>"well-established reputation for bias and shoddy journalistic practices - moreso than other news outlets."
How about a citation for that assertion? Reputation among whom?
Buzzfeed News is rated in the category "Most Reliable for News" and "Reliable for news, but high in analysis / opinion content" similar to WSJ and, CNN, and FT.
I wouldn't say so, maybe in right wing circles - the perception of it is a left-leaning site or just a silly top-ten list site, but my hack friends say that they actually make pretty good and honest content. Obviously nobody is perfect.
Media Bias Fact Check is completely unscientific, with a highly subjective methodology that leads to patently ridiculous aggregate ratings about high integrity, highly trusted sources.
Your other comment in no way refutes or undermines the journalistic integrity of Buzzfeed, which is high.
To give an example how this kind of database and software can be used, based on their example:
- Someone "Alex" is convicted, a data point is created
- "X" has a scar on their left leg and a tattoo on their right ankle of a skull, this is registered in their data, along with any other info (height, probable weight, ethnicity, age, what they look like etc.)
Later, a citizen reports a crime, doesn't really see the person but noticed the start of a tattoo on the perp's right ankle under a sock. This info, along with vague height info, etc. could then be used to define Alex as a suspect. Powerful, data comprehension.
Potentially controversial part: not everyone has a data point, and since the article states it, there are probably more data points on black and brown communities. So data comprehension and links found and going to be found in this already-existing pool. A possible solution to this is creating data points for everyone, but that has its own controversies (whether it's a government or a private entity).
To play Devil's advocate, Palantir gets a lot of flack for licensing their software to governmental institutions, e.g. ICE, the US army, and for being affiliated to Peter Thiel (see the other comment). Unlike Thiel, Alex Karp (Palantir CEO) is left leaning. Palantir has worked with numerous administrations (Bush, Obama). Their view is that they don't see or pick any sides, but the morality of that is up for everyone to decide, I believe if you don't pick a side, one is often picked for you.
I don't think the devil needs any more advocates, tbh, and I kind of wish people would stop refusing to take responsibility for the arguments they put forth by hedging in that way.
> Powerful, data comprehension
Correlation, maybe -- not the same as comprehension.
> A possible solution to this is creating data points for everyone, but that has its own controversies
Or you know, creating fewer data points -- maybe even none -- in general. Why is this something that needs to be software-driven? Why increase automation? What is the evidence that policing and criminal investigation is more effective (and by what standards!) when using Palantir's tech?
I question the assumption that increased automation and digitization of every process is necessarily an improvement, or useful, and I feel like this line of argument takes that as an assumption.
> Their view is that they don't see or pick any sides, but the morality of that is up for everyone to decide, I believe if you don't pick a side, one is often picked for you.
"Not seeing sides", or not picking them, is taking a position though. Declaring neutrality doesn't remove the concrete effect of your actions, or change the fact of your material support for a given model of policing or government. As Can Duruk argues convincingly in his latest newsletter[1] one of the best examples of that is low voter turnout in US elections: that benefits one side pretty clearly, so to declare you won't take sides and won't vote is a practical and real show of support for the status quo (to take just one example).
> Correlation, maybe -- not the same as comprehension.
The problem with this argument is that, although it is right (I agree!) it will not carry the debate.
_The_ problem here is that there is a large group of political moderates who trust authority. Arguing that the authorities will misinterpret and/or abuse statistical evidence will not sway their opinions.
It is difficult to convince that group to limit the power of government. The best argument I came up with was an appeal to a fundamental right to privacy; but that has pretty much all but evaporated in today's world. Privacy is too expensive and might be politically untenable. That argument started to collapse in the hysteria after 9/11 2001.
How much power the government has is certainly one line of argument. I'm more interested in how we hold the government accountable for the exercise of the power it has.
In general, I don't think limiting the power or size of the government is the solution, although I'm all in favour of shrinking certain parts of it in very specific ways: You can make your prison management arm smaller by privatising jails and prisons, but you haven't really fixed anything if you do that; on the other hand, I would be delighted to see defence and policing budgets the world over diverted into infrastructure, schools, and housing, and there are many very good examples of how those can be managed well by both central or local government.
If the government is unaccountable, then anyway it's useless to talk about its size, it can simply accrue more power. I think America is especially (although not uniquely) feeling the pain of lack of accountability at the moment, not least because the electorate as a whole is increasingly misrepresented by the outcome of elections, there is no culture of building political consensus, and as a result every vote becomes a zero-sum contest.
It doesn't matter if it is more effective, but it is cheaper than employing people connecting dots.
Also the generation currently in politics would feel left behind if they didn't employ such pattern matching systems, even if there is no evidence of crime reduction.
It is part of the quantification of everything. You don't just get a number, you get several thousands and the judgement on you is calculated on the fly.
I think people responsible here are still the ones that decided to buy it. They are responsible the market exists. Also not a mystery since those people are mostly evaluated by how many people they put behind bars.
> I think people responsible here are still the ones that decided to buy it. They are responsible the market exists. Also not a mystery since those people are mostly evaluated by how many people they put behind bars.
That's letting the people selling (and producing) the snake-oil off the hook, which I'm not willing to do that easily.
If you know what you're making is at best making lemonade out of statistical noise, or at worst actively harming people and society, and it's being marketed as an intelligent system that gets results, well then you bear a big chunk of that moral responsibility.
>"Not seeing sides", or not picking them, is taking a position though.
This has always been a simplistic take on disagreement, action, and belief. It deems agnosticism untenable and inherently precludes the possibility of nihilism.
It would be extraordinary - and flatly in tension with common sense - to believe/claim that agnosticism and/or nihilism were not live options for some domain.
E.g., when you have an oppressive status quo, and the options are to resist it and argue for change, or accept the ways things are, and thereby support them, then there is no agnostic position. This isn't a game: sometimes you simply don't get to sit it out.
> and inherently precludes the possibility of nihilism
I am quite happy to preclude nihilism entirely.
> It would be extraordinary - and flatly in tension with common sense - to believe/claim that agnosticism and/or nihilism were not live options for some domain.
I could give a lot of answers to this, but I think the simplest one is that what you regard as "common sense" is not necessarily the same as what I do, so I don't find that a very convincing counterargument.
The right tack to defending your position is not to double-down and say nihilism is impossible. How could a nihilistic take in any domain be impossible?
We aren't talking about whether agnosticism and/or nihilism are unreasonable. We're talking about whether they are impossible tout court.
Ah you see, the mistake you make is thinking I am interested in defending my position to the unreasonable. You can take any position you like, I don't have to entertain that position, or even get dragged into a discussion about it.
It's an old tactic the one you're taking, which is to carry off your arguing partner into the weeds of oratorical flourishes and pseudorationalist blow-by-blow point-making. One can make any point they like "tout court", that doesn't mean I am required (or indeed that any useful purpose is served) to spend any mental energy formulating a substantive reply.
I'm not sure what you are arguing. I'm arguing that political nihilism is a possible position, and that being a political nihilist is different than being a supporter/opponent of the status quo.
Your argument seems to be that there are only ever two possible political positions: pro-status-quo or anti-status-quo. Is it really that simple/non-complex?
> I don't think the devil needs any more advocates, tbh, and I kind of wish people would stop refusing to take responsibility for the arguments they put forth by hedging in that way.
my kneejerk reaction is that I kind of hate this mindset. people should be allowed to consider whether there is any internal logic to the opposing position without actually committing to it. at the same time, I have come to agree that devil's arguments in favor of an already dominant position can be harmful in broader discourse. if you already have a representative sample of american voices, for instance, there's no need to make the american status quo's argument for it. but this conversation is taking place on hacker news, where I daresay the consensus is opposed to palantir and other companies working with law enforcement. in this case, the devil's argument is actually the minority position, and I do feel it adds to the discussion. when you spend most of your time in filter bubbles that cater to positions you already hold, it's easy to end up only engaging with strawman versions of the outside world.
> but this conversation is taking place on hacker news, where I daresay the consensus is opposed to palantir and other companies working with law enforcement
I don't share your confidence of that assessment, but it would warm my heart to think so.
> people should be allowed to consider whether there is any internal logic to the opposing position without actually committing to it.
I'm gonna stand by this one, I think, and disagree with you, although I think I see where you're coming from.
What I see a lot more than testing for internal logic of a position is people advancing their sole, or only substantial discussion point, with the "devil's advocate, but" position. If a consistency test is what you're after, you can attempt to perform an explication of someone else's position without advocating that position yourself, and without accidentally advocating it by presenting it unchallenged.
If I can omit the "devil's advocate" disclaimer and it still reads as a position earnestly advanced by the writer, then they are not playing at anything, they are making a real argument.
In warfare one of the common aphorisms is "know thy enemy" and I think this is very much lost on many people these days. So often I see people fighting against some thing without any understanding of how that thing works or why its that way. So they end up fighting some badly drawn caricature, and many of their arguments end up being unsound as they reflect the caricature and not the opponent.
This is especially disappointing in politics, where at times in the past political dealing was centered around achieving a consensus, these days it seems to be based more around wielding power and whatever tricks are needed to get the thing to stick. Unsurprisingly since the opposite side has nothing to get them to stick to the plan, if they get into power they try to wipe it out.
So, I think understanding your enemy puts you in a position to make strategic choices to lock them into a path that achieves your goals and makes it hard for them to block them. So, with all the caveats about looking into abysses and fighting monsters, we should understand our enemies.
> A possible solution to this is creating data points for everyone
A better solution is to use better statistics.
If you think you need to collect redundant extra data about innocents just to improve inferences using the data you have, you're doing statistics wrong.
"The Prosecutor’s Fallacy is most often associated with miscarriages of justice. It’s when the probability of innocence given the evidence is wrongly assumed to equal an infinitesimally small probability that that evidence would occur if the defendant was innocent. Consequently, highly improbable innocent explanations have led to the assumption of guilt,"
It's essentially when information that was used to identify the suspect is presented as confirmation of their guilt.
> The witness described a tall, white, black-haired man with a tattoo on his right ankle. Only .001% of the population match this description, so there is only a .001% chance that the suspect is innocent.
Of course, this ignores the fact that they compared the description against a database of one million people, among whom you would expect to find 10 coincidental matches.
For the same reason more data is required to select a model than to validate a model in machine learning.
There is no problem that the information on past offenders is searchable.
There is a bit of a problem that information on non-offenders is sucked in as well, but even that is moderately ok as long as anyone can see what the system has on them (currently it's not the case), who accessed it and why.
Perhaps the access data should be released with a delay to protect current investigations, or sealed altogether for a very good reason with appropriate authority.
I don’t judge you for taking the job, wanting to be paid well, maybe believing in some idealistic variation of the company’s mission that can be twisted to purport to help people.
It’s not about you or your reasons. It’s about the destructive force Palantir represents if it can go unchallenged while amassing deep connections to public sources of revenue in municipal police forces and national defense or intelligence agencies, and ultimately branching into corporate security and espionage and facilitate employers spying on employees.
Please quit helping the leadership of Palantir. It’s an awful, awful thing for the world - much more morally awful than rank and file corporate misbehavior. Palantir is a difference in kind - enabling & facilitating a vastly more oppressive future and stamping out pathways to achieve privacy and individual liberty in a big data world.
Be the person who can proudly tell your friends and family, your children, one day when looking back that you had the courage to quit and you put society’s welfare above your own potential path to luxury.
To everyone else: don’t consider ever working for Palantir. Don’t be lured by the pay or fooled by a false depiction of the company mission. It’s not helping people. It’s not a civic duty. It’s not a positive force for the world.
One time I interviewed with Palantir. First time I used Zoom. This was yeeears before it became an issue that they don't encrypt communications and I brought it up that I thought it was weird they're relying on third-parties for something that felt... vague but also security related?
In order to suss out exactly what apprehensions I was having about their corporation (forgive me, I knew they were operated by Peter Thiel and this and all that but I was willing giving to give them a fair shake since I was a lot younger and a lot more naive), basically what I asked was a simple question, "what exactly does Palantir do? Because I have read your marketing and releases and everything but I'd like to hear it straight from the horse's mouth with an employee."
"We leverage data."
Oh I see! Well it's 12:00PM and I see we're out of time, thank you so much for taking the time to speak with me and looking forward to next steps.
I never hear back and spent the next few weeks horrified by the interaction.
Addendum: What got me the interview was the fact that I studied under one of the chief scientists behind PredPol. I understood how the technology worked and the basic logistics of how to implement it. Regardless of whether I believed in it, I knew how it worked. I think that's what they were after.
34 comments
[ 0.23 ms ] story [ 76.0 ms ] threadHow about a citation for that assertion? Reputation among whom?
Buzzfeed News is rated in the category "Most Reliable for News" and "Reliable for news, but high in analysis / opinion content" similar to WSJ and, CNN, and FT.
[1]http://gatewayjr.org/how-a-popular-media-bias-chart-determin...
No, not really:
https://mediabiasfactcheck.com/buzzfeed/
https://mediabiasfactcheck.com/washington-post/
https://mediabiasfactcheck.com/financial-times/
I left CNN out because they are nearly as bad as Fox these days
Thanks for the links, I'll keep these around next time I have to bring this up again.
But still have a nice Day.
Your other comment in no way refutes or undermines the journalistic integrity of Buzzfeed, which is high.
- Someone "Alex" is convicted, a data point is created - "X" has a scar on their left leg and a tattoo on their right ankle of a skull, this is registered in their data, along with any other info (height, probable weight, ethnicity, age, what they look like etc.)
Later, a citizen reports a crime, doesn't really see the person but noticed the start of a tattoo on the perp's right ankle under a sock. This info, along with vague height info, etc. could then be used to define Alex as a suspect. Powerful, data comprehension.
Potentially controversial part: not everyone has a data point, and since the article states it, there are probably more data points on black and brown communities. So data comprehension and links found and going to be found in this already-existing pool. A possible solution to this is creating data points for everyone, but that has its own controversies (whether it's a government or a private entity).
To play Devil's advocate, Palantir gets a lot of flack for licensing their software to governmental institutions, e.g. ICE, the US army, and for being affiliated to Peter Thiel (see the other comment). Unlike Thiel, Alex Karp (Palantir CEO) is left leaning. Palantir has worked with numerous administrations (Bush, Obama). Their view is that they don't see or pick any sides, but the morality of that is up for everyone to decide, I believe if you don't pick a side, one is often picked for you.
> Powerful, data comprehension
Correlation, maybe -- not the same as comprehension.
> A possible solution to this is creating data points for everyone, but that has its own controversies
Or you know, creating fewer data points -- maybe even none -- in general. Why is this something that needs to be software-driven? Why increase automation? What is the evidence that policing and criminal investigation is more effective (and by what standards!) when using Palantir's tech?
I question the assumption that increased automation and digitization of every process is necessarily an improvement, or useful, and I feel like this line of argument takes that as an assumption.
> Their view is that they don't see or pick any sides, but the morality of that is up for everyone to decide, I believe if you don't pick a side, one is often picked for you.
"Not seeing sides", or not picking them, is taking a position though. Declaring neutrality doesn't remove the concrete effect of your actions, or change the fact of your material support for a given model of policing or government. As Can Duruk argues convincingly in his latest newsletter[1] one of the best examples of that is low voter turnout in US elections: that benefits one side pretty clearly, so to declare you won't take sides and won't vote is a practical and real show of support for the status quo (to take just one example).
[1]: https://themargins.substack.com/p/all-politics-is-bad-says-c...
The problem with this argument is that, although it is right (I agree!) it will not carry the debate.
_The_ problem here is that there is a large group of political moderates who trust authority. Arguing that the authorities will misinterpret and/or abuse statistical evidence will not sway their opinions.
It is difficult to convince that group to limit the power of government. The best argument I came up with was an appeal to a fundamental right to privacy; but that has pretty much all but evaporated in today's world. Privacy is too expensive and might be politically untenable. That argument started to collapse in the hysteria after 9/11 2001.
In general, I don't think limiting the power or size of the government is the solution, although I'm all in favour of shrinking certain parts of it in very specific ways: You can make your prison management arm smaller by privatising jails and prisons, but you haven't really fixed anything if you do that; on the other hand, I would be delighted to see defence and policing budgets the world over diverted into infrastructure, schools, and housing, and there are many very good examples of how those can be managed well by both central or local government.
If the government is unaccountable, then anyway it's useless to talk about its size, it can simply accrue more power. I think America is especially (although not uniquely) feeling the pain of lack of accountability at the moment, not least because the electorate as a whole is increasingly misrepresented by the outcome of elections, there is no culture of building political consensus, and as a result every vote becomes a zero-sum contest.
It does not have to be like this.
Also the generation currently in politics would feel left behind if they didn't employ such pattern matching systems, even if there is no evidence of crime reduction.
It is part of the quantification of everything. You don't just get a number, you get several thousands and the judgement on you is calculated on the fly.
I think people responsible here are still the ones that decided to buy it. They are responsible the market exists. Also not a mystery since those people are mostly evaluated by how many people they put behind bars.
That's letting the people selling (and producing) the snake-oil off the hook, which I'm not willing to do that easily.
If you know what you're making is at best making lemonade out of statistical noise, or at worst actively harming people and society, and it's being marketed as an intelligent system that gets results, well then you bear a big chunk of that moral responsibility.
This has always been a simplistic take on disagreement, action, and belief. It deems agnosticism untenable and inherently precludes the possibility of nihilism.
It would be extraordinary - and flatly in tension with common sense - to believe/claim that agnosticism and/or nihilism were not live options for some domain.
Agnosticism is untenable in some scenarios, sure.
E.g., when you have an oppressive status quo, and the options are to resist it and argue for change, or accept the ways things are, and thereby support them, then there is no agnostic position. This isn't a game: sometimes you simply don't get to sit it out.
> and inherently precludes the possibility of nihilism
I am quite happy to preclude nihilism entirely.
> It would be extraordinary - and flatly in tension with common sense - to believe/claim that agnosticism and/or nihilism were not live options for some domain.
I could give a lot of answers to this, but I think the simplest one is that what you regard as "common sense" is not necessarily the same as what I do, so I don't find that a very convincing counterargument.
We aren't talking about whether agnosticism and/or nihilism are unreasonable. We're talking about whether they are impossible tout court.
It's an old tactic the one you're taking, which is to carry off your arguing partner into the weeds of oratorical flourishes and pseudorationalist blow-by-blow point-making. One can make any point they like "tout court", that doesn't mean I am required (or indeed that any useful purpose is served) to spend any mental energy formulating a substantive reply.
Your argument seems to be that there are only ever two possible political positions: pro-status-quo or anti-status-quo. Is it really that simple/non-complex?
my kneejerk reaction is that I kind of hate this mindset. people should be allowed to consider whether there is any internal logic to the opposing position without actually committing to it. at the same time, I have come to agree that devil's arguments in favor of an already dominant position can be harmful in broader discourse. if you already have a representative sample of american voices, for instance, there's no need to make the american status quo's argument for it. but this conversation is taking place on hacker news, where I daresay the consensus is opposed to palantir and other companies working with law enforcement. in this case, the devil's argument is actually the minority position, and I do feel it adds to the discussion. when you spend most of your time in filter bubbles that cater to positions you already hold, it's easy to end up only engaging with strawman versions of the outside world.
I don't share your confidence of that assessment, but it would warm my heart to think so.
> people should be allowed to consider whether there is any internal logic to the opposing position without actually committing to it.
I'm gonna stand by this one, I think, and disagree with you, although I think I see where you're coming from.
What I see a lot more than testing for internal logic of a position is people advancing their sole, or only substantial discussion point, with the "devil's advocate, but" position. If a consistency test is what you're after, you can attempt to perform an explication of someone else's position without advocating that position yourself, and without accidentally advocating it by presenting it unchallenged.
If I can omit the "devil's advocate" disclaimer and it still reads as a position earnestly advanced by the writer, then they are not playing at anything, they are making a real argument.
This is especially disappointing in politics, where at times in the past political dealing was centered around achieving a consensus, these days it seems to be based more around wielding power and whatever tricks are needed to get the thing to stick. Unsurprisingly since the opposite side has nothing to get them to stick to the plan, if they get into power they try to wipe it out.
So, I think understanding your enemy puts you in a position to make strategic choices to lock them into a path that achieves your goals and makes it hard for them to block them. So, with all the caveats about looking into abysses and fighting monsters, we should understand our enemies.
A better solution is to use better statistics.
If you think you need to collect redundant extra data about innocents just to improve inferences using the data you have, you're doing statistics wrong.
This has come up with DNA databases too.
Prosecutor's Fallacy:
https://en.wikipedia.org/wiki/Prosecutor%27s_fallacy
https://www.cebm.ox.ac.uk/news/views/the-prosecutors-fallacy
"The Prosecutor’s Fallacy is most often associated with miscarriages of justice. It’s when the probability of innocence given the evidence is wrongly assumed to equal an infinitesimally small probability that that evidence would occur if the defendant was innocent. Consequently, highly improbable innocent explanations have led to the assumption of guilt,"
> The witness described a tall, white, black-haired man with a tattoo on his right ankle. Only .001% of the population match this description, so there is only a .001% chance that the suspect is innocent.
Of course, this ignores the fact that they compared the description against a database of one million people, among whom you would expect to find 10 coincidental matches.
For the same reason more data is required to select a model than to validate a model in machine learning.
There is a bit of a problem that information on non-offenders is sucked in as well, but even that is moderately ok as long as anyone can see what the system has on them (currently it's not the case), who accessed it and why.
Perhaps the access data should be released with a delay to protect current investigations, or sealed altogether for a very good reason with appropriate authority.
I don’t judge you for taking the job, wanting to be paid well, maybe believing in some idealistic variation of the company’s mission that can be twisted to purport to help people.
It’s not about you or your reasons. It’s about the destructive force Palantir represents if it can go unchallenged while amassing deep connections to public sources of revenue in municipal police forces and national defense or intelligence agencies, and ultimately branching into corporate security and espionage and facilitate employers spying on employees.
Please quit helping the leadership of Palantir. It’s an awful, awful thing for the world - much more morally awful than rank and file corporate misbehavior. Palantir is a difference in kind - enabling & facilitating a vastly more oppressive future and stamping out pathways to achieve privacy and individual liberty in a big data world.
Be the person who can proudly tell your friends and family, your children, one day when looking back that you had the courage to quit and you put society’s welfare above your own potential path to luxury.
To everyone else: don’t consider ever working for Palantir. Don’t be lured by the pay or fooled by a false depiction of the company mission. It’s not helping people. It’s not a civic duty. It’s not a positive force for the world.
In order to suss out exactly what apprehensions I was having about their corporation (forgive me, I knew they were operated by Peter Thiel and this and all that but I was willing giving to give them a fair shake since I was a lot younger and a lot more naive), basically what I asked was a simple question, "what exactly does Palantir do? Because I have read your marketing and releases and everything but I'd like to hear it straight from the horse's mouth with an employee."
"We leverage data."
Oh I see! Well it's 12:00PM and I see we're out of time, thank you so much for taking the time to speak with me and looking forward to next steps.
I never hear back and spent the next few weeks horrified by the interaction.
Addendum: What got me the interview was the fact that I studied under one of the chief scientists behind PredPol. I understood how the technology worked and the basic logistics of how to implement it. Regardless of whether I believed in it, I knew how it worked. I think that's what they were after.