The Approaching ‘Singularity’

In science fiction and
technology, the term "The Singularity" is
often used to reference that point in time when computers surpass human intelligence.
In 1997, IBM’s Deep Blue computer beat the leading human chess champion,
Garry Kasparov. Tonight, IBM’s computer, Watson (named after the company’s
founder), will take a seat on the Jeopardy quiz show to challenge the
competition’s top two human players: Kenn Jennings, who has won $2.5 million,
and Brad Rutter, who won $3.25 million.

Watson is not your average desktop computer. It is comprised
of 10 racks filled with IBM POWER 750 servers that are able to perform 80 trillion
operations per second on 2,880 processor cores. With its 15 terabytes of memory,
Watson is able to scan two million pages of content in less than three seconds.
Special software created by a team of 25 developers over four years is not
only able to extract facts but is also aware of inferences and context. The
software can distinguish between "bat" as in baseball and "bat" as
in Dracula, as it uses Natural Language Processing (NLP) to extract facts from
original documents.

Anyone who watches Jeopardy has an appreciation
for the cryptic answers that are often presented. A contestant must construct
a valid question that produces the answer before any of his opponents. Watson’s
NLP is combined with IBM’s
DeepQA (Deep Question and Answer) technology to finally make it possible for
computers to compete. The general goals for this technology are to retrieve
natural language text from multiple sources such as technical reports, novels,
dictionaries, encyclopedias, etc; understand, synthesize, integrate and rapidly
reason over the extracted knowledge; and to deliver a meaningful response in
natural language.

But the technology in Watson has far reaching application
beyond challenging game show contestants. It has potential in customer eelationship
management, regulatory compliance, contact centers, help desks, medicine and
numerous other fields. The critical factor is its ability to work with natural
language. Using NLP, it can tell you the best widget on the market by searching
and compiling all the customer reviews on the internet. It can review all the
weather related news reports to determine how different market areas will react
to weather phenomena. How about reviewing consumer purchases and demographic
data based on local customs to answer the question of who will purchase a new
product? It can scan social networking sites to give a company warning when
its reputation has begun to suffer or a competitor’s
has started to rise. Some of these things were previously possible, but they
often involved tedious efforts to edit data into a standard format. Now the
data does not have to be pre-edited and the answers come faster and more accurately
because they take into consideration more factors.

There is also a potential dark
side to this technology, however. Will there still remain a role for highly
paid experts or will people become merely data gatherers feeding facts to the
computer? Will machine learning make it more difficult to find people capable
of making the decisions necessary to lead a business? I have already given
up trying to remember phone numbers and the GPS has replaced my need to reference
maps. I won’t even talk about the use of calculators
in schools. As we give up using our minds in everyday tasks, do we risk losing
our ability to think all together?

BrainTrust

Discussion Questions

Discussion Questions: What do you think Watson-level artificial inteligence will mean for the retail industry? In what areas of the industry do you see this type of “brain power” first being applied?

Poll

18 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Dick Seesel
Dick Seesel
13 years ago

Aren’t we already in the age of artificial intelligence? After all, RetailWire panelists have commented recently on the use of RFID technology to “talk” to consumers with digital signage when they pick up a particular pair of shoes. And there are multiple other examples of sophisticated “algorithms” driving consumer behavior, from Google’s search engine technology to Amazon’s “recommendations” to its shoppers. But I believe there is still a strong “human factor” driving new uses for this application. We haven’t turned into robots yet!

Paula Rosenblum
Paula Rosenblum
13 years ago

I always thought a “Singularity” is a physics term, to describe a once-in-forever event, like the Big Bang. But your point is taken.

There are actually 2 different questions here. One is, will computers become “smarter” than we are? Probably. If it’s a memory debate, I lose already!

The other is, will people stop being social creatures? Not likely. Retailers made a TERRIBLE mistake in the early 2000s trying to use self-service as a proxy for customer service in stores. Consumers have recognized it’s MUCH more convenient to just serve yourself at home. This creates a couple of things–too much price sensitivity and almost no brand loyalty (your loyalty is as good as your last free shipping event). But people need reasons to get out and engage with other sentient life forms.

So I see this as a crossroads for retailers–where they must bring employees back INTO the stores, not take more out of them.

Fabien Tiburce
Fabien Tiburce
13 years ago

Great topic! I am keenly interested (call that a hobby) in the semantic web and this is right up that alley.

This kind of system is actually a break from earlier, cruder attempts. Earlier AI attempts were not AI at all. They were “bags of tricks,” full of if/then statements coded to emulate understanding by spitting out answers but, by design, not engineered to be intelligent (no “concept,” no linkage, etc…). The IBM system (and other semantic answer engines like http://www.wolframalpha.com/) actually use natural language processing to construct linked concepts (an object representation of concept in computer programming). This is the premise of the “semantic web” which always seems to be the “future” of the web (but this future hasn’t materialized). The problem with these types of tools is that they more or less require widespread semantic adoption to be successful. But no one wants to invest in the infrastructure without existing demand.

This “chicken and egg” situation is familiar, it also affects the so-called “hydrogen economy.” In other words, I find the technology both interesting and promising but I am afraid it could be a decade or more until we start seeing this type of system commonly used.

Dan Berthiaume
Dan Berthiaume
13 years ago

What it should mean is that e- and m-commerce should become much more customer-centric. For example, while Amazon.com does a pretty good job using algorithms to suggest books based on purchase history, advanced AI would make those suggestions much better and also more timely. Online cross- and up-selling could become revolutionized.

Marge Laney
Marge Laney
13 years ago

I agree with Paula 100%. Brick and mortar retail will go away if retailers don’t give customers a reason to leave their couch.

Ralph Jacobson
Ralph Jacobson
13 years ago

Many examples have been suggested as to why we need not fear innovation. Remember that it is still humans who perform the development of these amazing machines. Therefore, our brains will be challenged into the foreseeable future to continuously improve technology that serves the public.

Watson is a great example of the analytics capabilities that are required today; yes, today, to uncover patterns in the seemingly random shopper traffic in retail stores.

The quiz show (www.ibm.com/Watson) will demonstrate the incredible power of the technology, however, the potential for business growth for retailers and CPGers alike by utilizing this tool is virtually limitless.

Dan Raftery
Dan Raftery
13 years ago

Great topic for this forum. Bill Bitner’s discussion questions about ‘The Singularity’ have no “if” context. No “when” either, for that matter. I guess we have to drink the kool aid of AI.

While I don’t disagree with the point about losing mental capabilities as we transfer them to machines, the way language changes over time (I am only familiar with English and assume it is the same in other languages) will be an interesting challenge for the AI creators.

I had the opportunity to hear a commercial from an old 1940s radio broadcast of Casablanca. It was a hoot. The subject was simple: face soap (“toiletry soap” back then). But he storyline was clearly rooted in the times.

The speed at which the language is changing is of course accelerating by the very machines that we are discussing. So, another fun question is: how will AI both cope with and participate in the human evolution of language?

Regarding the use of Watsons in the food business–get real. There’s no margin in this business for that, at least not until other businesses use it to the degree that the cost drops a lot.

Ed Rosenbaum
Ed Rosenbaum
13 years ago

Forgive me for being a questioner; but how will we continue with any social interaction if we divert to artificial interaction? Social interaction is important to the success of the selling model. I doubt any artificial means will ever replace the human interaction. Except with those people who prefer not to speak face to face with another person. This gives them an option. They can shop and not have to speak with anyone who counts.

Paul R. Schottmiller
Paul R. Schottmiller
13 years ago

The computer brain power (analytical analysis) has not historically been the real challenge for retailers. The challenges tend to be in the quality and timeliness of the input data and the operational execution, particularly at the store level. Using the Jeopardy analysis, it doesn’t matter if you are the smartest contestant, if you can’t “ring-in” to answer.

While human interaction will be necessary in some segments, more and more interactions will involve self service technology assists vs. in-store associates (either from the retailers or other 3rd parties) and even the human interactions are likely to be increasingly from remote associates (i.e. video). There is significant innovation in the machine/human interface (both fixed and mobile…saw many of these at NRF this year) and I expect to see elements of these working their way into the store shopping experience resulting in new store labor models. More simply put, fewer store associates in-store but a better customer experience overall.

Dan Frechtling
Dan Frechtling
13 years ago

AI has advanced since the 60s with gains in processing speed, memory, and quantitative models. In retail, the quality of data has improved not only with NLP, but also with new data sources such as demographics and purchase history.

AI already drives in-store digital, as Richard notes. It also already serves operations, merchandising, and marketing. Distributed intelligence powers inventory management. Analytic processing enables site selection, marketing mix modeling, demand forecasting, and so on.

More broadly, AI can process and calculate but not interpret or decide at a strategic level. It can follow pre-programmed rules when tracking storms or directing air traffic but needs human control when rules change.

And the rules often do change. What happens when goals change from minimizing working capital to maximizing product availability? Or increasing sales per SKU instead of growing overall brand sales? Or targeting mainstream instead of high end buyers?

I’ll disregard the Jeopardy rules and phrase the answer as an answer. Because rules change, humans can push tasks to machines, but they can’t relinquish decision control to Watson.

Gene Detroyer
Gene Detroyer
13 years ago

COMMENT #1: Paula and Marge, there is already very little reason for the shopper to leave their couch (or desk). A retailer can put the best people in the world in their stores and it still does not change the equation. Service is not interaction with real people. It is the result of the entire purchase process.

COMMENT #2: Practical intelligence is based not on IQ, but on the memory (already mentioned above) and recognizing patterns. In both cases, computers will win. Where can they be applied? Almost everywhere. Where can they help us the most? Let’s try healthcare and diagnosis. What is the biggest problem? It is still GIGO. But, as computers get “smarter,” they will eliminate the garbage.

Just imagine a computer versus the politicians in solving today’s U.S. budget problems. But, we don’t want to try that. The truth might come out. Americans don’t understand simple arithmetic.

Craig Sundstrom
Craig Sundstrom
13 years ago

“What is unlikely to be unsurpassed in our lifetime?”
> The human brain. (Dan may not have been up for the challenge, but I thought I’d give it a try.)

Watson sounds very impressive, but I believe it is still “thinking” at a very rudimentary level…if even at that. And won’t we always be bound by the axiom that a person can’t construct something smarter than themself?

Lee Peterson
Lee Peterson
13 years ago

I find this discussion both humorous and sad in that, to me, it’s already happened! Computer programs have already taken over retail–and for the most part, look at the results: boring! This is especially a huge affliction in specialty retail.

I wonder what retailers thought this past Christmas when they went into an Apple store and saw 20+ employees (red T’s on) in a 5000 sq. ft. store. Probably, “gee, at $4000 a sq. ft., they can afford it.” But I would say that one begets the other; excellent customer service brings about increased business. And this, from a computer company!

I also think that computers have taken over the merchandising systems to the point where there’s no “gut” instinct anymore. You can see it in the assortments. They’re all ‘risk free’. Again: boring (maybe J. Crew gets a pass).

So to me, I think we should go the opposite direction: more human elements; better training, better recruiting better professional development, more risk taking.

But, if Apple gets rid of their fantastic staff in lieu of some talking head…I may change my tune.

Carlos Arámbula
Carlos Arámbula
13 years ago

I love the idea of using these capabilities in inventory control and other logistical situations. The ability to use real time data would allow retailers to cater to consumer’s demands and respond to any changes in consumers behavior prompted by issues out of the norm (weather, trends, cold season, and so forth).

Think how effective a drug chain can be if data indicates a surge in colds, or allergies, in a particular geography: It would carry sufficient inventory to supply the heightened demand and it would even allow for geography specific promotions and advertising.

I don’t see a replacement of human expertise. Instead I see human involvement evolving to another level with the use of new technology.

Charlie Moro
Charlie Moro
13 years ago

Just think for a minute how great a breakthrough this is for medicine. Doing diagnostic reviews with something that is interactive with the full breath of every option known to man…Retail…ok, not so significant. But instead of going shopping and hitting Bing to see where else I can buy it and for how much, I can save the trip and ask while I am in the car who has the most options, lowest prices, Zagat rated, etc. And for the retailer, the option of maybe having that knowledge as a sales person at point of sale, as an interactive tool…limitless.

Shilpa Rao
Shilpa Rao
13 years ago

In the structure of a human brain, there are two important parts which are involved in making a decision–the old brain (cerebellum) and the new brain cerebrum. The old brain is responsible for “flight or flight” activities and the final decision is made by it. It is responsible for our basic instinct. This part of the brain is also considered emotional. Often when we refer the decision of “heart,” that’s actually the old brain at work. Whereas the new brain is more analytical, it processes data. But it only gives input to the old brain to make decisions.

Watson is a good emulator of our new brain. Will it beat our analytical new brain? Absolutely it will, no doubt about it.It has more memory and processing power. But decisions are not made by the new brain and that’s what makes us human. We are an emotional being. We make decisions first and try and give it analytical logic later. I think with computers working for our analytical brains, we will evolve emotionally.

I totally agree with Paula’s point that retailers made a terrible mistake of taking employees out of the store. After-all if the emotional connect is absent; loyalty is a mere plastic card. It’s time to reestablish that emotional connection; or else we will be soon fighting a lost battle.

Bob Phibbs
Bob Phibbs
13 years ago

We’re one step closer to SkyNet.

Dr. Linda Whitaker
Dr. Linda Whitaker
12 years ago

I think instead of posing the question “will machines take over?” the proper question is, can we learn to live together and exploit each others’ strengths, and both come out ahead? Can we exploit the human machine relationship, so that both benefit from each other?

In order to do this, computers need to be able to speak and understand the language of the experts interacting with them. While I am not suggesting we will ever pass the Turing test, Watson was able to communicate with humans in a specialized application.

However, Watson was programmed to compete against humans, and not work with them. For computers to support retail, they need to speak the appropriate language. We have already found that you can teach (i.e. program) computers to use business strategies such as, like “What happens when goals change from minimizing working capital to maximizing product availability? Or increasing sales per SKU instead of growing overall brand sales? Or targeting mainstream instead of high-end buyers?” But this is limited in scope, and programmatic in communication.

On the opposite side of the fence, computers are starting to be used to ‘listen’ to human conversations via Twitter, Facebook, etc, and use these discussions to predict the stock market or, closer to home, to serve up retailer promotions. And in a sense, the feedback measured by the success rate of these promotions is a way in which humans talk back to machines.

It does not seem so far off to assume that human consumers and retail experts could directly communicate with systems to change the way that the consumer is served. But since humans are the final decision makers in the retail supply chain, computers will always be a slave to them and not the other way around.