A Great Simian or just a Monkey

Month: August 2016

cognitive pre-requisits

Tell me what I do not know!

Artificial Intelligence will be a huge part of our daily lives and in our businesses from now on, but what we see today is not always Artificial Intelligence or my preferred term, Augmented Intelligence. It is just re-packed machine learning with a conversational interface. AI and cognitive are not only algorithms it is so much more. Many high-end AI / Cognitive products today are not mainly algorithm, that goes for Google DeedMind as well as IBM Watson.

Without further rambling, I thought I would state my position on the topic. It is actually pretty simple. A product, service or process that will be an AI-enabled or cognitive shall have one single pre-requisite in my opinion.

An AI, cognitive product or process should tell us things we do not already know. If not, it is not AI or Cognitive!

If it does not meet that “general” term, it is not AI or Cognitive, period! AI and cognitive per se is to think, learn and reason etc, but to provide value in a business and to assist us, we need to guide our software in a way that they provide information, solutions and data that we could not generate ourselves with resources at hand.

The future of software in the enterprise is about elevating both us as humans, and our business. Efficiency and speed is fantastic, but that is nothing new, it is what we have today!

I think that in 5 years most processes and things we do in a company will be elevated by cognitive elements. AI and Cognitive will be everywhere.

Photo: From a morning walk last week in my new hometown.

watson cost

What does Watson Cost? What is the Price?

UPDATE: The Watson price table below is now updated as of April 24th 2017. Differences from the prior table are documented in my Watson Pricing and Language Updates post, where I have documented the differences in pricing as well as changes in what languages Watson supports since my last post on Watson language support from Dec 2015. Only the table below is updated. The table from Aug 2016 is still available here.

How much does Watson cost? What is the price model? Enterprise software price models are never easy and when entering the API-era it does not get simpler, but it often gets more open. This is all applicable to Watson. Unfortunately, the Watson cost is split, much like what languages that Watson supports, it is listed in different places for each service / API. For my own pleasure (not entirely true) I created a cheat sheet. It is included below both as a link to the actual Google Spreadsheet as well as an embedded table.

Worth noting is that these are the public pricing and I am fairly confident that they stick pretty hard to this structure, but there are discounts involved for partners and enterprises, these are not entirely public, but since being involved with some products / projects I know they are pretty fair.

Also worth noting is that most of the services has a premium offering to move from the public shared cloud and instead go for a “single tenant instance” / private cloud setup that will ensure better encryption and compute-level isolation. I have an upcoming post on security coming up as well.

The source of the prices included in the document is from both the Watson Services Catalog as well as the Watson services on Bluemix.

This post is more of sharing then trying to make a statement or write an intelligent post, it is simply sharing an aggregated sheet I made and maybe others can find it useful in some way.

Watson Cost and Price Model Cheat Sheet

UPDATE: 31st of July: New pricing for Discovery Service and Discovery News. Post from IBM: “New pricing for Watson Discovery Service“.

Watson Price & Cost table

The cover photo is from a visit to NASDAQ in NYC a while back

watson products

Watson Products, what are they?

Is it all about software you download and install or is it APIs in the cloud? Well, depending on who you talk to, you might get different answers on what Watson Products that exists.

Ginni Rometty (CEO and Chairwoman of IBM) always uses the term APIs and that they are services in the cloud. These services are used build the end-customer products. Internally at the local IBM office (non-Watson unit) there is often a mixed voice. Many struggle to change mindset to a API business model instead of traditional software license model and in Watsons case that means talking about Watson Analytics (sigh), Watson Explorer (WEX) or any of the other nifty packaged products they have released, like Watson Company Advisor and Watson Knowledge Studio. Many of these products are hybrid Watson products. A mix of classic software and Watson services.

The pre-packaged way, is in no way a bad approach, but I see it mainly as a way to package and make it easy for the customers to make a decision instead of having to build from scratch.

In my mind the only core Watson products are the services presented at the Watson Developer Cloud Watson service catalog. Those services are the core and the foundation of Watson. Watson is X amount of APIs all available in the cloud. It is also those services I would recommend all companies to primary work with.

14 Watson Services Available today

IBM have been consolidating the services lately and currently there are 14 different services available (from double the amount just a few months ago, the consolidation was needed, there was a big overlap previously). These services then contain from one to many APIs. AlchemyLanguage has 13 APIs and Dialog only one etc.

As of today, the Services are the following, I included an extremely short description, since some can be hard to grasp when reading the IBM lingo on each service.

  1. Alchemy Language
    A great service that analyzes text and sense sentiment, companies mentioned, person mentions, language used etc. It can also extract keywords which is very useful.
  2. Conversation
    A graphical interface to help designing Bots. A combo of Dialog and NLC I would say.
  3. Dialog
    A very “simple” API that only manages a scripted dialog.
  4. Document Conversion
    Converts text and documents to a format that Watson can use to learn from.
  5. Language Translation
    A service that translates text from one language to another.
  6. Natural Language Classifier
    Helps with deciding the intent of the input a user sends to Watson. Very useful since we all use different ways to express in different domains etc.
  7. Personality Insight
    A very cool API that helps us understand what type of person we interact with or receive information from.
  8. Retrieve & Rank
    The Jeopardy brain. This API is the one that answer your question and replies with a recommended solution on your problem, including evidence and ranking.
  9. Tone Analyzer
    What tone do clients use on your Facebook page or in customer service?Separate the happy clients from the angry ones etc. This API helps you understand.
  10. Speech to Text
    Obvious I guess
  11. Text to Speech
    Also obvious I guess
  12. Visual Recognition
    What is in the picture, is it a 40 year old man or a product from your company, this API helps you recognize things in images.
  13. AlchemyData News
    An archive of news and blogs. One of the least impressive APIs imho.
  14. Tradeoff Analytics
    Which product of these 5 is best for ME? TA helps me decide.

If you wonder what languages each API supports, read my post “What languages does Watson support” where I list language support per API.

4 Watson Products that are pre-packaged

As of today the Watson offerings (pre-packaged products) are:

  1. Watson Engagement Advisor
    Essentially a pre-built bot framework for e.g customer service
  2. Watson Explorer
    Search on steroids. Search both structured and unstructured data, build dashboards etc.
  3. Watson Knowledge Studio
    A graphical interface to build and train your machine learning models.
  4. Watson Company Advisor
    A high-octane version of company profiles. Find out whatever you want about any company and compare to yours etc.

These are the Watson Products as of today.

 

Augmented Intelligence instead of Artificial Intelligence

Artificial Intelligence is a term that, for me, only causes confusion and also builds up caution towards new tech. It indicates that AI will replicate our intelligence and some are even afraid of AI taking over our jobs etc. It is a term that stunts the development and our progress to more successful companies. A term that much better describes what it is about is Augmented Intelligence.

Augmented Intelligence is actually what it is all about. Augmented send the notion of evolving and increasing our existing knowledge. The intelligence that systems will assist us with is not artificial, it is real and elevated. It is systems that are taught by humans and then elevated to assist and scale human expertise to help businesses and applications to assist people with well-defined tasks and help them make more informed decisions.

This heavily connects to the fact that collaboration between human and computer is a very important part of becoming a successful business in the future. This collaboration is about systems helping humans and humans helping systems.

As a side note, augmented intelligence also fits much better as term in the landscape of cognitive. Artificial Intelligence is one part of cognitive, but many tend to mix cognitive with Artificial Intelligence. If we replace Artificial Intelligence with Augmented Intelligence, it is much easier and more relevant…..at least for me.

 

Microsoft Cortana vs IBM Watson

IBM Watson is a brain and Microsoft Cortana is a personal assistant. That is the simple conclusion of Yesterdays Cortana vs Watson (sort of) at The Conference in Malmö. My biggest surprise was that Microsoft Cortana is 100% scripted. Watson is a cognitive platform that actually could complement Cortana’s great features as personal assistant.

Currently in Malmö in Sweden, the conference The Conference (ehh) is happening and I was very much looking forward to the session Artificial Intelligence in Services. Primarily to listen to Deborah Harrison from Microsoft, since my knowledge about Cortana is limited. To summarize, Cortana was a huge disappointment to me. After listening to Deborah I posted the below tweet that described my immediate feelings about Cortana (excuse the typo).

So, Cortana is:

  • Scripted
  • Only a personal assistant
  • Personality and tone driven

As Deborah say already in the first minute of her talk (embedded at the end of the post) “Nothing is programmatic, we write it all”.

This leads me back to my recent post where I say that the future is about cognitive platforms and not about Bots (aka personal assistants).

Cognitive Platforms are the future, not Bots and AI

And to continue with the embeds, the comparison between Cortana and Watson was by Felix Segerbrecht described in a true and funny (with a touch of sexism irony included) way.

Cortana is a Personal Assistant and Watson is a Brain.

The speakers were senior and knew the topic way better than most (both speeches embedded below). Deborah Harrison from Microsoft is one of the original architects behind Cortana as well as head of the team that creates all the content that Cortana is communicating (again, Cortana is all scripted). Her talk was focused on how Cortana communicates and how she interprets different situations. It seems that Cortana is more of a Siri for the enterprise (guess that Microsoft is still enterprise focused).

Maya Weinstein is design lead at IBM Watson in New York. Maya talked to the Robot Neo (the real robot, not an app), gave the standard examples of oncology (cancer treatment) and kids education (Element Path) etc. She gave the standard variety of examples of use cases for Watson from concierge services (Connie the Concierge at Hilton Hotels as an example) to dating apps and medical examples. Naturally, Jeopardy was mentioned as well.

To put it blunt, Cortana could be built with just three Watson APIs, if you add a few more Watson could do even more, but again, I hope that does not happen, they should not compete, rather compliment each other. Watson is a cognitive platform and not a product in that way. Watson could even be the cognitive parts (aka the Brain) of a very good personal assistant (Cortana). Cortana is the UI / UX and asks Watson for help with all the cognitive parts she is asked to help her users with.

Nevertheless, it is interesting times and this is a true paradigm shift in technology and business…..and it is starting now.

We are now entering the World of Cognitive!

Maya Weinstein’s talk at The Conference in Malmö 16th of August 2016

 

Deborah Harrison’s talk at The Conference in Malmö 16th of August 2016

Top image is from my daughters origami book

ai chatbots and cognitive platforms

Cognitive Platforms are the future, not Bots and AI

Wherever you turn, whatever magazine you read, you see Bots, more Bots. Most of them sprinkled with Artificial Intelligence. In my view, this is a flawed and wrong assumption. The future is a platform that makes processes cognitive. A cognitive platform that will enable processes to think, learn, reason, solve problems and communicate in the expected domain and in the expected language. Bots and Artificial Intelligence are part of this platform.

The future is cognitive platforms!

Just a bot is just plain stupid, it does what you tell it to do, it is very much like traditional programming, the program does what it is programmed to do, nothing more, nothing less. It is a fun interface and an innovative communicative interaction. To escape the risk of endless loops and repeating the same stupid standard answer over and over again, we sprinkle some artificial intelligence on the poor Bot. Now we have a Bot that understands what you are saying to it and also can reply with a bit more elaborative answer then the basic standardized sentences. Does that mean that the Bot will actually assist us in our daily life and improve our performance as humans?

The above is true to a certain degree, in some cases, it will improve our lives, but mainly in the way of speed to information, the Bot will not tell us things we could not find out in other ways, it will just find it fast and deliver it to us. Such Bots are already here and have been for decades. Bots are simply put another type of interface.

The sprinkled AI is mainly about language and algorithms, a subset of the real meaning of AI. Languages in the sense that the Bot will understand you, classify what you intend to communicate and also often do this in several languages. A flaw today is that we often confuse machine learning and algorithms with the true meaning AI. It cannot be AI when most of the time it is a formula created in a spreadsheet that we then put to work on extensive amounts of data to help us to provide insight and conclusions to the user or the company. That is just machine learning. This is often done in very impressive ways (Spotifys Discover Weekly is a good example here), I have personally built companies mainly based on machine learning, so I am not in any way negative to the practice. It is just that it often (not in the Spotify case, just to be clear, they use machine learning) is described as AI and that it is the overall future of tech and in that I disagree. Why? Well, let’s start by looking at what AI is according to Wikipedia

Colloquially, the term “artificial intelligence” is applied when a machine mimics “cognitive” functions that humans associate with other human minds, such as “learning” and “problem solving”.

So, according to Wikipedia all is fine, cognitive and AI is the same, but that is not the way it is translated into today’s tech-world. Most describe AI as a combination of Machine Learning, Natural Language Processing and Algorithms. With this in mind, I prefer the “cognitive” term instead of AI. The AI term is in many ways already destroyed by those who claim to do AI, but mainly does Machine Learning and NLP based on structured data.

Lets not forget that 80% of todays information today is unstructured, not even possible to interpret for most systems today.

Today 80% of the information we have at hand is unstructured, that means that very few computers can work with that data in a valuable way. To be able to understand millions of medical journals or law books from all over the world the computer need to be able to learn, reason etc as well as to learn the domain specific language. You cannot apply an algorithm from a spreadsheet to accomplish this, a software that is very close to a human brain is needed. A cognitive platform learns, communicates and does problem-solving.

The Cognitive Doctor vs the Human Doctor (example)

Over the years we have done tremendous research on cancer. We also have a huge amount of medical journals on cancer patients. Let’s play the role of the doctor. You think you have read a lot about cancer, and this specific version of cancer. You have also been in practice for 20+ years so you have seen your fair share of cancer patients passing by as well as written a lot of research papers on the topic, which would classify you as a very knowledgeable and experienced doctor within the field.

A patient walks in with the diagnoses of cancer, the specific cancer type that you are experienced in. You look at the journal and talk to the patient. Given the research reports you have done over the years, as well as the one you have read from others in combination with your experience in the field you come to the conclusion that a specific type of treatment would be the best way to progress. Hopefully, the patient gets well and all is good.

Now, if the doctors working process would be cognitive the cognitive parts of the process would read ALL available research, ALL available medical journals. This in combination with getting all the data and info about the person incl. historical sickness and traumas etc that the patient might have. Again, let’s not forget that most of this data is actually unstructured and not even readable to most computers today. The cognitive process looks at all data (structured and unstructured) and swiftly replies to the Human Doctor with three alternatives (including the info these recommendations are based upon) for treatments for the patient, including a ranking on which alternative might be most suitable for this patient.

Now, the truth is that no doctor in the world can keep all information inside his / her head, the truth is that most of us tend to do what we have seen previous success with. What a cognitive enabled process now have given us is three alternatives for the Human Doctor to consider. It is still the Human Doctors decision, but the help presented might give new angles and new info that the Human Doctor actually had not considered. The decision is still the doctors, the decision is now only based on a more trustworthy base of data and information. Still presented to the patient by its doctor without even knowing that a computer actually helped a lot in providing the best treatment for you.

Is this pure fiction? Actually not! 

What is the future then? It is cognitive platforms!

For me, the success in this space lies in how well a company makes its processes cognitive. Cognitive will never be a specific product (like a Smart Bot) it will be integrated into existing (or new) business processes. Damn, that sounds boring, but let’s be honest here, most companies have defined ways of working aka processes. Most of these processes can be vastly improved by making them cognitive. To be clear, it is not new processes it is about making existing processes cognitive and by that more efficient, more valuable and more productive.

To be able to do this a process needs to be able to access different cognitive functions and consume them where appropriate. I could be to read, learn and understand a law book (and solve legal obstacles) or thousands of medical journals (like the cancer doctor example above) or it could be to find certain objects in millions of images or to analyze incoming communication, voice or e-mail, to be able to give a better customer experience, it could also be to give a quick, accurate  and personal (tone, personality) reply (in the right language) to a customer that interacts with the company.

This is only provided by a cognitive platform and not by a bot or AI (as positioned today by tech companies), but both are important parts of the cognitive platform.

PS I think I will break down this rant rambling post in a few more specific posts on the topics of cognitive platform, cognitive process etc Stay tuned DS

Photo from the movie Big Hero 6 by Disney

Powered by WordPress & Theme by Anders Norén