Generate SARs (as result of "fusing" and analysis of the data from license plates readers and the likes) Poptip will expand their ability to "fuse" various unstructured data like social feeds and phone conversation transcripts.
And what exactly is it that you think Palantir's products do? :)
Palantir is a private company whose product's development was incubated and developed at taxpayer expense following 9/11 (they essentially researched all of their technologies on the taxpayer dime on govt contracts, then took those technologies and created their product). It's an asshole company that doesn't give a shit about the constitution or human rights.
They also use Congressman Hunter as a paid shill to push their product all over the government as if its a panacea to every data integration issue. Its not.
The set of companies that benefit from taxpayer-funded research is enormous, and it does not follow from membership in that set that a company "doesn't give a shit about the Constitution or human rights".
Congrats on the acquisition but I'm still a little confused/surprised by this. The two (seemingly)disparate markets that Poptip and Palantir operate in have little cross-over. Is there a longer-term vision that I'm not seeing?
They do NLP analysis of social media feeds [Facebook, Twitter, etc] which is one of the products Palantir sells to Governments already to find "terrorists".
If the more accurate spying is used in the same manner as the less accurate type, then obviously more accurate is preferable. More often, the 'more accurate' version is accompanied by significant improvements in breadth and usability making the 'spying tool' responsible for much more spying behavior.
A good example is the history of tracking cars. Privacy concerns with police tracking were originally non-existent since the first versions were police literally following cars, so the number of potential targets was limited by the number of officers. Then came tracking beacons that were physically attached to the car, enabling a single officer / command center to track hundreds or thousands of vehicles. Undoubtedly more accurate and detailed, but mildly more invasive. The latest iteration of this is to simply track every car on the road via cameras with OCR or through RFID on toll passes. After you've collected a history of every vehicle in an area, you can pick-and-choose which ones are interesting. This is astronomically more detailed since you now have a history of where every baddie was at various points in time, but I think most would agree that it's much more invasive than ever before.
Still sounds like a net positive to me. Arguments against widespread surveillance are pretty much always founded on the assumption that the surveillance will be inaccurate. If it's only being expanded because it's accurate, the only problem is ideological, not practical.
> Arguments against widespread surveillance are pretty much always founded on the assumption that the surveillance will be inaccurate.
That assumes that there exists some objective, mutually agreed-upon measure of "accuracy".
For example, Glenn Greenwald has said that there is no definition of what constitutes "terrorism" that is not inherently based on race[0]. This isn't saying that the measures we take to prevent "terrorism" are racist; it's saying that the actual pursuit - the goal itself - is inherently racist, albeit implicitly.
The argument in support of that claim is too long to go into here[1], but suffice to say that, if you believe that the word "terrorism" is simply a euphemism for rationalizing racist behaviour or suppressing legitimate political dissent, then no, you're not going to be happy when people find more accurate ways to target those individuals.
[0] You may or may not agree with this premise personally, but that's not the point.
I really can't take seriously the idea that the goal of the NSA or any other part of the US government is any kind of racial cleansing or suppression. It seems far more plausible that using race as a judge of threatfulness is a result of not having anything more accurate to go on. In which case, better targeting would actually make the program less racist.
I generally fall into the same anti-paranoia camp as you seem to, as in "how bad can they really be?" But then I read articles like one posted here a while back, which I'm having trouble finding (anyone?), that described massive, tens-of-millions-of-dollars government research contracts ostensibly to "study the Arab Spring." The article talked about the discomfort of researchers, who were surprised to be pressured by the Pentagon into focusing their efforts on "potentially disruptive" American groups like the Tea Party or Occupy Wall Street.
Again, I can't find the article, but I remember my takeaway from it was that the Pentagon is actively pursuing research into how to identify, control, and suppress grass roots movements and protests in the US.
So, maybe we haven't seen evidence of suppression yet, but I suspect that's because there has yet to be an adequate trigger event in the US, or an "American spring" of sorts. When that happens, which I'm sure it will at some point, we will see the NSA putting their data to use against American citizens.
> I really can't take seriously the idea that the goal of the NSA or any other part of the US government is any kind of racial cleansing or suppression.
Perhaps not consciously, no, but we're talking about this in the same lens that people use when describing the war on drugs as "The New Jim Crow"[0], to assert that the war on (some) drugs is inherently a race- and class-driven fight, even if not done consciously by most parties involved.
> It seems far more plausible that using race as a judge of threatfulness is a result of not having anything more accurate to go on.
Again, you're assuming an uncontroversial, objective, and non-racist definition of "threatfulness". This is one of the few instances in which the phrase "begging the question" can be used correctly in an online discussion.
It might seem at first that such a definition would obvious, but that's actually very far from the case, as people like Greenwald have explained.
I don't want to try and explain this reasoning any further myself, not for lack of patience, but simply because Greenwald has already written about this this very eloquently and at great length, so anything I could write in this comment would be an ersatz rehashing of what he's said.
His essays on this are very easy to find online, so out of respect to his beliefs, I'd encourage you to read it "straight from the lion's mouth" rather than rely on whatever summary I could muster.
Because you believe the danger of government targeting dissident or other unpopular groups accurately outweighs the benefits that the proposed program adds to terrorist catching.
Terrorism is a low risk in the US, these programs seem to add little to just having investigators follow leads and intentionally infiltrate groups or paying informants or similar, and the US government has a long record of targeting groups that the powerful don't like.
You're missing the point. Even when the spying is used in "bad" ways, it's better for the spying to be accurate. For example, rather than randomly throwing students in jail for searching for the wrong library books, they'd only be jailing real dissidents. Of course it's wrong to do either, but if they're going to do it, everyone is better off if they do it right.
Objectively define "real dissidents" with a definition that can never inaccurate, as well as impossible to abuse by any and all that have power within such system.
Congratulations. I wish I'd visited Palantir in NYC; I've spent some time in Palo Alto and know people in DC (and downrange), but never met any of their NYC people. It's pretty amazing how I'd had the company pigeonholed as "government contractor" (which is a huge business in its own right), but the commercial stuff seems to be growing even faster.
Having worked at a startup that failed to find a customer base in the consumer world for the NLP software that analyzed social media, I'm not surprised the only application turns out to be defense industry.
There doesn't seem to be many uses for that type of NLP other than surveillance and brand monitoring.
Let's talk business here, because I find the numbers behind Palantir both astonishing and fascinating.
Their burn rate looks to be around $275m/yr based on employee head count and they've received ~$900m in funding:
- Seed - In-Q-Tel (~$2m)
- Debt Financing - $8.3m
- Series C - $35m
- Series D - $90m
- Secondary Market (unk amount, 137 investors)
- Series E - $50m
- Series F - $70m
- Series G - $56m
- Secondary Market (unk amount, 137 investors)
- Private Equity - $196.5m (Founders Fund)
- Private Equity - $390.4m (Feb 2014)
Their last round was almost $400m in Feb of 2014. That's almost as big as all other rounds combined. Their "valuation" (for whatever that's worth) based on the last investment round is ~$9billion.
As of August 2013 they were not yet profitable but estimates for 2013 revenue were around $450m and <$300m in 2012.
Other revenue estimates (they're all over the place honestly it seems like every time they give an interview it's a different number)
They've publicly announced several time they have no interest in being acquired or go public. Yet one of their investors claims on Quora that they'll probably go public in 2015 or 2016.
Some questions for people who understand this mess better than I.
- What's their exit strategy honestly look like? At $9b valuation, I can't imagine too many companies who might be interested in buying them.
- All the private equity investment tells me they're burning money faster than they can find investors, keeping it private keeps their financials private, which smells like trouble to me if they were to go public. Am I right in this analysis or wrong?
- Does dilution basically make everybody who has shares or options through 11 (eleven!) rounds of investment basically holding onto worthless paper? Even if they go public? A review of glassdoor on the company says that people are underpaid and salary capped, they must be holding on for a big payday somewhere. Are they going to end up disappointed?
Anything else that might be interesting to discuss? Their fundraising seems really crazy and unusual from anything I've seen before.
Here's a longform history [1] of the company which shows they spend over $1m/yr on lobbying (IBM spent $6m, Raytheon $7.5m, Lockheed Martin, $14.5m, Boeing, $15m). To put that in percentage terms, if Palantir is worth $9b
Palantir spends .012% of their value on lobbying
IBM spends .003%
Raytheon .026%
Lockheed Martin .027%
Boeing .017%
This article also lists 16 distinct financing events, bringing dilution of early investor shares even more to mind.
Some people want to build actual sustainable businesses. Why are you assuming they have a secret exit strategy when they've stated they're not interested in exiting?
If you switched to the bull side, you would basically have the opposite article and it would likely have as much merit or more.
They are obviously going to IPO at some point. With the hassle of IPOing, the nature of their business and the easy access to private capital, they'd be dumb to IPO sooner than later.
The secondary market for shares is extremely active and employees can sell a limited number. So almost anyone has probably cashed out as much as desired at this point.
> The secondary market for shares is extremely active and employees can sell a limited number. So almost anyone has probably cashed out as much as desired at this point.
That's an interesting point. Can you go into more about how that might work? I had assumed they were just internally selling shares to other employees, but could it be they're selling shares to external buyers? I'm not clear on how that works.
Employees typically would not buy other employees' (or investors') shares.
Basically buyers and sellers need to find each other, agree on terms, send it to the company for right of refusal and then transact (and pay a large legal fee).
Different companies have different degrees of willingness to entertain the secondary market. With a stated desire to remain private for an extended period, Palantir may be more amenable than most to secondary.
I would imagine they are probably operating like Amazon: spending all their money on growth, mainly on acquiring talent. If they get big enough, they will be able to close huge deals with the government. Maybe the investors see this as a safe bet to hedge their portfolio, kinda like a value investment. My 2 cents..
Palantir is a govt contractor. They need to be big to be heard and be in charge. They want to be Lockheed, Raytheon and Boeing. They can't be big without hiring tons of employees so they are hiring tons of employees.
Exactly. Government contracts don't require the number of people to get the job done. They require the number of people they SPECIFY are required (whether or not that number is correct) to get the job done. Translation: Government contracts are by their very nature overstaffed as fuck.
One of the things I've been reading is that quite a bit of their business seems to be shifting to their financial product from their government product. So I'm not sure how much their government business is still a major part of the company.
I don't think it's exactly that, but vaguely I get the same feeling. It's like if they are in the big league. Not being public IMHO it's a way of keeping a discreet profile more than anything else.
Raytheon, Lockheed, and Boeing all hire ex-Pentagon/Defense officials as well so the official "lobbying #s" are probably low for all of them. ;)
I think Palantir's goal is to burn as hot as it can for as long as it can, then coast on the residuals of its government contracts. [e.g. Make the kind of "open bid contacts" that require automatically fails anyone not Palantir]
Plus, it's not that helpful to look at lobbying spend as a percentage of revenue. It takes a certain amount to be in the game. Then you get some decreasing returns.
Easy but wildly incorrect due to the exponential factor.
For example, if you assume a 15% dilution over 11 fundraising rounds, then it's 0.0167% which is +$1.5MM at a $9B valuation.
at 20%: +730k
at 25%: +$380k
I picked 15% because it seemed a good middle ground. Seed rounds are often <15%, Series A/B is around 20-30%, and later rounds target <5-10%. Because Palantir did so many rounds, I would guess only a few were significantly (> 15%) dilutive.
Palantir is pretty mysterious to me. Frankly, I don't see the demand for their software being high enough to warrant their size and number of employees.
I guess I'll just have to wait and see what happens.
