How reliable is machine learning (ML) and artificial intelligence (AI)? Our team at Quantflix tested this by doing some research on the much raved about Xiaomi IPO and tested this against the results from PCMag Asia’s Editor’s Choice for “The Best Self-Service Business Intelligence (BI) Tools of 2018”, IBM’s Watson Discovery.
“Losing China Has Taken Wind Out of Xiaomi’s IPO Sails” – Bloomberg
“Investors at the Hong Kong roadshow said the share pricing was expensive even at the lower end.” – Financial Times
“The new valuation range is far below the $100 billion touted by sources earlier this year and below the more recent $70 billion plus valuation target that some analysts and investors see as aggressive.” – The New York Times
“But there’s a real risk that Xiaomi’s internet business will not achieve the same success overseas it’s had in China.” – Quartz
“Hong Kong retail investors give Xiaomi’s IPO the cold shoulder, put off by high financing cost and valuation.” – South China Morning Post
These are all quotes extracted from articles published by the above publications. Based on our market research, all 10 of the publishers we identified (other names include Channel NewsAsia, CNN, Reuters, Yahoo, and ZDNet) had negative sentiments towards Xiaomi’s IPO.
However, when we compared these results with those from IBM’s Watson Discovery, we found that the results contradicted our research.
IBM Watson Discovery Sentiment Analysis for Xiaomi Valuation
Source: IBM Watson Discovery
For instance, IBM’s Watson Discovery returned a ‘Sentiment Analysis’ of 25% Negative Sentiment and 75% Positive Sentiment when we did a search for “Xiaomi + Valuation” with regards to news articles in the past week. The results came as a surprise to us as it turned out that of the 4 news sources identified, all of them had extreme sentiments of either positive or negative (typically, it would be a mixed sentiment along a spectrum leaning towards either side).
Furthermore, we did a sense check on these results by going through the articles that returned positive sentiments, namely that of Bloomberg and TechNode. In both cases, we found that the sentiments towards Xiaomi were either neutral or negative, far from the 100% positive sentiments that the Watson Discovery claims it to be.
For a topic like Xiaomi’s IPO valuation which has received much buzz from the media, one would think that there would be sufficient information available for a platform like IBM’s Watson Discovery to pull off an accurate analysis of the market sentiments. Unfortunately, in this case, the popular BI platform did not live up to our expectations.
While we are not suggesting that the Watson Discovery is undeserving of its positive reviews, our team at Quantflix believes that as much as ML and AI is a good starting point for speeding up processes, there are undoubtedly still loopholes present that require a human mind to refine and improve the results of ML and AI. For example, our Quantflix team will still constantly speak with our managers to verify and scrub data despite having machine learning capabilities. While ML and AI may not be perfect at this current point in time, who knows how much more we will be able to do with them 5 years later?
For more information about our article or about Quantflix, please email us at enquiries@hedgequery.com.