At the recent AnalyticsX with SAS Software, I had the pleasure of interviewing Oliver Schabenberger Executive Vice President, Chief Operating Officer and Chief Technology Officer at SAS. Schabenberger had some very insightful thoughts on on Artificial Intelligence, and how customers approach this topic. I want to thank Mr Schabenberger for his valuable time and it was my privilege to meet him and spend time with him to discuss these topics.
Schabenberger commented that the original ambition of artificial intelligence was regarded as easily solvable. He pointed back to Herbert Simon’s assertion that The machine will be capable of doing any work that a man (or a woman!) could do’. Now, we smile at the optimism, but, at the time, the industry entered the first winter of AI since expectations were not met.
The AI industry started again with the work of people like Rosenblatt(1958) in perceptrons, and then adoption of neural nets after the work of Rumelhart and McClelland(1986) which, despite a lot of efforts, brought about more disappointment and wrought another winter for AI.
Nowadays, however, things have changed drastically and AI is having
another boom. As Schabenberger noted, at one point, working on AI would
get you ridiculed at a cocktail party. Now, you don’t get invited to the
party unless you are working in AI. AI is back, and suddenly the market has a slew of AI experts.
So what changed? Why has AI come out of hiberation and back into the sunlight?
Today we have a different source of knowledge throughout the world:
Data. Data-driven is an adjective that we hear everywhere, and
businesses are rethinking their data, which has led to rethinking
Today, we are building better algorithms than ever. At the heart of
today’s Artificial Intelligence revolution is data science. We create
models using a combination of data science, artificial intelligence and
Analytics. The models impact many arenas, from healthcare, insurance,
education, medicine, finance and so on.
Artificial intelligence is no longer about rules or backpropagation.
It’s all about processing data and imbuing human-like intelligence in a
system, with varying degrees. After some time, the system develops its
own logic and there are plenty of examples where programs have learned
to program themselves through being programmed implicitly.
Fundamentally, the artificial intelligence boom is an analytics boom.
As it stands now, Artificial Intelligence is enabled by massive data
volumes, cloud technology and digital transformation, empowered by
advances in computing.
AI impacts people who manage organizations. Previously, managers would ask the question tell me how you did that. Now, this question is now replaced by show me the data.
This data driven approach to artificial intelligence has caused very
powerful transformation in the industry, but people are still confused
by what Artificial Intelligence really means. Schabenberger cuts a neat
and disciplined distinction between Narrow and General Artificial
Rethinking Narrow and General Artificial Intelligence
Shabenberger’s insight is that people’s expectations on Artificial
Intelligence can be distinguished by the tasks it is expected to do, as
well as its autonomy. Does it really think and work in a real
environment, or is it dedicated to a narrow task, which it does well?
Artificial General Intelligence
The goal of AGI is to create a general thinking machine. It is a
machine that has a thinking capacity. In Artificial General
Intelligence, the definition includes self-sufficiency and broad
human-like general intelligence, where the intelligence can be
extrapolated from one situation to another and the system can learn over
time. Perhaps, if it is to be human-like, it can forget over time, too,
and choose areas to focus?
Schabenberger calls this approach Artificial General Intelligence (AGI).
In this interpretation of Artificial Intelligence, the goal is to have
the machines behave, think and be like humans. It is to attempt to
realise human intelligence in hardware and software. It is distinct from
the narrow form of Artificial Intelligence because it focuses on a
breadth of activities, rather than a very focused attempt of using
human-like intelligence in one domain with a high degree of expertise.
Narrow Artificial Intelligence
In contrast to this perspective, there’s another approach to
artificial intelligence, which Schabenberger calls Artificial Narrow
Intelligence. In contrast to AGI, these artificial intelligence systems
do not think. These systems are applying and executing algorithms, not
thinking. They may also learn from evidence, but the data and the
modelling fundamentally comes from humans in some way.
For example, let’s take Siri, Alexa and Cortana. This is a narrow
form of artificial intelligence which solves very specific tasks. These
systems are purpose built systems, aimed at a specific objective. These
systems are executing algorithms that are programmed implicitly.
What does the distinction mean for industry?
The precise distinction will allow businesses to formalize and
understand what they mean when they talk about AI. With AI, everyone has
an opinion, and they can mean different things. By focusing on
outcomes, this means that business ideas can be generated that can be
actionable and achievable by focusing on an understanding of AI that is
Narrow, rather than the general, sweeping definition of Artificial
At present, data specialists and developers are the ones developing
the system in silos, with no ethics discussion, or values imbued.
However, these decisions cannot be made in a vacuum, with no form of
ethics discussion. We are all impacted by data, and we will all be
impacted by Artificial Intelligence. We cannot assume to have the
ethical superiority of the uninvolved, pointing fingers when things go
wrong. If we are putting God in the machine, let’s make it one that we
can and want, to live with.
And what do all these Artificial General Systems have in common? They
do not exist, and we have absolutely no clue have to build them,
according to Schabenberger. We should have a conversation about what if we could get there. How does that impact ethics? How does it impact jobs? Our planet?
To summarise, Schabenberger’s distinction is very clear for
businesses to understand, and to direct their discussions on artificial
intelligence. Bringing this clarity is essential when people use the
same term to mean different things, and the neat distinction will
facilitate successful discussions and outcomes for business.
Rosenblatt, F. (1958). The perceptron: A probabilistic model for information storage and organization in the brain. Psychological Review, 65(6), 386-408. http://dx.doi.org/10.1037/h0042519
Rumelhart, D.E; McClelland, James (1986). Parallel Distributed Processing: Explorations in the Microstructure of Cognition. Cambridge: MIT Press. ISBN 978-0-262-63110-5.