Artificial Intelligence is embedded in popular culture with incredible authors on the topic such as Asimov, Heinlein, and many many others writing on the topic. It is arguably only in the last 10 years, however, that we have seen the technology starting to have use cases that show a real promise to stand the test of time. Indeed, previous false springs for AI have caused a strong whiplash and lead to ‘AI Winters’ in the 70’s and 90’s.
Is this time going to be different? How can we tell? In this article, which is the first of larger series in the topic, Elinvar’s Director International, Niall Bellabarba puts AI under the spotlight. By highlighting some of the recent developments in the field and providing some key information, Niall identifies ‘fluff indicators’, some items that might indicate truths and, finally, what might come next.
Fluff, Fluff, and More Fluff
Artificial Intelligence is an area with massive investments and is becoming ‘mainstream’ with self-driving cars and voice-powered assistants like Cortana or Siri. The presence of the topic so centrally in popular culture and media is causing an inflationary pressure on news outlets to reduce the burden of proof, sensationalize and generally over-state the short-term potential of this technology and possibly not reflecting enough on the longer-term impacts. Here are some fluff indicators a prudent reader should look out for:
Vertical vs Horizontal AI
The technology is currently best as ‘vertical’ as incapable of automating processes that are well defined and doing so to great precision. For instance, keeping a car on the road even in presence of heavy traffic or listening to audio waves and answering an (ever increasing) range of voice commands are examples of ‘vertical’ task.
From this form of intelligence, no other form of intelligence can originate. To be even clearer, an extreme example: a self-driving car is not going to start studying stock market fluctuations. If within the media one were to witness a hopeful ‘halo effect’ of intelligence ‘spreading’ horizontally to another area the reader should see this as a fluff indicator. ‘Horizontal’ intelligence is orders of magnitude harder as the intelligence would need to learn skills it was never originally intended to have.
Suspicious eyebrows should also raise when technical complexity and buzzwords are overplayed and a clear use case from the technology is not readily presented. Given the vertical nature of the technology right now, the end use case should be simple. A self-driving car is (as fascinating as it is) clear to grasp, but if technical complexity appears to be the only discerning feature of a proposition within the media or a company website then it is unlikely any substance resides under the bonnet. In a commercial context, it is possible the use case is still in development and for business IP reasons it is not ‘published’ but if the complexity does not quickly disappear once ‘in person’ interactions commence the fluff flag should be raised.
Providing guarantees when Failure & Research is the name of the Game
In this ‘media-propelled but nascent industry,’ it is tempting to over promise and offers guarantees to future clients. Such guarantees should be taken with a pinch of salt, as the reality is that even within the most advanced and big budget, firms do fail in releasing functioning AI-based products.
For this reason, if a firm is truly in the AI space but is providing a guarantee of success also when not fully aware of key items such as the real business objective, the underlying data size or quality or the time available to produce results, then one should be highly suspicious.
Foundations of Facts
If the opposite of the above were to be witnessed, this would be an encouraging aspect to be worthy of further attention. Therefore, one should look for simple and narrow applications, technical complexity explained relatively simply and strong caveats around its applicability or possible error rates. Beyond this, there are other characteristics that should be encouraging:
Quick Demo & Happy Customers
Nothing speaks louder than the power of this technology than the ability to go through the use case personally to witness the ‘a-Ha’ moment. To this end, videos and online demos should be strong factors for encouragement. Just as vitally, nothing speaks louder than a happy customer who can in his own words explain the business rationale or the return on investment the product, service or change the technology brought to his / her company.
Technical complexity quickly broken down
In the event the technical innards where an essential part of the proposition (this would need a simple explanation in itself), then this explanation should be succinct and verifiable.
There are a finite number of Machine Learning engines and schools of thought (see Master Algorithm, by P. Domingos). If the innards of an AI proposition are explained by clarifying: the Machine Learning technique being adopted, if the starting point was an open source library or if everything is an ‘in-house’ build and the key items that will cause the ML engine to fail. Finally see and meeting the teams of Computer Science PhDs propelling the firm in is paramount. This area is one where, without this intellectual capital, the firm is not likely to be a serious AI player.
Using Vs Building
In many areas of human progress, we ‘stand on the shoulders’ of our predecessors and in the AI field this is even more true. In the voice recognition and natural language processing sphere the room for improvement over readily available open source solutions is near zero. This implies that all players in the AI / virtual assistant / chat bot space are with near certainty using ‘off the shelf’ technology and in the event they were not, they would be at a massive technological disadvantage versus those who do.
A strong fact indicator would be a firm that is quite open to the fact it uses AI technology for business purposes but does not intend to pass off as having an AI core competence to build this further.
This aspect of ‘only using’ AI technology does not make the firm any less of an AI firm because the correct usage of these still requires advanced technological skills and understanding. Indeed, it is a signal of earnest, as it is clarifying its internal core competencies. Collectively, we should hope for a massive proliferation of using AI firms and 20 to 50 large global players building AI technology.
Predictions are hard; especially about the future
In this section, we try to extrapolate the trends of the last years into the future. Also given the financial industry focus of Elinvar and its clients, it seems appropriate to list some predictions on the impact of AI on the financial world.
More Use Cases, More Hype, More Knowledge
The ‘mainstream’ use cases will become increase exponentially, currently, we haven’t even finished the proverbial ‘tip of the iceberg’. These will propel more media frenzy which will in the short term probably cause more confusion over ‘horizontal vs vertical’ AI abilities. However, in the medium term, as this technology becomes more ‘day to day’, we should expect more-widespread knowledge and inability to create headline-grabbing (and largely misleading) statements as is possible today.
Rising China & Hunt for Talent
It can be expected that China will rise and (eventually) technologically overtake most, if not all, Western nations in this area. Though currently there is no such superiority, the disparity of volume of investments (public and private), University publications and valuations for FinTechs indicate (to me at least) the long-term future of AI is in the East. Given the economic and geopolitical returns of deploying this technology in the financial sphere, we should expect at some point Chinese acquisitions of Western firms even with strong valuations.
In this context, the hunt for talent will further globalize and increase. The stars of AI will be Machine Learning PhDs, and predictably they will choose their careers paths in-line with academic and economic growth opportunities. We should expect even stronger competition among tech companies to attract the strongest minds and company managers in this space.
Robotic Beta & Human Alpha & eventually entirely AI based
Within finance, we can expect vertical automation brought forward by digital advisory automation firms (such as Elinvar) to become endemic, considerably more so than is the case now. There are few financial sector firms that use investment processes which could be described as AI based, but this will (in time) come.
In time, more elaborate AI-based models will emerge, but as to their success, only time will tell. Some examples: usage of pattern recognition across trading patterns to exploit signals of insider trading and ‘follow them’ (see Venture Beat) or using virtual traders that ‘live or die’ based on their fund performance allowing the new traders to learn from ones who ‘died’ (see Euklid). These are not even a beginning to what will come and what could be created.
For some investors, the human interaction will still be necessary in the key stages of financial decision making. Therefore, it is likely client advisors will be assisted through AI applications developing tailored investment strategies for their clients.
Despite the above might not be the message in the mainstream media (and possibly a disappointment after reading so far into this), the topic of AI is largely inappropriate for the vast majority of firms. Most firms are not even remotely ready for the adoption and usage of advanced technologies like AI. The extent of the usage of Robotic Process Automation RPA is further proof that firms are resisting true end to end digitalization as such they would record and automate with manual legacy processes that should long ago have been removed and streamlined or entirely digitized.
AI without a Digital foundation is like fitting F1 tyres on a Donkey.
Niall Bellabarba, Director International
Therefore, as the demand for AI on a customer side with mainstream applications will increase, the pressure on businesses to ‘skip’ generations of tech will increase as well.
At Elinvar we see that the winners will be the ones that come to the conclusion that it is not realistic for them to develop and maintain a core competence of technology and therefore choose to focus on their own core competencies, while partnering with digital & technologically native ventures to help them effectively compete in 2020 and beyond.