Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

IBM Watson API

developer.ibm.com

Post reply on HN
Print view View original

71–80 of 88 posts

Re: IBM Watson API

#71

Post by finin »

Rooster61 wrote: post #58
aabajian wrote: post #14

Earlier quoted context omitted.

Google / Wolfram Alpha / Watson are great at answering broad questions. 1. "Who was the 12th president?" - Zachary Taylor 2. "What color wine is cabernet sauvignon?" - Red 3. "Is a ferret a rodent?" - The ferret is the domesticated member of the Order Carnivora, Family Mustelidae and Genus Mustela. A common misconception is that ferrets are rodents. The real challenge is answering niche questions: 1. What size are th…

You bring up a good point, but it seems as if Watson was designed with this in mind. If you notice in the JSON response, it lists this query as a factoid class. It may handle different queries with different attributes differently, such as focusing on certain portions of its corpus or changing what aspects of its search results are more heavily weighed. A query identified as a factoid might be researched and judged v…

Factoid is a word coined by Norman Mailer for "an item of unreliable information that is repeated so often that it becomes accepted as fact". http://en.wikipedia.org/wiki/Factoid

Re: IBM Watson API

#72

Post by nradov »

tluyben2 wrote: post #7
readerrrr wrote: post #3

Out of curiosity I googled the same request. https://www.google.com/search?hl=en&safe=off&q=%22His+1983+h... . I think this might be useful if Watson was being feed with a medical database. Otherwise I don't see any need for it; is there any? edit: Watson as a legal consultant would be great. There might be a product in that, not as an replacement for a lawyer but more as guide/search tool.

I would think Watson can replace most lawyers (and MDs, and ...) of this world. Most of them don't think and just rehash stuff they learnt, just like Watson does. Sure for the exceptions you need actual people, but that is the same time when you would go from your corner lawyer to a more prominent one and when your doctor would forward you to an expert anyway.

Watson can't replace most physicians. Physicians have to physically interact with a patient to gather a relevant medical history as well as subjective and objective observations about the patient's symptoms. Robots and machine vision systems are nowhere near being able to fill that role.

Watson might be able to partially replace some specialists that primary care physicians use for consultations. When PCPs are unsure about a diagnosis or proper plan of care they will often consult with a specialist for advice via phone or e-mail. So in that case the PCP has already gathered at least some preliminary data and could feed it to a computer. But even for that use case Watson won't be able to provide same level of back-and-forth interaction that's often necessary to achieve the correct result.

Re: IBM Watson API

#73

Post by matznerd »

dalek2point3 wrote: post #59
pesenti wrote: post #15

IBM is about to make these APIs (and many others) much more accessible as part of BlueMix ( https://ace.ng.bluemix.net/ - the IBM PaaS/Heroku). I lead the team in charge of developing the Watson platform. Ask me questions!

I'm curious about the "copyright" field. Do you return the original sources from where Watson learnt the information he is presenting? What are the major sources? Have you faced copyright or legal restrictions to access information and has this affected Watson's ability to answer questions in a certain area?

When you create a Watson instance, you basically load it up with what they call a "corpus" of information. They are supposed to check the source of all your info etc that you have copyright to it. I don't know what it will be like once it is opened up more, but as of now they have pretty tight hands on Watson and will only accept projects with monetization and with revshare agreements...

Re: IBM Watson API

#74

Post by matznerd »

logicalman wrote: post #66
pesenti wrote: post #15

IBM is about to make these APIs (and many others) much more accessible as part of BlueMix ( https://ace.ng.bluemix.net/ - the IBM PaaS/Heroku). I lead the team in charge of developing the Watson platform. Ask me questions!

Is BlueMix a Heroku competitor or compliment? I want to develop on the Watson API but I also want to use a stack I'm familiar with. Also, do you have range estimates for the price of accessing the Watson API?

I went for an info session and a few events for Watson a couple of months ago at the IBM office in NYC. They are doing pricing with revenue shares. A lot of devs, including myself, were not happy about this. They basically were not open to projects that weren't directly making money off of Watson...

Re: IBM Watson API

#75

Post by matznerd »

cdnsteve wrote: post #36
pesenti wrote: post #15

IBM is about to make these APIs (and many others) much more accessible as part of BlueMix ( https://ace.ng.bluemix.net/ - the IBM PaaS/Heroku). I lead the team in charge of developing the Watson platform. Ask me questions!

1. How does Watson know to share only public data and not confidential data if it comes across it? How does it gather data? 2. Does it cache answers or does it try to "learn" and modify/improve the response each time?

1. Watson has to be trained on a specific corpus of information. Each instance is it's own Watson, and needs to be trained. They say that each Watson starts off as a kindergartner. 2. Watson needs to be trained by an expert using sample questions and it learns from that.

Re: IBM Watson API

#76

Post by matznerd »

jere wrote: post #44
pesenti wrote: post #15

IBM is about to make these APIs (and many others) much more accessible as part of BlueMix ( https://ace.ng.bluemix.net/ - the IBM PaaS/Heroku). I lead the team in charge of developing the Watson platform. Ask me questions!

>Data: at least 50 percent of content is unstructured, and sufficient volume exists Not quite what I expected. Does this mean the developers provide the data?

A given instance of Watson is unique based on it's corpus (documents that have been uploaded to it) and it's level of training and fine tuning. So far each instance has had to be trained from scratch, although they are making some Watson instances available that will come pre-trained in specific fields (like medicine).

Re: IBM Watson API

#77

Post by matznerd »

nradov wrote: post #72
tluyben2 wrote: post #7

Earlier quoted context omitted.

I would think Watson can replace most lawyers (and MDs, and ...) of this world. Most of them don't think and just rehash stuff they learnt, just like Watson does. Sure for the exceptions you need actual people, but that is the same time when you would go from your corner lawyer to a more prominent one and when your doctor would forward you to an expert anyway.

Watson can't replace most physicians. Physicians have to physically interact with a patient to gather a relevant medical history as well as subjective and objective observations about the patient's symptoms. Robots and machine vision systems are nowhere near being able to fill that role. Watson might be able to partially replace some specialists that primary care physicians use for consultations. When PCPs are unsure…

The volume of medical data being created and published is rapidly increasing, and it would take a doctor to read something like 160 hours per week to stay on top of their field. There are only 168 hours in a week, so there is really no way a doctor today can keep up with what is going on . The idea of Watson is to be able to take large amounts of unstructured data (i.e. an entire patient history and current symptoms) and be able to find a solution. I don't think Watson will replace a doctor or physical interaction anytime soon, but a device connected to Watson or similar solution will most definitely be used to augment the doctor's diagnosis and treatment solution.

Re: IBM Watson API

#78

Post by matznerd »

MartynX wrote: post #34
ilaksh wrote: post #9

So you just ask it any random question and it knows everything? Or only things that come up on Jeopardy? I don't see an API for feeding it information.

I would assume that they are working closer with companies that they trust to feed in information. Having that on an what will eventually be a public API is recipe for disaster.

Each instance of Watson is unique and has to be trained as such based on it's "corpus" (set of data) and actual feedback on the quality of it's answers by experts. The public API sounds like it will allow access to specific flavors of pre-trained and data-filled Watsons, like the food- recipe one or some basic medical ones.

Re: IBM Watson API

#79

Post by chris_j »

atratus wrote: post #57
pesenti wrote: post #15

IBM is about to make these APIs (and many others) much more accessible as part of BlueMix ( https://ace.ng.bluemix.net/ - the IBM PaaS/Heroku). I lead the team in charge of developing the Watson platform. Ask me questions!

The Q&A is not particularly interesting...are there other api's we can experiment with? Also, is there any precedent for getting clojure up on bluemix?

(Another) IBMer here. As far as I am aware, Bluemix doesn't have built-in support for Clojure but a few of us have played with it and there are a few ways that you can run Clojure apps:

I wrote a blog post about my experiences running a Hello World app using the Heroku buildpack for Clojure which you can find here:

http://mytediousblog.blogspot.co.uk/2014/07/running-clojure-...

In addition, since Bluemix natively supports Java apps, you can export a Clojure app as either an uberjar or an uberwar and run it directly on Bluemix that way. So for example, if you create an uberwar of your Clojure webapp then you can push it to Bluemix by doing:

  cf push your-app-name -p target/my-app-uberwar.war
... and Bluemix will run your app on a WebSphere Liberty Profile app server.

I'm happy to help if you have any questions - contact details in my HN profile.

Re: IBM Watson API

#80

Post by bkeroack »

thinkpad20 wrote: post #53
Ixiaus wrote: post #39

Earlier quoted context omitted.

BlueMix is actually pretty exciting to me; I think IBM has finally stepped up as a competitor in the "cloud" ecosystem where Microsoft has really failed.

"Really failed"? According to Forbes in July of this year[0], Microsoft is second only to Amazon in the cloud market, and gaining. I'm not sure how that counts as a failure, except in the sense that Microsoft makes a popular whipping boy. [0] http://www.forbes.com/sites/georgeanders/2014/07/28/amazon-i...

I suspect (but don't have confirmation) that this includes Microsoft's SaaS products like Exchange Online/O365, which are very successful.

I get the impression that Azure specifically as PaaS/IaaS, OTOH, is quite a bit less so. At least compared to AWS.

Post reply on HN