from Hacker News

AI and Mass Spying

by hendler on 12/5/23, 2:09 PM with 328 comments

by somenameforme on 12/5/23, 3:45 PM
This is a political problem, not a technological one. The USSR (alongside Germany and others) managed effective at scale spying with primitive technology: paperwork for everything and every movement, informants, audio surveillance, and so on. The reason such things did not come to places like the US in the same way is not because we were incapable of such, but because there was no political interest in it.
And when one looks back at the past we've banned things people would never have imagined bannable. Make it a crime to grow a plant in the privacy of your own home and then consume that plant? Sure, why not? Make it a crime for a business to have the wrong opinion when it comes to who they want to serve or hire? Sure, why not?
Going the nuclear route and making the collection of data on individuals, aggregated or otherwise, illegal would hardly be some major leap of reach of jurisprudence. The problem is not that the technology exists, but that there is 0 political interest in curtailing it, and we've a 'democracy' where the will of the people matters very little in terms of what legislation gets passed.
by klik99 on 12/5/23, 3:57 PM
I don't know why this isn't being discussed more. The reality of the surveillance state is that the sheer amount of data couldn't realistically be monitored - AI very directly solves this problem by summarizing complex data. This, IMO, is the real danger of AI, at least in the short term - not a planet of paperclips, not a moral misalignment, not a media landscape bereft of creativity - but rather a tool for targeting anybody that deviates from the norm, a tool designed to give confident answers, trained on movies and the average of all our societies biases.
by thesuperbigfrog on 12/5/23, 3:16 PM
The new Google, Meta, Microsoft, etc. bots won't just crawl the web or social networks--they will crawl specific topics and people.
Lots of cultures have the concept of a "guardian angel" or "ancestral spirits" that watch over the lives of their descendants.
In the not-so-distant technofedualist future you'll have a "personal assistant bot" provided by a large corporation that will "help" you by answering questions, gathering information, and doing tasks that you give it. However, be forewarned that your "personal assistant bot" is no guardian angel and only serves you in ways that its corporate creator wants it to.
Its true job is to collect information about you, inform on you, and give you curated and occasionally "sponsored" information that high bidders want you to see. They serve their creators--not you. Don't be fooled.
by bonyt on 12/5/23, 3:31 PM
I think another aspect of this is mass criminal law enforcement enabled by AI.
Many of our criminal laws are written with the implicit assumption that it takes resources to investigate and prosecute a crime, and that this will limit the effective scope of the law. Prosecutorial discretion.
Putting aside for the moment the (very serious) injustice that comes with the inequitable use of prosecutorial discretion, let's imagine a world without this discretion. Perhaps it's contrived, but one could imagine AI making it at least possible. Even by the book as it's currently written, is it a better world?
Suddenly, an AI monitoring public activity can trigger an AI investigator to draft a warrant to be signed by an AI judge to approve the warrant and draft an opinion. One could argue that due process is had, and a record is available to the public showing that there was in fact probable cause for further investigation or even arrest.
Maybe a ticket just pops out of the wall like in Demolition Man, but listing in writing clearly articulated probable cause and well-presented evidence.
Investigating and prosecuting silly examples suddenly becomes possible. A CCTV camera catches someone finding a $20 bill on the street, and finds that they didn't report it on their tax return. The myriad of ways one can violate the CFAA. A passing mention of music piracy on a subway train can become an investigation and prosecution. Dilated pupils and a staggering gait could support a drug investigation. Heck, jaywalking tickets given out as though by speed camera. Who cares if the juice wasn't worth the squeeze when it's a cheap AI doing the squeezing.
Is this a better world, or have we just all subjected ourselves to a life hyper-analyzed by a motivated prosecutor.
Turning back in the general direction of reality, I'm aware that arguing "if we enforced all of our laws, it would be chaos" is more an indictment of our criminal justice system than it is of AI. I think that AI gives us a lens to imagine a world where we actually do that, however. And maybe thinking about it will help us build a better system.
by miki123211 on 12/5/23, 9:44 PM
I personally find the censorship implications (and the business models they allow) far more worrying than the surveillance implications.
It will soon be possible to create a dating app where chatting is free, but figuring out a place to meet or exchanging contact details requires you to pay up, in a way that 99% of people won't know how to bypass, especially if repeated bypassing attempts result in a ban. Same goes for apps like Airbnb or eBay, which will be able to prevent people from using them as listing sites and conducting their transactions off-platform to avoid fees.
The social media implications are even more worrying, it will be possible to check every post, comment, message, photo or video and immediately delist it if it promotes certain views (like the lab leak theory), no matter how indirect these mentions are. Parental control software will have a field day with this, basically redefining helicopter parenting.
by jillesvangurp on 12/5/23, 3:28 PM
Both were always going to be kind of inevitable as soon as the technology would get there. Rather than debating how to stop this (which is mostly futile and requires all of us to be nice, which we just aren't), the more urgent debate is how to adapt to this being the reality.
Related to this is the notion of ubiquitous surveillance. Where basically anywhere you go, there is going to be active surveillance literally everywhere and AIs filtering and digging through that constantly. That's already the case in a lot of our public spaces in densely populated areas. But imagine that just being everywhere and virtually inescapable (barring Faraday cages, tin foil hats, etc.).
The most feasible way to limit the downsides of that kind of surveillance is a combination of legislation regulating this, and counter surveillance to ensure any would be illegal surveillance has a high chance of being observed and thus punished. You do this by making the technology widely available but regulating its use. People would still try to get around it but the price of getting caught abusing the tech would be jail. And with surveillance being inescapable, you'd never be certain nobody is watching you misbehaving. The beauty of mass, multilateral surveillance is that you wouldn't ever be sure nobody is not watching you abuse your privileges.
Of course, the reality of states adopting this and monopolizing this is already resulting in 1984 like scenarios in e.g. China, North Korea, and elsewhere.
by HackerThemAll on 12/5/23, 2:55 PM
Soon in the name of "security" you'll have your face scanned on average every few minutes and it's going to be mandatory in many aspects of our lives. That's the pathetic world IT has helped to build.
by brunoTbear on 12/5/23, 3:49 PM
Schneier is wrong that "hey google" is always listening. Google does on-device processing with dedicated hardware for the wake-words and only then forwards audio upstream. Believe it or not, the privacy people at Google really do try to do the right things. They don't always succeed, but they did with our hardware and wake-word listening.
Am Google employee, not in hardware.
by dkjaudyeqooe on 12/5/23, 4:54 PM
This is why AI software must not be restricted, so that ordinary people and the civic minded can develop personal and public AI systems to counter corporate AI. The future of AI is adversarial.
Now freedom to develop AI software doesn't mean freedom to use it however you please and its use should be regulated, in particular to protect individuals from things like this. But of course people cannot be trusted, so you need to be able to deploy your own countermeasures.
by nojvek on 12/5/23, 7:26 PM
It’s not that AI will enable mass spying, mass spying is already there.
AI enables extracting all sorts of behavioral data across decades timespan for everyone.
The devils argument is in a world where the data is not used for nefarious purposes and only to prosecute crime as passed by governments, it leads to a society where no one is above the law and equal treatment for all.
However that seldom goes well since humans who control the system definitely want an upper edge.
by bmislav on 12/5/23, 4:30 PM
We actually recently published a research paper on exactly this topic (see https://llm-privacy.org/ for demo and paper). The paper shows that current LLMs already have the reasoning capability to infer personal attributes such as age, gender or location even when this information is not explicitly mentioned in the text. Crucially, they can do this way cheaper and way faster than humans. So I would say that spying scenarios mentioned in this blog post are definitely in the realm of possibility.
by zxt_tzx on 12/5/23, 3:41 PM
I tend to think the surveillance/spying distinction is a little fragile and this more a continuation of what Bruce has previously written insightfully about, i.e. the blurring of lines between private/public surveillance and, as the Snowden leaks have revealed, it's hard to keep what has been collected by private industry out of the hands of the state.
However, a more recent trend is companies that sell technologies to the state directly. For every reputable one like Palantir or Anduril or even NSO Group, there are probably many more funded in the shadows by In-Q-Tel, not to mention the Chinese companies doing the same in a parallel geopolitcal orbit. Insofar as AI is a sustaining innovation that benefits incumbents, the state is surely the biggest incumbent of all.
Finally, an under-appreciated point is Apple's App Tracking Transparency policy, which forbids third-party data sharing, naturally makes first-party data collection more valuable. So even if Meta or Google might suffer in the short-term, their positions are ultimately entrenched on a relative basis.
by miyuru on 12/5/23, 3:03 PM
The TV show "Person of Interest" portrayed this beautifully and it came out 12 years ago.
Strange and scary how fast the world develops new technology.
by 127361 on 12/5/23, 3:22 PM
Time to decentralize everything. I think we are already in the early stages of this new trend. We can run AI locally and hard drives are so large we can have a local copy of an entire library, with millions of ebooks, in our own home now.
That is in addition to generating our own energy off grid (so no smart meter data to monitor), thanks to the low cost of solar panels as well.
Bye bye Big Brother.
by troupo on 12/5/23, 3:06 PM
Will? It already has. China has had its surveillance for ages. And it's been spreading in other countries, too. Example: https://www.404media.co/fusus-ai-cameras-took-over-town-amer...
by renegat0x0 on 12/5/23, 3:39 PM
The author makes one mistake. He said that Google stopped spying on gmail.
- they started spying on user's gmail
- there was blowback, they reverted
- after some time they introduced "smart features", with ads again
Link https://www.askvg.com/gmail-showing-ads-inside-email-message...
I do not even want to check if "smart features" are opt-in, or opt-out.
by sarks_nz on 12/5/23, 5:24 PM
Nick Bostrom proposed the "Vulnerable World Hypothesis" which (amongst other things) says that a technology as powerful and accessible as AI requires mass surveillance to stop bad actors using it as a weapon.
https://nickbostrom.com/papers/vulnerable.pdf
It's disturbing, but also hard (for me) to refute.
by sambull on 12/5/23, 2:48 PM
next time they try to root out whatever 'vermin' defined; it will be a quick natural language prompt trained on ingested data from the last 2 decades to get that list of names and addresses / networks. AI is going to make targeting groups with differing ideologies dead simple. It will be used.
by yonaguska on 12/5/23, 3:08 PM
It's already happening. See this DHS memo issues on August 8th - page 3.
https://www.dhs.gov/sites/default/files/2023-09/23_0913_mgmt...
Fortunately the DHS has put together an expert team of non-partisan, honest, Americans to spearhead the effort to protect our democracy. Thank you James Clapper and John Brennan- for stepping up to the task.
https://www.dhs.gov/news/2023/09/19/secretary-mayorkas-annou...
And just in time for election season in the US AI is going to be employed to fight disinformation- for our protection of course. https://www.thedefensepost.com/2023/08/31/ussocom-ai-disinfo...
by 1-6 on 12/5/23, 3:25 PM
AI allows companies to skirt laws. For example, a company may be forbidden from collecting information on individual people but that rule doesn’t apply for aggregated data.
AI can be a deployed ‘agent’ that does all the collection and finally send scrubbed info to its mothership.
by px43 on 12/5/23, 3:12 PM
Never in the history of humanity has such powerful privacy tech existed for anyone who wants to use it.
Using common off the shelf, open source, heavily audited tools, it's trivial today, even for a non-technical 10 year old, to create a new identity and collaborate with anyone anywhere in the world. They can do research, get paid, make payments, and contribute to private communities in such a way that no existing surveillance infrastructure can positively link that identity to their government identity. Every day privacy tech is improving and adding new capabilities.
by intended on 12/5/23, 5:27 PM
If this argument hinges on summarization, then I have to ask - what is the blasted hallucination rate ?
I tried exactly this. Watched 4 talks from a seminar, got them transcribed, and used ChatGPT to summarize this.
Did 3 perfectly fine, and for the 4th it changed the speaker from mild mannered professor into VC investing superstar, with enough successes under his belt to not care.
How do you verify your summary is correct? If your false positive rate is 25% - 33%, thats a LOT of rework. 1 out of 3.
by TheLoafOfBread on 12/5/23, 4:01 PM
Then people will start using AI to generate random traffic noise to fool AI watching them.
by erikerikson on 12/5/23, 8:11 PM
Author: welcoming to knowing. This is unavoidable. Outside of extreme measures that will themselves mark, the network effects of use will overwhelm any effort to evade.
The question I think is how too navigate and what consequences will follow. We could use these capabilities to enslave but we could also use them to free and empower.
Scams rely on scale and the ineffective scaling social mechanisms to achieve profit. Imagine if the first identification of a scam informed every potential mark to which the scam began to be applied. Don't forget to concern yourself with false positives too, of course.
The injustice of being unable to take action in disputes due to a lack of evidence would evaporate. Massive privacy, consent, and security risks and issues result so will we be ready to properly protect and honor people and their freedoms?
At the end of this path may lay more efficient markets; increased capital flows and volumes; and a more fair, just, equitable, and maximized world more filled with joy, love, and happiness. There are other worse options of course.
by __jambo on 12/5/23, 7:23 PM
Because this is so depressing I am going to try think of positive aspects:
The flip side to this is the government had power because these activities required enormous resources. Perhaps it will go the other direction, if there is less of a moat other players can enter. Eg all it takes to make a state is a bunch of cheap drones and the latest government bot according to your philosophy.
Maybe it means government will massively shrink in personelle? Maybe we can have a completely open source ai government/legal system. Lawyers kind of suck ethically anyway, so maybe it would be better? With low barrier to entry, we can rapidly prototype such governments and trial them on smaller populations like iceland. Such utopias will be so good everyone will move there.
They still have to have physical prisons, if everyone is in prison this will be silly, but I suppose they can fine everyone, not so different from lowering wages which they already do.
by mmh0000 on 12/5/23, 4:40 PM
There's an excellent documentary on how sophisticated Government Surveillance is and how well it's tuned to be used against the general population:
https://www.imdb.com/title/tt0120660/
by mullingitover on 12/5/23, 8:16 PM
This kind of thing will probably never fly, because Americans expect that they can break the law and in most cases will not suffer any consequences. Anything that threatens this would be, in their eyes, oppression. Imagine if you immediately got a text informing you of your $300 speeding ticket within a few seconds of you going a mile per hour over. People would riot.
However, Americans expect that the law is enforced vigorously upon other people, especially people they hate. If AI enabled immediate immigration enforcement on undocumented migrants, large portions of the population would injure themselves running to the voting booth to have it added to the Constitution.
It's the whole expectation that for my group the law protects but does not bind, and for others it binds but does not protect.
by willmadden on 12/5/23, 6:22 PM
This could be used to expose government and monied corruption, not just for surveilling peasants.
What is the market for this short term?
I think this could greatly curtail government corruption and serve as a stepping stone to AI government. It's also a cool and disruptive startup idea.
by sebastianconcpt on 12/6/23, 1:34 PM
Yesterday I've watched again "Enemy of the State" from 1998 and the movie was entertaining but the fact that the evil force is coming from just one bad actor felt so naive compared to real life that it was tragicomical.
by thesz on 12/5/23, 11:54 PM
AIs, LLMs in patricular, go in one direction, just like humans usually do.
What other humans do to cicumvent that? Yes, they found a way to alternate direction, in case of London cockney, by using rhymes [1].
[1] https://www.theguardian.com/education/2014/jun/09/guide-to-c...
If you need to fool your AI of choice, rhyme the concepts!
For a demo, ask your AI of choice about "Does basin of gravy likes satin and silk?" (decode yourself)
The article above is from 2014 and is hardly is used when I asked questions using Cockney parlance.
You are welcome. ;)
by I_am_tiberius on 12/5/23, 8:07 PM
Somewhat related: A "Tell HN" I posted today but was shadow-banned after it started trending: https://news.ycombinator.com/item?id=38531407
by darklycan51 on 12/5/23, 3:13 PM
Think about it this way.
Every service has access to the IPs you've used to log on, most services require an email, phone number some debit/credit cards and or similar personal info. Link that with government databases on addresses/real names/ISP customers and you basically can get most peoples accounts, on virtually any service they use.
We then also have things such as the patriot act in effect, the government could if they wanted run a system to do this automatically, where every message is scanned by an AI that catalogues them.
I have believed for some time now that we are extremely close to a complete dystopia.
by FrustratedMonky on 12/5/23, 8:53 PM
This very scenario was one of the key threats in book Homo Dues and that was over 5 years ago.
Russia could do surveillance, but was limited by manpower.
Now AI solves this, there can be an AI bot dedicated to each individual.
Wasn't there another article on HN just day, that Car Makers, Phone, Health monitors all can now aggregate data to know 'your mood' when in an accident? To know where you are going, how you are feeling?
This is the real danger with AI. Even current technology is good enough for this kind of surveillance.
by Arson9416 on 12/5/23, 3:52 PM
I have a friend that is working as a contractor building AI-powered workplace spying software for the explicit purpose of behavior manipulation. It gives the employees and employers feedback reports about their behavior over chat and video. For example, if they used microaggressions, misgendered someone, or said something crass. This same friend will then talk about the dangers of dystopian technology.
People don't know what they're creating. Maybe it's time it bites them.
by elric on 12/5/23, 10:23 PM
We can barely get people to care about the implications of types of surveillance that they do understand (CCTV everywhere, Snowden's revelations, etc). It's going to be nigh impossible to get people to care about this enough to make a difference.
Heck, even if they did care, there's nothing they can realistically do about it. The genie's out of the bottle.
by RandomLensman on 12/5/23, 2:59 PM
Not sure why this sidesteps the potentially quite different legal settings for spying and surveilling.
by EVa5I7bHFq9mnYK on 12/5/23, 5:59 PM
Thinking of finally creating a private server for receiving emails. For sending email it is known to be very difficult, but I send very very few emails these days, will outsource it to sendgrid or similar.
What's the best docker image for that, simple in configuration?
by wseqyrku on 12/5/23, 8:35 PM
So far you were a line in the log. Now someone is actually looking at you with three eyes.
by righthand on 12/5/23, 4:38 PM
Only if you continue to invest time and energy into. Not everyone puts 99% of their life online. When’s the last time you left your house to do something without a cellular? Or compromised by not ordering something you think you need?
by pockmockchock on 12/5/23, 4:26 PM
Next would be to use/create devices that are using Radio for Communication and payments, devices like Satslink as an example. It drives innovation at the same time, so I wouldn't be too concerned.
by naveen99 on 12/5/23, 2:39 PM
The limiting factor is usually, someone has to pay for the spying and the punishment. A lot (most?) of troublemakers just aren’t worth the trouble of spying or punishing.
by fsflover on 12/5/23, 5:07 PM
Another ongoing discussion: https://news.ycombinator.com/item?id=38530795
by blueyes on 12/5/23, 6:26 PM
this has been true for a long time. most data is garbage, so the data collection phase of mass surveillance punched below its weight in terms of consequences felt by the surveilled. AI is one way to extraction actionable meaning from that data, and when people start feeling the everyday consequences of all that collection + meaning extraction, they will finally understand.
by _Nat_ on 12/5/23, 3:26 PM
Seems inevitable enough that we may have to accept it and try to work within the context of (what we'd tend to think of today as) mass-spying.
I mean, even if we pass laws to offer more protections, as computation gets cheaper, it ought to become easier-and-easier for anyone to start a mass-spying operation -- even by just buying a bunch of cheap sensors and doing all of the work on their personal-computer.
A decent near-term goal might be figuring out what sorts of information we can't reasonably expect privacy on (because someone's going to get it) and then ensuring that access to such data is generally available. Because if the privacy's going to be lost anyway, then may as well try to address the next concern, i.e. disparities in data-access dividing society.
by uticus on 12/5/23, 9:22 PM
This sums up so many things well and clearly. It's so quotable.
- The money trail: "Their true customers—their advertisers—will demand it."
- The current state of affairs: "Surveillance has become the business model of the internet..."
- The fact that not participating, or opting-out, still yields informational value, if not even more so: "Find me all the pairs of phones that were moving toward each other, turned themselves off..."
This isn't a technological problem. Technology always precedes the morals and piggybacks on the fuzzy ideas that haven't yet developed into concrete, well-taught axioms. It is a problem about how our society approaches ideals. Ideals, not ideas. What do we value? What do we love?
If we love perceived security more than responsibility, we will give up freedoms. And gladly. If we love ourselves more than future generations, we will make short-sighted decisions and pat ourselves on the back for our efficiency in rewarding ourselves. If we love ourselves more than others, we won't even care much about social concerns. We'll fail to notice anything that doesn't move the needle against my comfort much.
It's more understandable to me than ever how recent human horrors - genocides, repressive regimes, all of it - came about to be. It's because I'm a very selfish person and I am surrounded by selfish people. Mass spying is a symptom - not much of a cause - of the human condition.
by moose44 on 12/5/23, 3:58 PM
Is mass spying not already going on?
by jacobwilliamroy on 12/5/23, 3:04 PM
A friend of mine was recently a witness for the FBI. He was working in a small office in the middle of nowhere and happened to have a very loud argument with the suspect. A few minutes later he left the building and when he was about to start his car, he got a call from an agent asking him if he wanted to be a witness in the case they were working on.
by forward1 on 12/5/23, 6:08 PM
Mass spying is yesterday's news. The thing we all need to worry about is behavioral management at scale, which influences politics, relationships, religions and much more. This is the true "hidden" evil behind social media and "AI" which few apprehend let alone can do something about.
by pier25 on 12/5/23, 5:11 PM
Will? I'd be surprised if NSA etc hadn't been using AI for years now.
by graphe on 12/5/23, 3:45 PM
>Mass surveillance fundamentally changed the nature of surveillance.
Computers create and organize large amounts of information. This is useful for large organizations and unempowering to the average person. Any technology with these traits are harmful to individuals.
by ysofunny on 12/5/23, 4:48 PM
"if everybody is being spied on, nobody is being spied on"
???
by barelyauser on 12/5/23, 10:31 PM
The average HN user defends the "common guy" or "the masses", perhaps because he fears being perceived as condescending. I've come to the conclusion that the masses don't deserve any of this. Many drink and drive, indulge in destructive addictions (not only to themselves). Many can't bother recycling or even maintaining a clean home environment, waste time and resources in every activity they engage in and don't even care for their neighbors well being (loud music, etc).
Concluding remarks. As man succeeded in creating high mechanical precision from the chaotic natural environment, he will succeed in creating a superior artificial entity. This entity shall "spy" (better described as "care" for) every human being, maximizing our happiness.
by CrzyLngPwd on 12/5/23, 4:00 PM
"has", not "will".
by aaroninsf on 12/5/23, 11:38 PM
"This is a political problem..."
No. The US has always had political problems wrt surveillance; and many places have had much worse ones.
Now, all of us can anticipate a very different, all but inevitable, and very much worse political problem.
AI is a force multiplier and an accelerant.
As such it is a problem, an obvious one, and a very very big one. This is just one of many ways in which force multiplication and acceleration, so pursued, and so lauded, in so many domains, may work their magic on preexisting social and political evils. And give us brand new ones, per Larkin.
by CrzyLngPwd on 12/5/23, 3:45 PM
More people will be criminalised, and fines will be just another tax we must pay.
by indigo0086 on 12/5/23, 5:14 PM
"Bitcoin will enable terrorism" "Ghost guns will enable criminals to commit murder" "Social media will enable hate speech"
AI will be a useful and world changing innovation which is why FUD rag articles like this will become more prevalent until it's total adoption, even by the article writer themselves
by gumballindie on 12/5/23, 3:06 PM
I beg to differ. The correct term is not “will” but “is”.
by godelski on 12/5/23, 7:13 PM
There's a lot of speculation in the comments so I want to talk about the technology that we have __TODAY__. I post a lot about being in ML research and while my focus is on image generation I'm working with another team doing another task but not going to state it explicitly for obvious reasons.
What can AI/ML do __today__?
We have lots of ways to track people around a building or city. The challenge is to do these tasks through multi-camera systems. This includes things like people tracking (person with random ID but consistent across cameras), face identification (more specific representation that is independent of clothing, which usually identifies the former), gait tracking (how one walks), device tracking (based on bluetooth, wifi, and cellular). There is a lot of mixed success with these tools but I'll let you know some part that should concern you: right now these are mostly ResNet50 models, datasets are small, and they are not using advanced training techniques. That is changing. There are legal issues and datasets are becoming proprietary but the size and frequency of gathering data is growing.
I'm not going to talk about social media because the metadata problem is an already well discussed one and you all have already made your decisions and we've witnessed the results of those decisions. I'm also not going to talk about China, the most surveilled country in the world, the UK, or any of that for similar reasons. We'll keep talking in general, that is invariant to country.
What I will talk about is that modern ML has greatly accelerated the data gathering sector. Your threat models have changed from governments rushing to gather all the data that they can, to big companies joining the game, to now small mom and pop shops doing so. I __really__ implore you all to look at what's in that dataset[0]. There's 5B items, this tool helps retrieve based on CLIP embeddings. You might think "oh yes, Google can already do this" but the difference is that you can't download Google. Google does not give you 16.5TB of clip filtered image,text, & metadata. Or look into the RedPajama dataset[1] which has >30T tokens and 5TB of storage. With 32k tokens being about 50 pages, that's about 47 billion pages. That is, a stack of paper 5000km tall, reaching 5x the height of the ISS and is bigger than the diameter of the moon. I know we all understand that there's big data collection, but do you honestly understand how big these numbers are? I wouldn't even claim to because I cannot accurately conceptualize the size of the moon nor the distance to the ISS. They just roll into the "big" bin in my brain.
Today, these systems can track you with decent accuracy even if you use basic obscurification techniques like glasses, hats, or even a surgical mask. Today we can track you not just by image, but how you walk, and can with moderate success do this through walls (meaning no camera to see if you want to know you're being tracked). Today, these systems can de-anonymize you through unique text patterns that you use (see Enron dataset, but scale). Today, these machines can uncanny valley replicas of your speech and text. Today we can make images of people that are convincingly real. Today, these tools aren't exclusive to governments or trillion dollar corporations, but available to any person that is willing to spend a few thousand dollars on compute.
I don't want to paint this as a picture of doom and gloom. These tools are amazing and have the potential to do extraordinary good, at levels that would be unimaginable only a few decades ago. Even many of these tools that can invade your privacy are benefits in some ways, but just need to consider context. You cannot build a post scarce society when you require humans to monitor all stores.
But like Uncle Ben says, with great power comes great responsibility. A technology that has the capacity to do tremendous good also has the power to do tremendous horrors.
The choice is ours and the latter prevails when we are not open. We must ever push for these tools to be used for good, because with them we can truly do amazing things. We do not need AGI to create a post scarce world and I have no doubt that were this to become our primary goal, we could easily reach it within our lifetime without becoming a Sci-Fi dystopia and while tackling existential issues such as climate. To poke the bear a little, I'd argue that if your country wants to show dominance and superiority on the global stage, it is not done so through military power but technology. You will win the culture wars of all culture wars and whoever creates the post scarce world will be a country that will never be forgotten by time. Lift a billion people out of poverty? Try lifting 8 billion not just out of poverty, but into the lower middle class, where no child dreams of being hungry. That is something humans will never forget. So maybe this should be our cold war, not the one in the Pacific. If you're so great, truly, truly show me how superior your country/technology/people are. This is a battle that can be won by anyone at this point, not just China vs the US, but even any European power has the chance to win.
[0] https://rom1504.github.io/clip-retrieval/
[1] https://github.com/togethercomputer/RedPajama-Data
by boringg on 12/5/23, 6:50 PM
Is this really an insight to ANYONE on hackernews? How is this article providing anything new to the conversation except to bring it back into discussion?
by blondie9x on 12/5/23, 5:19 PM
Real question you have to ask yourself. Is AI spying and AI law enforcement Minority Report in real life?
by Spivak on 12/5/23, 2:52 PM
This happened during the last big technological advancement -- search. Suddenly it became possible for a government to theoretically sift through all of our communications and people online made constant reference to it talking directly to their "FBI agent."
But it was and still is a nothingburger and this will be the same because it doesn't enable anything except "better search." We've had comparable abilities for a decade now. Yes LLMs are better but semantic search and NLP have been around a while and the world didn't end.
All the examples of what an LLM could do are just querying tracking databases. Uncovering organizational structure is just a social graph, correlating purchases is just querying purchase databases, listing license plates is just querying the camera systems. You don't need an LLM for any of this.
by Nilrem404 on 12/5/23, 3:14 PM
Surprised Pikachu Face
Seriously nothing new or shocking about this piece. Spying is spying. Surveillance is surveillance. If you've watched the news at all in the past 2 decades, you know this is happening.
Anyone who assumes that any new technology isn't going to be used to target the masses by increasingly massive and powerful authoritarian regimes is woefully naive.
Another post stating what we all already know isn't helping or fostering any meaningful conversation. It will just be rehashes. Let me skip to the end here for you:
There is nothing we can do about it. Nothing will change for the better.
Go make a coffee or tea