Can Watson save IBM? | Hacker News Reader

Can Watson save IBM?

ft.com

26 pointsapsec11230 comments

Watson of course is PR-ware, a sort of flux capacitor for selling seven figure consulting services.

IBM does truly deserve credit for making Linux a safe choice for corporate IT. Few people remember this, but believe me I sat in on IT focus groups in 2001-ish.

Every. Single. Person. In every single session in every city said "there's no way we could use it in production". In every group, there was one who added "... but if IBM..." (literally as a sentence fragment like that), then the whole room would agree out loud.

Spooky. And then IBM came out with a big Linux campaign and the rest is history. Anyway, that happened.

It doesn't make a business for IBM. But it was a good contribution. That too was a way for IBM to sell consulting services. Which is all they really are.

Thank you for this reminder, we owe IBM a lot for their early Linux efforts.

I was pushing Linux into production, hard, for a very large company in the late 90s, with only limited success. It wasn't until 2001, because of IBM, when the nearly fanatical resistance to Linux started to fade.

I know that there is a lot of PR around Watson but it is not just PR. There is a genuine effort to put AI-related technologies in the hands of as many developers as possible. If you want to check for yourself what we are doing, please go to: http://ibm.com/watsondevelopercloud.

We are trying to expose state-of-the-art technology in speech, dialog, text analytics, vision, machine translation, etc as a coherent platform. I am happy to answer (without any PR BS) any question you may have.

Ok, here is a question. I went to the site and tried the document conversion service. I gave it three very clean PDF files containing CS articles. In all cases, what I got in response was "The input document failed to be converted because An invalid XML character (Unicode: 0x0) was found in the element content of the document."

Tika is able to parse them without a glitch, and has been doing so for a few years at least.

So I tried with some sample HTML taken from the main italian news sites. In this cases, it parsed it, but returned some garbage HTML tags and no content. BoilerPipe or similar libraries are able to extract a clean body of the news in all cases.

How is that state of the art technology?

Thanks for giving it a try. I agree that particular test does not sound like state-of-the-art. But don't judge the whole package just based on one test of one API. It'd be great to have access to these files (you can email them to me at my YC ID @gmail.com). Best would even be to post your experience on our forum: https://developer.ibm.com/answers/smartspace/watson/. We are usually pretty responsive.

Sorry I was harsh. I tried the translation service and it work better.

Anyway, I don't remember exactly all the articles I tried, but two of them were

https://www.cs.princeton.edu/~chazelle/pubs/mst.pdf https://www.cs.ubc.ca/~condon/papers/chungcondon96.pdf

while one news that failed to parse was

http://www.repubblica.it/economia/2016/02/09/news/borse_9_fe...

I tried other articles and news, but I do not recall each of them exactly

We will check them and get back to you. Thanks!

Hey pathsjs, sorry for the bad experience... Nevertheless, thanks for the feedback. TLDR: I pushed a fix for the bad character issue, and those PDFs should convert now.

The long version: It has to do with the underlying structure of the PDF; some of the characters in the above PDF have glyphs for display but don't actually map the characters to code points. So, when we pull out the text, they come through as invalid characters, which we should have filtered out. This is an issue we've seen with (all?) PDF viewers; the text you copy from a sentence isn't always what you expect... But, we're aware of that shortcoming and are looking at some ways to improve the quality.

In regard to the extra content in the news articles, we're not currently trying to do what BoilerPlate does. If you want to include or exclude specific content from a page, we have config options to do that via XPaths. Though, we're always open to ways of improving our services, and incorporating something like that would probably be useful.

I've asked many IBM employees to explain to me what Watson actually does (like for real, not the marketing BS that's on TV) that would make me want to buy it. Nobody could explain it. Literally some where just like "yeah, I think they're still trying to figure out how they can make a product out of it."

The Jeopardy thing was cool but since then it's turned into this mystery black box that will supposedly solve all your problems, but nobody's can seem to point to any real successes that would matter to a business. Honestly it seems mostly like marketing fluff built on some midly proprietary distributed computing technology.

I don't know that this is "something that would matter to a business" per se, but here's something else they're working on with it: https://www.ibmchefwatson.com/community

You can check and use many things that Watson does by using trying and using our cloud services: http://ibm.com/watsondevelopercloud. I know there is a lot of PR around Watson. But there are also real efforts and some state-of-the-art technology in speech, dialog, text analytics, vision, machine translation, etc. I am happy to answer any question you may have.

It seems like it may be a "solution without a problem" so to speak.

The issue for me is does IBM need to be saved? Their business practices have continued to be patent laden and closed. If they went down the road in a more open nature I feel that they wouldn't be in the place they are.

The last 15 years of IBM has basically been open.

1) They're in the top 5 re: Linux contributors. (Guess who else is in there? Microsoft.)

2) They dropped their JVM (J9) for OpenJDK - again upstreaming their patches to the community.

3) Eclipse was founded by them in 2001 (later released into the public domain).

4) They released (at least part of) their ML system "SystemML" into the public domain as well, during the same month MS opened theirs and Google opened TensorFlow.

They aren't patent trolls in the sense of those Texas LLP's who lease a few hundred square feet and fire off scare-letters. They hold a lot of patents because they are one of the few companies that actually funds pure research. They make a ton of money by bilking banks/the feds/states who still run their RMV on AIX @ 2mill a year in support contracts, etc, who are vendor-locked in. A not insignificant amount of money goes to:

5) Post-docs from really smart uni's to publish theoretical comp sci work in more journals than you'd imagine, which guess what? is also public domain!

They're really not hurting the consumer at all. In fact, I honestly don't see how they could be MORE open.

------

[1] (And guess Who's not upstream'ing much at all but making billions off community written software? I'll give you a hint, its not Barnes & Noble! Even VMware donated back to FreeBSD. Support HN users who actually pay-it-forward[2] and use Tarsnap[3] instead of S3.)

[2] https://www.freebsdnews.com/2015/02/25/donate-freebsd-founda...

[3] www.tarsnap.com . Not affiliated with Colin but he's been a long-time HNer who isn't a conglomerate, taking GPL/BSD software, slapping a pretty coat of paint on it and selling it for huge margins. Pay a few cents extra and support startups. Karma, guys. < /preachy hippie shit>

IBM is the number one patent company for each of the following years. 1999 - 2015 I can't see much beyond this.

https://en.wikipedia.org/wiki/List_of_top_United_States_pate...

> 1) They're in the top 5 re: Linux contributors. (Guess who else is in there? Microsoft.)

quantity and quality of contributions doesn't mean that they are the most useful contributions though

Open Source has won in Enterprise and every company has examples. Microsoft has continued down that road especially the last few years. IBM has made some moves but the vast majority of its strategy is with closed source projects like Watson.

I really don't want to turn this into a Microsoft praise fest but similar to IBM, they've been a net-zero since about Ballmer era, and a net-positive since Satya took over. All of that Microsoft Research papers which silently go unnoticed in Silicon Valley (or published under the MSR, INRIA // MSR, Cambridge name and people don't do the math) but here's a few things MS has been responsible for

a: Asynchrony -- the async / await paradigm - if you use it in JavaScript it came from C#, which came from a MSR paper (or maybe a Microsoft Languages whitepaper.) A lot of the semantics took a very long time by very smart people to work out the execution logistics, and we get to reap the benefits.

b: Haskell -- paying SPJ's salary so he can work on Haskell full-time as a MSR Cambridge employee and implement a whole bevy of things like STM[1], the popularization of pure functions (immutable.js users, as well as CLJS)

c: Clojure -- if you use Clojure's STM, borrowed from work funded by Microsoft. For me, it's used every day in Datomic.

d: Verifiability -- a mind-blowingly large amount of formal verification stuff - ranging from the first formally verified OS + integrated language, to things like F*

e: Z3 -- no need to explain this if you're in any sort of reverse-engineering I'm sure your life has been made significantly easier since MS released this into the public domain

f: CoreCLR + the JS engine + their crossplatform 100% free IDE + opening up VS Community (which is about as as VS Professional)

g: Probably most importantly -- Well, just go here. http://research.microsoft.com/en-us/groups/rise/default.aspx. Basically 1 out of 4 things in there is already, or will be imminently, used by 'mainstream' languages after it moves out of the "only publishing grad students who eagerly wait for POPL every year" zone.

[1] I know STM comes from that '88 paper. I've read it. SPJ did too and brought it into the mainstream, so I give him credit.

IBM and Microsoft (and, presumably GoogleX but they havent been publishing) are one of the few corporations I'm aware of that dedicate more than a few resources to research-for-the-sake-of-research (n.b. MS has been doing this silently since 1991, and IIRC employ more than 100 staff just to produce work to the field of computer science further.)

Pretty trivial stuff in the context of an $81.7bn corporation. IBM makes the vast majority of its turnover and profits from closed, proprietary systems, most of which are in decline (it used to be a $107bn corporation).

> They're really not hurting the consumer at all.

The public pays indirectly to support the extremely expensive IBM systems that, partly for historical reasons, still run almost all Fortune 500 companies. You may say that "They make a ton of money by bilking banks/the feds/states", but ultimately it's consumers and taxpayers who foot the bills.

PS You forgot OpenPower ;-)

I considered that -- we're effectively subsidizing research. I feel way better about that, then giving defense contractors 300/hr to drag&drop some Telerik components into Visual Studio. I figure that's the lesser of two evils IMO. Some companies don't even pay taxes -take Apple[1] who posted 30 billion in profit (for those of you who haven't paid retained earnings taxes, that's ~10 billion that should have gone to the federal government, and if you live in California, your state wouldn't be broke anymore (or significantly less broke)). More than $100 billion dollars homed in a nominal 'headquarters' to rob hard working Americans. (Granted I'd be transfer pricing too if my corporation could afford to retain Skadden, haha.) On the upside, I think they pay the salary for a few LLVM guys.

We digress. Support other startups, if you have a startup - use Tarsnap. ;)

[1] http://www.theguardian.com/technology/2013/may/29/apple-tax-... (The Guardian does have a bit of a liberal slant, FYI. Think of them as the UK's NYT.)

The spending on pure research is trivial compared with IBM's profits. In its last financial year: "Total capital return to shareholders of $9.5 billion; dividends of $4.9 billion and gross share repurchases of $4.6 billion" http://www-03.ibm.com/press/us/en/pressrelease/48846.wss

IBM also spends a lot of money buying dozens of companies to bolster its results. These include These include Cognos ($5bn), Softlayer ($2bn), Netezza ($1.7bn), Kenexa ($1.4bn) and Trusteer ($1bn). http://www.zdnet.com/article/ibm-vs-microsoft-and-the-battle...

use Tarsnap instead of S3

I'm not entirely sure what you mean by this... I mean, Tarsnap is built on top of Amazon S3, so using Tarsnap will still mean that your data is ultimately stored in S3. The difference is that tarsnap would deduplicate, compress, encrypt, and sign your data first.

Why don't they just have watson play a few more rounds of jeopardy and use the winnings to support the company.

This would actually make for decent television. Man vs. Machine.

Watson is no longer the system that won Jeopardy. It is a collection of technologies that have nothing to do with machine intelligence and some that do. What is clear is that by no means has IBM won the game against FB, Google or Baidu. In many cases they can't do it. They have invested too much into a specific technology framework.

But what I believe is that IBM will survive. Their ability to sell will win over technology.

I feel Watson is very heavily marketing heavy. I've tried their APIs and their technology is very limited in functionality and accuracy. I have seen too many videos around "cognitive abilities" but their application at the moment from trial runs seems to be extremely crude.

IBM will just need to focus on what they are good at. And Cloud Server / Softlayer just isn't one of them to be honest.

https://en.wikipedia.org/wiki/Betteridge%27s_law_of_headline...

No, Watson cannot save IBM.

with gorilla gone, is there hope for man?

IBM is not a technology company - it is a technology holding company.

Watson has been around for a fair number of years; almost five years since the Jeopardy stunt. Aside from the story about Watson turning into a foul-mouthed asshole after being exposed to Urban Dictionary, I've not seen much that is particularly impressive out of it.

Reply on news.ycombinator.com