Is IBM's Watson Overhyped & Soon to Be Outdone?

IBM Watson may be the first name that comes to mind when you think artificial intelligence, ever since it beat out a human on the Jeopardy! quiz show back in 2011, but skepticism is mounting around whether the company will be the best -- or most profitable -- name in AI as the market picks up traction.

IBM Corp. (NYSE: IBM) makes a big deal about Watson. In fact, it's published over 200 press releases with Watson in the headline in the past five years and mentioned it at least 200 times in its prepared remarks on earnings reports since 2013, as well as investing more than $20 billion, according to Jefferies & Co. Inc. . The investment bank, however, believes the AI platform will struggle to keep up with competition, as the services component of AI deployments and the cost of doing business present problems for Watson adoption. (See IBM Watson Faces Tough Road – Analyst.)

"Our checks suggest that while IBM offers one of the more mature cognitive computing platforms today, the hefty services component of many AI deployments will be a hindrance to adoption," James Kisner, Jefferies senior vice president, IT hardware and communications infrastructure equities research, writes in a research note this week. "We also believe IBM appears outgunned in the war for AI talent and will likely see increasing competition. Finally, our analysis suggests that the returns on IBM's investments aren't likely to be above the cost of capital."

Jefferies acknowledges that Watson remains one of the most complete cognitive platforms available on the market today, but it sees it struggling to attract interest from companies that don't want to endure the significant consulting work to gather and curate the data required to take advantage of it. (See IBM Leads $15M Funding Round for AI Programming Startup.)

Other companies are also stepping up their AI recruitment efforts -- Jefferies says that Amazon.com Inc. (Nasdaq: AMZN) has ten times more job listings than IBM. (See IBM: AI Needs More Than Just Technology.)

What's more, companies have more choices when it comes to APIs. IBM dropped its pricing for Watson Conversations by 70% back in October, which -- in part -- leads Jefferies to determine that IBM barely recoups its cost of capital from AI investments. Jefferies estimates Watson plus associated pull-through revenue will only contribute between 3% to 5% to consensus earnings per share in 2019. (See Element AI Raises $102M to Bring AI to All.)

It sees competitors Nvidia Corp. (Nasdaq: NVDA), Pure Storage and, to a lesser degree, Mellanox Technologies Ltd. (Nasdaq: MLNX) gaining the most from both the uptick in AI adoption and in IBM's struggles.

For more on artificial intelligence in the world of telecom, visit the dedicated automation content page here on Light Reading.

While Watson was arguably the first and most well known name in AI, it's far from the only player today. AI is attracting interest from every major tech company and a slew of new startups. The technology is being explored by telecom service providers and enterprises spanning every major vertical as a way to improve their applications and customer service, better manage networks, improve security, do speech and image recognition and understand customer data. Jefferies says that private capital formation in AI has grown more than 50% to $4.25 billion in 2016 with nearly 1,900 startups in the space. (See Darktrace Raises $75M to Fight Cybercrime With AI and Eurobites: London AI Startup Thinks Big.)

The bank isn't the only one expressing skepticism about IBM's prospects in the AI market either. An article published in the MIT Technology Review last month illustrates the disillusionment many have had -- not in the tech that runs Watson, but in the hype around its potential, the expense of the system and the relative lack of progress so far.

Both the MIT article and Jefferies acknowledge Watson's technical superiority and its potential, but both are also skeptical it can live up to the hype -- or at least that by the time it does, others will have caught up.

— Sarah Thomas, Circle me on Google+ Follow me on TwitterVisit my LinkedIn profile, Director, Women in Comms

Page 1 / 2   >   >>
kq4ym 7/27/2017 | 8:27:37 AM
Re: IBM, Watson, cars, & AI I would guess IBM is fully aware of it's position in the market and how AI efforts of other competitors will be a challenge. I suspect IBM is looking for a bit different market, perhaps for companies willing to invest significantly in it's products that for now seem to demand a lot more to get up and running. The Watson effect bears watching to see how it grows or not over the coming years.
Joe Stanganelli 7/22/2017 | 3:15:19 PM
Re: IBM, Watson, cars, & AI @Phil: This goes to a brand methodology I've written on extensively and dubbed "Say-It-First Marketing". (link)

By inherently linking their brands to self-driving cars, Tesla and Google kind of "own" the space in the general public eye -- particularly when compared to the "legacy" automakers.
Phil_Britt 7/17/2017 | 8:16:49 AM
Re: IBM, Watson, cars, & AI As far as cars go, there are published reports that consumers are well aware of Telsa and Google, but not aware of the efforts of GM and Ford. So further advancements in that market could change the concept of the Big Three.

Joe Stanganelli 7/14/2017 | 3:27:34 PM
IBM, Watson, cars, & AI Part of that may be that companies working in the self-driving car sector seem to have gobbled up the lion's share of the AI talent.

Of course, IBM sure does love beating the Watson drum. I've seen enough IBM keynotes at conferences at this point -- all focusing on the history of Watson -- that I could deliver one myself.
HardenStance 7/14/2017 | 2:48:54 AM
Re: AI for Process Automation Re:"If the cost of the AI is not compensated by the additional headcount savings there is not much point."

That's no doubt true in the case of many applications but it's not the case in the security context.

No-one in their right mind is looking for significant headcount savings in the Security Operations Center nowadays.

Rather the expectation is that Watson et al should relieve much of the mundane pattern correlation work from SOC analysts, as well as assist them in deeper investigations.

That should free up SOC analysts to get onto the front foot in proactively investigating issues that pose the greatest threat of a high impact on the organization rather than just reactively fire-fighting as they often do now. 

There is tremendous value in that - and its a core part of where competition between vendors and between managed service providers is headed in the cyber security space.

danielcawrey 7/13/2017 | 7:25:00 PM
Re: AI for Process Automation One of the reasons Watson was marketed so much was based on margins. Big companies realize how much IBM charges, but most of us have no idea. My guess is Watson is a lot of money upfront without an organization really knowing what it is paying for. It's like buying a magical AI box that might not end up doing anything at all. 
James_B_Crawshaw 7/13/2017 | 6:13:32 PM
AI for Process Automation Aside from chat bots, one of the key applications of AI in the telco industry is in process automation. There are a number of companies offering Robotic Process Automation solutions today (essentially an alternative to BPO) although telco is a relatively late adopter, behind financial services. By combining AI with RPA the hope is that you can increase the level of automation in network operations even further.

You can read more about this in this article here:


There is of course a trade-off. If the cost of the AI is not compensated by the additional headcount savings there is not much point. To give you an idea, while a regular taskbot might cost you $5k per year, an AI-enhanced taskbot might cost $250k. That's clearly a lot more than BPO wages. Why such an exorbitant price tag for AI? Well, as the Jefferies report suggests it may well be all the associated consulting work which IBM Watson, Google DeepMind or third party SIs are charging to curate the data which feeds the AI engines. It's the old 'picks and shovels' principle at work in the AI gold rush. 
Sarah Thomas 7/13/2017 | 3:49:42 PM
Re: Watson For Cybersecurity: I Hear Nothing But Good Things Thanks for the feedback, Patrick! Great to hear from those in the field. Jeffries really wasn't doubting the superiority of Watson's tech, moreso issues around cost, training and competition. For something like security, companies are going to want the best of the best. It's too important to cut corners. It makes sense that IBM would have a stronghold here. I think they're a more trusted name than Amazon or Google.
HardenStance 7/13/2017 | 3:44:28 PM
Watson For Cybersecurity: I Hear Nothing But Good Things I can speak to what I've seen and heard around Watson For Cyber Security in the first half of this year.

The impression I've formed from three different customer sources is that Watson has few peers in terms of how well it is regarded for cybersecurity applicatons today.

Source #1. A couple of months ago, Ronan Murphy, CEO of Smarttech, a security services provider, positively purred with how well he has integrated Watson into his Security Operations Center (SOC) in Ireland. He stressed that whilst the output he saw from Watson was indeed disappointing initially, with feedback into the Watson developers improvements then came very rapidly.

Where did I meet Ronan? Alright, so it was at IBM Security's analyst event in London. IBM wheeled him out, Ronan duly delivered. But he did so with quite some verve. If he was acting it was one hell of a performance.

There's more.

Source #2: A couple of weeks before hearing Ronan's account I spent a day with some of Juniper's security folks. We talked about a lot of stuff – their own security strategy unsurprisingly – during which their team volunteered real enthusiasm for how Watson for cybersecurity is integrated into Juniper's security portfolio. The conversation around Watson was entirely spontaneous and unscripted. The enthusiasm of the Juniper folks – two of them – was one of real appreciation of Watson. It certainly wasn't your standard dutiful nod towards a so-so partner.

There's still more.  Source #3. Further back, in February I met with an exec at a tier one European telco who explained that they've been using Watson for some time and have been finding it very useful for cybersecurity. Not so much in terms of training Watson on network traffic but in terms of ploughing through tomes-worth of messages being exchanged by hackers on the deep web to gain insights into what's likely to be coming down the pipe. Very useful indeed, he told me.

So I doubt that the Jefferies analysis is totally off the mark. No doubt there's something in it. But there's a huge gap in the 'say:do' ratio afforded by so many AI-based solutions out there. Whilst pawing over recent infusions of funding for Deep Instinct, Crowdstrike, Cybereason and DarkTrace I received an email earlier this week from another start-up in their space. This company slammed one of the four aforementioned (by implication perhaps all of them) for having a ...wait for it..."legacy" solution.  

In the fluff-laden context of how AI plays in cybersecurity, I'm inclined to focus on the strong positives of something that from all the real accounts I'm hearing is delivering (very) well in the here and now. So I wouldn't be surprised if IBM's window of leadership is longer than the Jefferies folks are assuming, at least in the cybersecurity space.

I wouldn't under-estimate the value of Watson in terms of pulling through sales of other parts of IBM Security's portfolio, either. It's pretty much impossible to quantify - but that doesn't mean it isn't there.
bosco_pcs 7/13/2017 | 3:16:17 PM
Re: IBM response I used to know people working for IBM and work alongside with some on occasions but not anymore. So I could be wrong. 

It used to rely on the sink-or-swim strategy in developing human resources. Then there is the problem with silos. Its JIT pivoting doesn't help. Deep Blue may be the Jeopardy king but Amazon has been quietly hiring PhDs from MIT and CMU before they are in vogue. Finally, sometimes it has problems integrating acquired technologies. Softlayer may be in fashion but Netezza might not have gotten enough marketing support. Perhaps it is another silo problem. 

So who knows, Watson may end up to be another AS/400
Page 1 / 2   >   >>
Sign In