A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.1
Ian Goodfellow, Yoshua Bengio & Aaron Courville
Machine learning is essentially a form of applied statistics with increased emphasis on the use of computers to statistically estimate complicated functions and a decreased emphasis on proving confidence intervals around these functions[.]2
Ian Witten, Eibe Frank & Mark Hall
[W]e are interested in improvements in performance, or at least he potential for performance, in new situations.
Things learn when they change their behavior in a way that makes them perform better in the future.
Learning implies thinking and purpose. Something that learns has to do so intentionally.
Experience shows that in many applications of machine learning to data mining, the explicit knowledge structures that are acquired, the structural descriptions, are at least as important as the ability to perform well on new examples. People frequently use data mining to gain knowledge, not just predictions.3
The result of running the machine learning algorithm can be expressed as a function y(x) which takes a new digit image x as input and that generates an output vector y, encoded in the same way as the target vectors. The precise form of the function y(x) is determined during the training phase, also known as the learning phase, on the basis of the training data. Once the model is trained it can then determine the identity of new digit images, which are said to comprise a test set. The ability to categorize correctly new examples that differ from those used for training is known as generalization. In practical applications, the variability of the input vectors will be such that the training data can comprise only a tiny fraction of all possible input vectors, and so generalization is a central goal in pattern recognition.4″
“Data science isn’t woven into our culture; it is our culture. We started with it at the heart of the business, rather than adding it to a traditional organizational structure, and built the company’s algorithms around our clients and their needs. We employ more than 80 data scientists, the majority of whom have PhDs in quantitative fields such as math, neuroscience, statistics, and astrophysics. Data science reports directly to me, and Stitch Fix wouldn’t exist without data science. It’s that simple.”
“Emotibot, a Chinese caring robot solutions provider, has secured USD 30 million in series B round funding, led by China Development Financial, followed by Cathay Financial Holdings. Its existing investors Ecovacs and Advantech Capital participated in this new funding round. Light House Capital served as exclusive financial advisor in the transaction.
Founded in 2015, Emotibot is an artificial intelligence company engaged in deploying deep learning, Chinese semantic understanding, emotional calculation and computer science to provide emotional robot solutions. The robot can read, see, listen, remember, self-learn, and understand a user’s emotions, and the affective states, emotions and intentions of the speaker.”
“One of the world’s largest banks explains how it uses natural language processing and intelligent assistants to improve customer service on the web and mobile. Read this insider glimpse into the future of banking and user interfaces.”
“Retail investors are prone to two key mistakes. First, they tend to buy at the peak and sell at the bottom because these decisions are heavily driven by emotions. Second, people tend not to review their portfolios, particularly if they are already underperforming. This is another problematic behavioural trait that AI can help improve. We have a tendency to put off rebalancing our portfolios until it is too late,” she adds.
“A user could grant some level of autonomy to the AI assistant to invest objectively, and without emotion, on the user’s behalf. That said, ultimate control remains with the human and the AI assistant constantly keeps him in the loop on prospective investment decisions.”
Far from just the client-facing interactions, AI has the ability to augment a fund manager’s investment decision-making. AI will help finance professionals refine data points and catch patterns that they would not have otherwise detected on their own.
“By combining agent-based modelling with AI, fund managers or banking executives are able to map out a variety of financial, investment and lending scenarios, possibly even in real time,” says Durodié.
Even without resorting to advanced agent-based models, deploying some form of AI — for instance, machine learning in a forecasting and budgeting function within a hypothetical, global financial institution — could yield significant efficiency gains, says Durodié. “Suppose this bank is opening a Southeast Asian office in Kuala Lumpur. Typically, the headquarters would convene its leadership at the end of the year to plan its operational needs for the next 12 months. Then, headquarters would allocate budgets to each of its offices.
“With AI, this function can be conducted in near real time and with far greater efficiency. AI models could project business growth and its associated costs for the new Kuala Lumpur office, taking into account Malaysia’s very unique set of [economic, social and financial] circumstances. These models would then provide a much more targeted and personalised budget allocation for that particular office, over a three-month period, for instance. After all, given how everything moves so quickly, what 12-month predictions persist with absolutely no changes?”
“Apple, they write, stands out: its “Metal” API for iOS runs on a consistent chip platform and the GPUs in those chips are higher-performance, on average, “making Metal on iOS devices with GPUs an attractive target for efficient neural network inference.” Even then, however, the results of a “rigorous” examination of the speed of inference across six generations of Apple’s “A” series chips shows that within each generation of chip there is still “wide performance variability.”
“Programmability is a primary roadblock to using mobile co-processors/accelerators,” they write.
The newest version of Facebook’s “PyTorch” framework, unveiled this year at the company’s developer conference, is designed to “accelerate AI innovation by streamlining the process of transitioning models developed through research exploration into production scale with little transition overhead.” It also supports the “Open Neural Network Exchange,” or ONNX, specification backed by Microsoft and others.”
Online merchandising & personalisation How hard can it be? Join Micha Mokhberi, founder and CEO of Apptus, in a light hearted (but insightful) journey through the world of online merchandising and personalisation.
My favorite factoid from the HBR paper titled “Artificial Intelligence for the Real World,” is the breakdown of cognitive technology projects studied by Thomas H Davenport and Rajeev Ronanki.
The breakdown is likely representative of the current environment of projects:
- Robotics & cognitive Automation – 71
- Cognitive Insight – 57
- Cognitive Engagement – 24
- “On the AI assistant front, we saw Alexa and Cortana begin to work together.
- We documented how this fall Google, Facebook, and Amazon simultaneously fought major scandals while at the same time entering full pitch mode for smart displays, and we looked at the need for trust in AI assistant adoption.
- One of my favorites from Kyle Wiggers is about the danger that too much focus on apocalyptic AGI scenarios of the future will distract from pressing problems we face now.
- My favorite from former AI staff writer Blair Hanley Frank analyzed the way tech companies market AI solutions and proclaimed that Sensei, Watson, and Einstein must die.”