IZEA today just announced similar platform. Read This:
Veritone CEO Steelberg Ponders A.I. and His Stock Price
Veritone, a kind of service bureau for artificial intelligence analysis of unstructured data, has seen its shares soar since a mid-August article in Barron's. CEO Chad Steelberg spent an hour and a half at Barron's offices explaining the technology and the business model, and why his stock is not overvalued.
ByTiernan Ray Oct. 2, 2017 9:05 a.m. ET
On Friday, I met with Chad Steelberg, the CEO and co-founder of software maker Veritone (VERI), which went public in May. Steelberg was at the end of three weeks on the road meeting with investors. He had just come from three days in London. Swinging through New York, it would be worth his meeting with me, his people thought, and they extended the offer, and I accepted. He was accompanied by his head of marketing, John Ward.
Veritone stock has soared since I wrote about it in mid-August for Barron's print magazine, and it behooved me to get an update from Steelberg.
When I wrote about the company, its stock was 40% from its IPO price. I wrote about it as a company leading the charge in artificial intelligence whose beaten-down shares were worth a look. Veritone is a young company with minimal revenue. I was certainly surprised about the rapid rise in the stock.
Before our issue of August 21st came out, the price had closed that Friday at $7.87. The Monday the article appeared, shares opened at $8.45, and never looked back, rising 49% that day to close at $11.75.
This past Friday, the stock was up 23% at 45.45, rebounding from two massive sell-offs the prior two days, after hitting an all-time high of $74.92 on Wednesday.
Over the past several weeks, I've gotten numerous inquiries from readers ranging from whether the stock is overvalued to whether Veritone is a fraud to whether I am secretly an insider profiting off the stock. For the record, I have no personal connection to Veritone. Both times I've talked with the company, it was at the company's suggestion.
I make no profit in any way from the movement of Veritone stock, up or down. I'm bound by a rule at Dow Jones & Co. against profiting from what I write about. All us reporters are asked about this yearly by the company and have to attest in writing that we have obeyed the rule, on pain of losing our jobs. Any suggestion otherwise is a fantasy, but people will cast all kinds of aspersions, because in an Internet age, it costs people nothing to cast aspersions.
I understand Veritone shares have been hard to borrow, making it difficult to short. The stock did take a dive on Wednesday and on Thursday, following a negative tweet from noted research outfit Citron. It was helped Friday by a note from Roth Capital’s Brian Alger, who attended meetings with Steelberg in London, and who assigned a $62 price target.
On Friday, as his shares soared once again, I asked Steelberg if his stock is overvalued. “No,” he said. As to why not, he said “I think we value the company in ways different from Wall Street. There is a hockey stick for our business that is still to come” as its success becomes apparent. He said the hockey stick happens next year.
“Artificial intelligence is a new category for investors,” said Steelberg. “People still trying to grasp the impact of it, and of what we’re doing.”
To explain that impact, Steelberg and I talked for about an hour about how he sees the business evolving. Steelberg is not only conversant in his own business, he shows an interesting grasp of things going on in tech and in A.I. in particular. He spoke at length about how Alphabet’s (GOOGL) Google is innovating in its data center, and about other new young companies, such as San Francisco startup Brain Chip.
He began his presentation with a video of a Nascar race, from the cockpit camera. The ordinary approach to analyzing such a video, like other computer analysis of such “unstructured data,” doesn't yield much information about what's going on. Using a bunch of “engines,” algorithms tuned for a given type of analysis, it's possible to glean a lot more from every video frame. Things such as when a corporate logo pops up in the video, or the reaction of the crowd seen from the cockpit. Veritone acts as service bureau, of sorts, for computer analysis. Nascar pays Veritone money on an hourly basis for such analysis, depending on the value of things identified in the video, such as corporate logos.
With the help of data scientists Steelberg has hired, such as Wolf Khan, a former Lockheed and NASA scientist, Veritone is intent on finding patterns in unstructured data where “the question is not yet known,” as Steelberg puts it. “Can you unify all problems in compute?” is how Steelberg poses the A.I. question. Can one achieve not just “artificial narrow intelligence,” or ANI, as he calls today's A.I., but “cognition”?
“You can't today, and one reason is, you need to know all the problems in advance,” and having to define the questions to be asked limits what one can search for in unstructured data. By using more engines, or algos, Veritone can look in more ways at the data and come up with results before even being able to formulate a question. These are the “known unknowns,” if you will.
For Steelberg, the limits of Google and others in A.I. has always been the relative paucity of their engines or algos, even though they have vast amounts of data. The data is piling up for everyone, like Nascar, as events generate reams of data, and so the challenge is to bulk up on engines.
“The limitations have prompted us to look at the second derivative,” he says. “Could we understand the idiosyncrasies of those engines themselves?
“Each engine has a training data set and a language and a lexicon,” he observes. By learning about all the engines for natural language or another domain, “we can normalize” across their capabilities. “We can get a confidence score” for how good a given engine is.
His company has 120 engines spanning 14 categories of analysis including natural language processing, text processing, identifying objects in video, etc. He and his data scientists have been assembling a database of engines in development around the world, and they believe there are currently 5,500 such engines that are viable, locked inside various companies.
He is pursuing those engines via two tracks. One is to acquire companies. Veritone has communicated to investors it will be acquisitive, something that Roth’s Alger believes will take the company from what would be just $39 million in revenue next year, on its own, to perhaps $100 million. (Veritone is projected to have just $17.7 million in revenue this year, according to the average of four estimates on FactSet.)
In addition, however, Steelberg sees the company’s service bureau as a hub for third-party engines that is analogous to Salesforce’s (CRM) Force.net development platform. Force lets various application makers of cloud software sell to Salesforce customers and Salesforce makes money running the platform. Similarly, Veritone’s “aiWare” software, running in the data center, is a kind of operating system. Both third-party engine developers, and developers of higher-level applications, can plug into aiWare, the way device drivers and user applications run on an operating system. Just like Salesforce, Veritone would take a 30% cut from what those app vendors make.
As more app vendors join the aiWare platform, hopes Steelberg, more customers find the platform attractive and sign up to bring their data. CBS (CBS) did a trial with Veritone to see if A.I. could improve the ad targeting of its Huston radio station over the course of 30 days. After being satisfied with the results, the broadcaster has signed a $1.7 million deal with Veritone to cover all its stations, nationwide.
As more companies like CBS sign up to use the aiWare platform, said Steelberg, more and more valuable content becomes combined in an anonymized fashion, so that the whole platform learns about how to best analyze unstructured data, and grows in value, a kind of network effect that grows and grows, he hopes.
All that is to come. To sell customers and partners and app developers and engine makers is a lot of work ahead for this new young company. He's confident, of course. “Outside of Google, we have the best data scientists on the planet,” is one of the many things that supports such confidence. Using various cloud services such as Microsoft’s (MSFT) Azure is an important go-to-market and channel strategy. “They just want us to put as many engines as possible on Azure,” he says with a smile of Microsoft’s attitude. “They just love us grinding away on Azure” because it boosts Azure’s compute cycles, and it is a battle for compute cycles with Google and with Amazon’s (AMZN) AWS.
Before we ended our discussion, I read off to Steelberg a litany of the risks raised by skeptics. Veritone is just a rehash, some claim, of the advertising business he and his brother created with dMarc, which was sold to Google to ignite online radio ad sales, for $1.24 billion in 2006. Little ever came of a prior businesses, his first, AdForce, which he sold to Internet conglomerate CMGI in 1999 for half a billion dollars. The patents they have are from that legacy ad business, not really for A.I., per se. The company was a “going concern” at the time of the IPO, and had disclosed in its IPO prospectus “material weakness,” and had restated its financials.
Steelberg’s response is that “all that stuff is basically grabbed from the safe harbor section of our prospectus,” and is therefore not original research and not original insight on the part of critics.
“It is true that we were a going concern at the time, but we have resolved that with the proceeds of our offering.”
“It is true that most of our original revenue has been from the buy-side of media,” and not A.I., but that is changing as the platform becomes mature, he says. “SaaS is now the fastest-growing part of our business.”
“We think about the business a little more long-term than most of Wall Street, about the long-term metrics as we have more touch points with each customer” to validate things. “I know this from having built ecosystems in the past; we spent months building AdForce before there was any revenue coming from that platform.” Hence, the notion of a knee in the curve coming in 2018.
“I will let my successes speak for themselves,” said Steelberg, his voice becoming quieter, and offering almost a shrug of the shoulders. “AdForce was my first company, and your first company, when you are in your 20s, is not really a company as much as it's an extension of yourself.”
I asked him, in closing, if he wished to reply to Citron’s tweet. “You know what, I'm not going to comment on a tweet,” Steelberg replied.
Which seems to me an intelligent response. There's little to say about 140 characters worth of aspersions, thrown together with little thought and little to justify it. If I were him, I wouldn't dignify it ether. Citron have produced voluminous reports in past that substantiated their points, so perhaps at some point they'll produce something on Veritone that merits a response.
In the meantime, this is a young company with modest revenue. A lot of the valuation depends on how much confidence one puts in the future success of what Steelberg has outlined. My sense is that success is far from guaranteed, but that the software and the technology assets, and the brain power to do something with it are real. People who suspect the company is a fraud likely do not know what they are talking about. If you've bought since the $8 level, you may want to take some profits. If you've bought closer to recent levels, how much you want to hold on is relative to your risk tolerance for what is a promising but a speculative venture.
