What to expect from Artificial Intellignece during and after the Covid-19 crisis? [April 2020]
 

What if creativity and a culture of innovation were not enough to innovate in the long run and on a large scale ? [Nov 2018]

What to expect from Artificial Intelligence during and after the Covid-19 crisis ?

 

 

 

 

 

 

 

 

 

 

 

[[[Photo propriété du Conseil de l'Europe]

 

As soon as the coronavirus pandemic developed, several start-ups and laboratories around the world announced the discovery of potential sources of treatment thanks to AI (Artificial Intelligence) (1).

Will AI also be a 'game changer' in this industry as already proclaimed in many others?

 

It is too early to say for sure (2), but let’s take the opportunity to understand precisely the impact of AI on businesses in “normal” situations or in crises, and the consequences for leaders.

 

First of all, let's remember that the widely publicized word AI is sometimes a 'catch-all' term which includes various technology fields from the digitalization of processes to platforms and robotics.

 

What makes AI unique today is in fact Machine Learning / Deep Learning (ML / DL) - the ability for computers to perform tasks and solve problems without being explicitly programmed for each of them but thanks to independent 'learning' via mathematical and statistical approaches (3).

 

AI as defined above is today implemented in various use cases, at pilot stages or beyond, and in many industries which range, without wanting to be exhaustive, from finance (portfolio management, fraud detection, cybersecurity) to health (advanced prevention, surgical assistance, radiological detection), education (personalized learning) or the automotive industry (autonomous vehicles).

 

Let us try to identify in a profusion of announcements some fundamentals.

Two examples of potentially significant crises in order to help our analysis

 

August 24, 2012: The day Amazon promised to deliver before we even had (too late) to think about it

 

This use of AI isn't pure science fiction, since Amazon has already filed a patent on predictive buying (4). The merchant's value proposition would be then radically changed from the traditional Order/Delivery to (pre) Delivery/Purchase.

 

This use case is based on the key added value of ML/DL, namely the ability to recognize objects and shapes (for example faces on a video surveillance) and more broadly correlations between digitized data. This then makes it possible to spot 'patterns'. In the case of Amazon, the need for products based on a purchase history, product characteristics and other variables identified from Amazon's customer base.

 

The ML / DL's advanced pattern recognition capabilities allow both diagnosis (what is the probability that this series of spots on a radio is a melanoma?) and prediction of future actions (what would a human driver do in the particular circumstances the car is facing?) or prediction of future states (what is the optimal planning of flight crew and planes?)

 

What ML/DL actually brings to businesses is this ability to generate relatively reliable predictions.

 

AI and reliable predictions: What consequences for managers?

 

On one hand, better predictions help make better decisions.

 

On the other hand, any human activity based on predictions is likely to be replaced by an AI predictive algorithm for a more or less significant part and whether this activity is itself very predictable or based on an implicit or explicit technical know-how.

 

Robots powered by AI will perform physical tasks in 'stable' environments after having 'learned' how to repeat actions of human workers (5). Activities centered on the acquisition and processing of information will also be automated - whether it is the processing of credit applications in a financial institution, online responses to specific requests or the diagnostic phase of a craftsman (6).

 

New 'native users ' of AI or competitors massively deploying AI are real threats to incumbents.

 

Decision-makers must therefore carry out classic strategic analysis:

  • Where do reliable predictions create competitive advantages in our business (and disadvantages if not achieved)?

  • Where precisely should we implement reliable prediction capabilities in our core business processes and support functions?

     

The next question is then quite often raised: Will AI decide for humans (7)?

A second scenario to deepen our analysis.

 

September 26, 1983: The day when American nuclear missiles were launched on the USSR

 

A date when a nuclear war was indeed very close.

The Soviet strategic missile defense center received that day from an advanced surveillance satellite an alert - 100% confirmed - about the launch of American intercontinental missiles. Tensions were then very high between the USA and the USSR (8) and such a signal of an imminent attack should have led to an almost automatic nuclear response from the USSR.

But on duty Colonel Stanislav Petrov ruled out that no information should be passed on to his superiors because it was a technical error according to him. The fact that it was an error was effectively confirmed later.

 

What does this crisis tell us?

 

Discernment and judgment are required when dealing with predictions.

 

Specially in the case of AI.
First, ML / DL learning is dependent on the quality and depth of the initial training data (9).

Second, qualitative and subjective information do not easily translate into digital data, in particular information dependent on the culture of the organization and its values.

For example, a taxi's GPS, even the latest generation one, can't yet fulfill the request: 'Take me to Neymar's favorite Parisian nightclub' (10), while a Parisian taxi driver probably would.

Finally and more fundamentally current AI recognizes patterns and makes predictions only according to a clear specific objective. Multiple, difficult to describe or subject to interpretation objectives cannot be achieved by AI today (11).

 

So human judgment remains necessary with regard to AI predictions.

Especially in unprecedented instances - for example the emergence of innovation - or in instances where emotional and contextual intelligence are required, for example in front of a client or a user - a classic case being the finalization and communication of a medical diagnosis to a patient.

 

AI and human judgment: What consequences for leaders?

 

On one hand, activities based on complex human interactions - such as managing and developing collaborators or managing stakeholders (team members, customers, users) will not be easily subsituted by AI tools, as long as these tools don't integrate unspoken, culture based and emotionally rich information. Physical and manual activities in an 'open' environment - not predefined and changing – will also not be easily substituted (12)

 

On the other hand, reliable predictions become more available in many jobs inside the organization .

Discernment and judgment start to be required at all levels in order to define the right objectives, to know how to interpret the results and to check their relevance.

 

There is in addition no guarantee that the AI training data (the past) may reflect all the possible conditions in the future. In the case of 'abnormal' conditions the decision-maker must be capable of 'taking back control'.
Delegating entirely too many daily decisions to AI tools may well lead to the 'downskilling' of the workforce. That would be extremely dangerous in the event of a crisis (13).

 

Leaders need therefore to answer two key questions before deploying AI within their organizations

 

  • To which extent are we ready to decentralize decisions? (14)

  • How to improve discernment and judgment inside the organization?

 

'Everything must change so that everything can stay the same'

 

Why end this article with a quote from the book 'The leopard' by Giuseppe di Lampedusa?

Because the emergence of ML / DL does not change the fundamentals of crisis and post-crisis management.

 

There are the Known Known – for example our knowledge of previous pandemics

There are the Unknowns who are Known – Questions are identified but we do not yet know how to answer - for example which substance(s) will help cure Covid-19?

In that case AI plays an important role.

 

There are also the Unknown Unknowns - the unforeseeable circumstances - for example, in the case of two skyscrapers being built, the possibility of two planes colliding almost simultaneously into these towers.

AI cannot provide an answer to a question not asked.

 

Finally, there are the Unknown Known - Facts that are known but whose meaning is not properly appreciated. A pandemic with a high rate of contagion and a low lethality like Covid-19 may well have been an Unknown Known for a certain period and in certain countries.

An AI tool may be able to make predictions on the future evolution of the pandemic, but no AI tool could under such dramatic circumstances sway decision makers.

 

The deployment of AI will have a major impact on any activity based on predictions, whether in a crisis or not.

 

However, the decision-maker's capabilities to discern and judge will remain key in order to be able to ponder on the decision and, if necessary, to 'take back control'.

 

Philippe Ginier-Gillet

April 2020

 

1) The secretariat of the Ad hoc Committee on Artificial Intelligence (CAHAI) and the Council of Europe have compiled a first set of articles from the media and other available public sources [Link: https: // www. coe.int/fr/web/artificial-intelligence/ai-and-control-of-covid-19-coronavirus]. The need to assess the use of AI after the crisis is emphasized

2) The tests, approvals and ramp-up of possible treatments or vaccines require that large players in the health world take over from startups.

3) Let us note that ML / DL is actually the 3rd 'Summer' of AI. There have been since the start of AI in 1956 already two 'Summers' and two 'Winters' - periods of disillusionment with the 'Summer' promises. Hence our caution when reviewing AI.

4) US 8615473 - 'Method and system for anticipatory package shipping', patent issued in December 2013,

5) For example the highly publicized Burger Robot from the start-up Creator aims at preparing on its own hamburgers for $6 a piece

6) Let us note that this does not inevitably imply the loss of entire jobs.

Because jobs very often include other non-substitutable activities (for example interactions with colleagues, collaborators, customers or users). And a business case may well identify high migration costs, less attractive cost/performance ratio of the technology, and strong legal and societal barriers to the implementation of AI tools.

7) Which would ultimately be the case if a so-called 'strong' AI emerges.

8) A Korean Airlines plane was shot down three weeks earlier by Soviet fighter planes.

9) Data from the surveillance satellite during the USA / USSR crisis were not cross-checked with data from geostationary satellites. The check would have highlighted a rare case of reflection of solar rays through clouds at high altitude. The surveillance satellite mistook them for missile traces

10) A highly publicized soccer player who currently plays for the PSG club in Paris,France

11) Colonel Petrov had during the USA /USSR crisis not only the 'recognition of a burst of light coherent with the trajectory of a missile' objective but also other objectives, such as' coherence with what is known from US military doctrine' (number of missiles in case of an attack) or 'estimated reliability of the detection system'

12) For example forestry work, cleaning of public space or part of construction work

13) As a dramatic example, the loss of manual piloting skills was identified as a key issue in the crash of the Air France flight AF447 Rio de Janeiro Paris in 2009 when speed sensors and electronic systems started sending erroneous information

14) We should note that Colonel Petrov was later punished not for his decision but for not documenting the incident in the official activity report of the defense center

WHAT IF CULTURE OF INNOVATION AND CREATIVITY WERE NOT ENOUGH TO INNOVATE IN THE LONG RUN AND ON A LARGE SCALE?

Innovation has become a major challenge for CEOs and General Managers in recent years. And 'calls to arms' are regularly issued. These calls are sometimes pompous ('Let's create the desire to innovative') sometimes anxious ('Disruptive barbarians are at the gates !') and often forceful ('Let's dare innovate'). But a large-scale sustained stream of innovations is not so often achieved in many organizations


Innovation is even reduced in some instances to an exercice in communication ('Let's take a 'selfie' with the Start-up founders of our incubator') or to short-lived projects launched right after the latest fashionable seminar on'how to innovate differently'.

It is not surprising under these conditions that the term 'innovation' sometimes creates cynicism ('a General Management imperative, among other General Management imperatives...') or frustration ('after the 'Proof of Concept'? nothing very significant...')

So how can innovation efforts be successful?
One temptation is to seek a miracle cure. But such a cure would be known if it existed !
Our experience shows that there are in fact different key success factors.

Let's try to go through them one after the other

"Learn to walk by learning to fall" (or a culture of innovation)

Decision Makers in established organizations often begin to pay attention to innovation only when traditional markets disappear or aggressive new competitors rise.

Which challenge then do 'innovation beginners' regularly face and bemoan? The lack of an 'innovation culture' and a 'risk adverse mentality'
The solution ? Workshops or even team coaching to 'kick off' innovation, to 'change the culture' and to call for employee initiatives. The objective is to transform behaviours inside the organization.
But let us acknowledge that changes in beliefs and behaviours are often slow and go through phases of resistance and demotivation. Success requires 'sponsors' and 'advocates' across the organization and a mix of participatory and (partially) authoritative management style - on a large scale and consistently over time. Fast renewal of management teams may not help on this matter.

 

"To a hammer every problem looks like a nail" (or ideation)

Any organization that starts to innovate very quickly realizes that the transition from idea to market success is very uncertain. It is therefore necessary to feed upstream the pipeline of ideas in order to succeed downstream. Finding the 'Big Idea' becomes a major goal for the organisation. The imperatives are to'be creative' and 'to get out of the box'.
The good news is that there are at least fifty methods of creativity: the off-site brainstorming seminar will very likely generate a number of ideas. The bad news is that tangible outcomes of the seminar are not always visible (if there is any actual follow-up to the seminar).

 

"Nobody can play a simphony alone, it takes an orchestra' (or transversal innovation processes)

There are two organizational obstacles to innovation in any organization .
First, silos (units and teams in the organization that focus solely on their areas of responsibility and power to the detriment of other units). The issue here is that many different functions inside the organization need to be involved in order to successfully deploy innovation.
Second, the ''Non Invented Here'' syndrome (the view that all developments should be carried out inside the organization). Opportunities to explore new concepts, ideas and offers are then limited.
The solution is to implement cross functional processes inside the organization and external "Open Innovation" involving Start-ups and Academics.
The challenge ? A Hackathon (a weekend dedicated to the rapid emergence of solutions to a given problem thanks to teams coming from outside the organisation) or an Open Lab (a place where among other things usability tests can be conducted) do not always lead to significant industrial deployments.

Other success factors must be taken into account to ensure the success of innovation efforts.
Let's try to describe them below

"There is no wind favourable for the sailor who does not know to which port he is sailing" (or an innovation strategy)

 

General Managers in quite a few organizations communicate about the need to innovate and each department or function then carries out innovation initatives. But do these actions really contribute to the success of the organisation's strategy?
An example? A manufacturer of home cleaning consumer goods launches a project to design new floor cleaning tools. The only problem? The offer will be easily copied within the next six months by the manufacturers to whom production is subcontracted....

 

In strategic terms, where is the the long term competitive advantage?

 

Experience shows that strategic alignment - a classic managerial challenge - is not easily achieved when it comes to innovation. Innovative projects cannot be deployed in any significant way if there is no enforced consistency across the organization and clear leadership and governance

 

"The devil is in details"
(or 'innovation compatible' business management processes)

 

Traditional business management processes are defined and optimized for established activities and in relatively predictable markets. These processes often become chocking points when it is necessary to move from the exploration phase of innovation to the exploitation (large-scale deployment) phase of innovation

Below some of these processes and the challenges they create for innovation

 

Budget planning:

One hand hand, the definition of precise and achievable objectives (such as ROI) is only valid for already established businesses. On the other hand, investment in a project portfolio - without any certainty of success per project - is only acceptable for a Venture Capitalist - risk diversification being its hedge.

But which planning process should be used for scaling up innovation?
Specific processes exist such as Discovery Driven Planning but that they are not yet well known and deployed

HR processes :

How to expose Managers, specially High potential Managers, to radical innovation and then assess them?
High risk and failure prone projects are better avoided during a carefully managed career! But then what about creating a 'culture of risk taking '?

Reporting structure:

How can a new innovative offer become the focus of Business Unit Managers when this offer still generates only a tiny amount of the BU revenues?

Purchasing processes 

How to manage Start-up suppliers when initial purchase volumes are small and profit sharing schemes challenging ?

 

Innovation faces less visible hurdles than reluctance to innovate or lack of ideas.
Established managerial processes are key roadblocks on the innovation journey.

 

In conclusion - Key success factors for innovation

 

Each organization faces specific innovation challenges. 'One size fits all' approach will not enable large scale ongoing innovations

Decision Makers and their teams must therefore step back, reflect on the innovation maturity of their organization and pay attention to neglected success factors for innovation .