence of AI on relative subject matters like Behavioural Sciences in everyday life, as in case of Weber using for daily commute is absolutely significant. This reveals how efficiently a variety of data set analyses can help us in shaping products that people actually use, effectively. Call that, Impact!👍
Leveraging Data- Airbnb:
The Idea of ‘Bed and Breakfast’ in un-known places are very strange to normal people. For travellers, belonging to any place is a thumping feeling from inside heart. A company which mastered a way to intrinsically unite both such ideas and mint millions of dollars, probably billions is, Airbnb. In last four and half years, with current evaluation of ~$31 billion on the market, Airbnb claims it had paid close to $1 billion in local taxes on behalf of hosts globally. This 12 year old company, has bought quality in service and thus, the reward. With such soaring growth, one can only think, What would make the platform so special?
Similar to Uber’s data wealth comprising of 10 billion of the ride, Airbnb boasts they have half a billion all time guests to 6 guests checking into one of their stays every second. With great volumes of users flocking to their app and website each passing second, the challenges are completely different for them in comparison to Uber. The major challenge appeared to me was, Airbnb as a service is mostly dependent on their user’s financial situation and planning ability. Also a strong need in forecasting the behaviour of users, weeks in advance, maybe even months. Real-time Personalisation in Search Ranking based on interactions, Categorisation and Delivering Travels based on Context, Intent Discovery and Classification in Text Messages, with rapid productionizing of ML models Airbnb is cruising in this frontier at rocket speed. This business is no longer wearing a hat of ‘alternate to hotels’ and moved into spaces of adventures and experiences based on their users demand. In this beautiful article, author Mihajlo Grbovic (Senior ML Scientist at Airbnb) tells, how they leveraged existing user data (research data) to introduce and scale their two sided Experience marketplace.
AirBnB built its end-to-end metrics forecasting platform, leveraging Neural Networks (NN), Natural Language Processing (NLP) and Bayesian inference, etc. around its platform specific challenges, one thing that stands out is their holistic approach in dealing with each specific challenge. Most of it is a combination of Research with user at ‘centre of focus.’ Here is a short video on how Dr. Theresa Johnson (Product Manager, Payments at Airbnb) explains their approach to specific challenges.