Snowden: "They've gone full mask-off: do not ever trust OpenAI or its products"

https://twitter.com/Snowden/status/1801610725229498403

by underlogic • 2 years ago

275 107 2 years ago

I don't need Snowden to tell me not to trust OpenAI. The feeling I got when reading just a few words when I first found out about the company was the same feeling one gets when walking into a used car shop run by the guy who bullied you in high school.

ranger_danger• 2 years ago

Many people DO need someone to tell them though. Would you rather nothing was said at all?

vouaobrasil2• 2 years ago

Of course not. I want more people to say it. I want everyone to say it, to condemn OpenAI, to condemn big tech ad nauseum. I want everyone to stop using big tech products as much as possible and I want big tech to go bankrupt.

threeseed3• 2 years ago

Of course there is a legitimate reason for having Nakasone on the board.

OpenAI is highly sensitive to regulation which is why they have such a large lobbying team trying to push the US government in their direction. He (a) brings political connections and (b) gives confidence that any advancements won't threaten US national security. It is pretty common amongst enterprise boards.

The reality is that the perceptions of what AI can do is impacting the world far more than what it can actually do.

bagels• 2 years ago

If you want to sell to the government, this is also a good idea.

contingencies• 2 years ago

Let's crowdsource some prompt engineering to win an election.

threeseed3• 2 years ago

According to the OpenAI CTO, ChatGPT5 is not going to break new ground.

So if you were Sam the smart thing to do is to get your platform embedded absolutely everywhere before the heat starts to come out of this iteration of AI.

tmpz22• 2 years ago

Reminds me of Henry Kissinger and others being put on the Theranos board.

How do we know these board seats are real and not just rubber stamping everything for the free pay check?

rchaud3• 2 years ago

Theranos' board had former generals on it, the plan was for them to open doors to getting Theranos products sold to the military.

jlaporte• 2 years ago

Yup.

And aside from being NSA director he was Commander of U.S. Cyber Command, experience highly relevant to both OpenAI's security and safety posture, and to selling to government and defense buyers.

pseudalopex• 2 years ago

It is common. Its legitimacy is controversial.

squigz• 2 years ago

The reality is that the perceptions of what AI can do is impacting the world far more than what it can actually do.

Could you elaborate on this? What impact is the perception of what AI can do having? And what impact is it actually having?

jononor• 2 years ago

Not parent, just giving my opinion. The perception of what it may become is causing pressure to change wide reaching laws and legal practice.

A) Copyright. The AI companies are blatantly disregarding practically everyones copyright when they train on all the data they can find online. The rights holders have not given them permission to do so.

B) Privacy. Companies are just taking all the data of their customers and using for training. This is directly in opposition with GDPR, recently established in the Eurozone. Companies, notably Meta, are arguing - oh but AI is so important, that we do not have to ask for users consent.

C) Compute/algorithms. Regulations on "AI" itself. With several actors, notably OpenAI/Altman saying it is so dangerous that only corporations like them can do it responsibly, aiming to do regulatory capture which benefits them.

Large forces from the biggest companies in the world are attempting to change the legal landscape, and manipulate the perception of politicians and the public to ensure sufficient support to get the changes through.

beefnugs• 2 years ago

The most charitable positive spin on this is that his connections will help stop government from going fact-less banhammer on AI in general. But jesus I would assume most peoples brains can go real negative on this one.

hecanjog• 2 years ago

The rest of the tweet (I assume) is quoted here in case you have twitter blocked too:

https://fortune.com/2024/06/14/edward-snowden-eviscerates-op...

hedora• 2 years ago

Heads up: The “article” page has a warning that they’re moving to x.com. Time to add an entry to your filter list.

(I clicked the link to see what fresh breakage they’d display instead of the current tweet. Oddly, the article link works. That hasn’t happened to me in ages.)

_lvbh• 2 years ago

I don’t have Twitter blocked but I can’t see anything on it. 80% of the time I get some generic error and “try again”

sroerick• 2 years ago

Is the implication that they’re training OpenAI on NSA data?

rglover• 2 years ago

Other way around. OpenAI and others are the new Google search. The likely suggestion is that OpenAI chats will just be another XKEYSCORE input and OpenAI's algorithms will assist in parsing/forming conclusions on existing data on behalf of the government (a likely bet as there are plenty of rumors swirling about government agencies integrating AI into their processes).

nashashmi2• 2 years ago

What choice do we have?

OutOfHere6• 2 years ago

You have a choice, and it is simple. Earn money. Buy GPUs. Download llama3 or comparable open models. Run them. After you do, to the maximum extent possible, stop using the OpenAI LLMs. Stop willingly feeding the surveillance industrial complex.

janalsncm• 2 years ago

Buy GPUs

I don’t think this is practical advice for the vast majority of people

EVa5I7bHFq9mnYK2• 2 years ago

That's too expensive, quality of llama3 answers is way below ChatGPT 4o or Clause Opus, and consumer GPUs are not competitive with specialized chips.

Need the ability to submit queries anonymously, with disposable accounts. How about an AI add-on for a VPN service?

worksonmine• 2 years ago

How did you manage to live before GPT? Was it that bad? I still haven't used it and I'm doing fine.

hammyhavoc• 2 years ago

But it's tHe FuTuRe bro

nashashmi2• 2 years ago

I’m not worried about losing chatgpt. Im worried that they will leverage gpt on nsa data. And open ai will be the agent.

lancesells• 2 years ago

Unless your job maybe mandates it why do you have to participate?

rchaud3• 2 years ago

You cannot opt out of a panopticon.

r721• 2 years ago

DuckDuckGo AI Chat: anonymous access to popular AI chatbots

https://spreadprivacy.com/ai-chat/

afpx• 2 years ago

Wait for yandex.ai or ccp.ai I guess? I hear mossad.ai is going to be pretty killer.

joemi3• 2 years ago

From the same tweet: "This is a willful, calculated betrayal of the rights of every person on Earth."

Sounds like a bit of a hyperbole to me.

elevatedastalt• 2 years ago

Dunno the details of this specific issue, but hyperbole is pretty standard for Snowden. I don't blame him too much. The guy has been through hell. But let's just say his Twitter proclamations aren't exactly neutrally worded.

FormerBandmate• 2 years ago

he's also probably watched like a hawk in russia. Who knows if that's even still his opinions

K0balt8• 2 years ago

To be fair, his experience in life would be considered hyperbole by most people whose most difficult challenge is whether or not they have time to get a coffee on their 20 minute commute.

His words reflect his personal experience as a person who feels compelled to point out legitimate dangers to a free and democratic society.

It turns out that “free and democratic “ societies react harshly to criticism of their powerful agencies.

Ironically, he has ended up having to flee to a place that is much, much less “free and democratic” but letting him freely criticise the rest of the world is politically useful for now, so he is allowed the bully pulpit.

andsoitis• 2 years ago

Concur. While being alarmist can help elevate a cause, it should be used sparingly lest you be seen as crying wolf.

Brian_K_White5• 2 years ago

Has he ever cried wolf?

K0balt8• 2 years ago

Crying wolf?

Please do elaborate. I am not aware of any information that Snowden has published that was not factually correct and informative.

I think that of all people, Snowden would have a much more nuanced understanding of what it means for a former NSA chief to join the board of a company like OAI.

imchillyb2• 2 years ago

OpenAI did the reasonable and logical thing here. They're easing themselves around regulatory hurdles, possibly avoiding Senate committee interrogations, and legitimizing themselves all with a single board seat.

Snowden's being disingenuous. He knows all of this and has chosen to ignore the above in service of an agenda.

What _actual_ choice did OpenAI have here? If they do not accept an NSA board member today, what would OpenAI be forced to accept later?

OutOfHere6• 2 years ago

OpenAI is very guilty of pushing for regulation of AI, leading to all these issues in the first place. They themselves created the poison of regulation for which they now purchased the antidote. They had a choice to not scaremonger or push for regulation, but to fight for freedom on constitutional grounds.

Also, they should focus on developing AGI, not on buying wortheless executives that the consumer or taxpayer has to pay for.

imchillyb2• 2 years ago

OpenAI is very guilty of pushing for regulation of AI, in order to be at the forefront of that technology.

Regulatory capture is SOP for tech bros.

AndrewKemendo• 2 years ago

OpenAI - whose CEO is Sam Altman, is financially backed by Microsoft, using Microsoft infrastructure and search and is now embedded in Apple, has crossed the line by adding the former NSA/CSS Commander to the board?

I’m a huge Snowden fan but cmon dude…Give me a break

OpenAI was never some indie thing

OpenAI is the most perfect physical manifestation of cynical, alienating, self-important narcissistic capitalism that has ever been seen

You would think LVMH would take this title, but there’s no more hypnotizing scrying device for burgeoning psychopathic narcissists (CEOs, “influencers”, rappers etc..) than the endless fawning solipsism on demand from a simulacrum of a human

20after42• 2 years ago

I think yours is the most under-appreciated take on this entire hn thread.

endless fawning solipsism on demand from a simulacrum of a human

ChatGPT: Every narcissist's best friend. Why settle for a stupid and fallible human when you can have an algorithmically optimized yes-man trained on all of human knowledge and available 24/7/365.

omoikane• 2 years ago

Source: The Verge

https://www.theverge.com/2024/6/13/24178079/openai-board-pau...

https://news.ycombinator.com/item?id=40680962 - Former head of NSA joins OpenAI board

Tronickle• 2 years ago

Yes, it's somewhat clear that some type collaboration between OpenAI and government is in place. But why it can be so dramatically bad?

elforce002• 2 years ago

Prob. because their track record is not that transparent, hehe.

oriel2• 2 years ago

Another one bites the dust. Added to the list, paid account canceled, data deleted, account deleted.

Its becoming easier and easier to step back from a service and consider if its essential (its not, as a rule), and remove it WHEN it becomes an abusive relationship.

OpenAI was not a good dog.

delichon2• 2 years ago

I ask so many questions of chatbots across so many topics and it has proven valuable many times. But I'm certainly untrusting and wary of them and don't really doubt that I am being surveilled in detail. But I struggle to find scenarios in which it could be used against me. I'd pay more for more privacy and am interested in hosting my own. But of my 99 problems this is pretty far down the list.

Once I sent some saliva to 23andMe. At least some of that data was stolen and I'm probably on a list of ashkenazim for sale somewhere. That's unpleasant. While I can't think of a problem with being surveilled by chatbots now maybe I'm just not being imaginative enough.

unyttigfjelltol• 2 years ago

If all your friends and all their friends said the sky was purple, even though at odds with your observation, would you be bold enough to declare it blue?

We'll be immersed in AI feedback and we'll all be influenced by the advice it provides. There won't really be a "source" in the sense Google sends you to BuzzFeed and you know you're reading trash. It'll all have the veneer of authority because ChatGPT said so. The product is trust. Would you ask a malicious thug the same questions and reveal your private details? No.

You can feel as untrusting as you want, but the streams of AI advice are like water and we're just like fish. It's influencing you, and if you don't realize how then you probably could pay closer attention.

k8svet• 2 years ago

Such a good reply. Even if somehow you avoid even creeping bit of AI (I'm trying as hard as I can), you will inevitably live amongst others that don't have the same information literacy or skepticism as some of us. And while using a chat bot might be personally innocent, any such users are absolutely feeding and training the beast and enabling OpenAI's next oversteps.

rchaud3• 2 years ago

OpenAI will presumably offer more than a chatbot in the future. Even so, you can already upload files into it for analysis.

I for one don't think it's a coincidence that Microsoft is pushing Windows Recall for consumers (keeping a permanent index of deleted files), pushing OneDrive as the default save location for O365 enterprise, and sinking billions into OAI.

oriel2• 2 years ago

At the risk of a tangent, the real problem with services like 23andMe isnt just that it has your (irrevokable) DNA or traits in it, its that they now have the DNA and traits of your entire extended family up and down the tree(s) to a significant degree. Using a service like that is essentially selling them out, and they have no say in it.

delichon2• 2 years ago

My older brother was pissed that I sent it in. I had a potential cousin contact me via 23andMe and I asked my brother about some possible common relations. He thought that we were probably being scammed and that I was a reckless idiot.

toastal• 2 years ago

Same as using GMail without PGP. GMail users are selling out the the rest of our data to Google by having one side of that conversation. It’s the same reason I am generally against using a gateway between Discord & IRC/XMPP/Matrix.

shirro• 2 years ago

Snowden is one of those guys like Stallman who you can disagree with and think their position is extreme but it is foolish to completely dismiss them. There is currently a gold rush to circumvent IP rights and exploit the works of billions without attribution or remuneration while society and legislators are sleeping.

cainxinth• 2 years ago

Richard Stallman has never done a single thing to make me believe he’s anything other than motivated by deeply held principles rooted in the right to privacy and other individual liberties.

Snowden’s true motivations are far less certain.

redeeman• 2 years ago

Snowden’s true motivations are far less certain.

based on what? do you understand what he risked and is sacrifising for letting you know what was illegally done?

newscracker• 2 years ago

In light of this, Apple’s choice to integrate with ChatGPT from OpenAI looks worse. Even with its privacy promises of not revealing much information and asking the user explicitly, these and other events in OpenAI should trigger serious concerns with Apple executives. The top Apple executives have said that they started with ChatGPT because it’s the best. Maybe this partnership won’t be around this time next year.

al_borland• 2 years ago

It sounded like they made it flexible so they could slot in other models. Maybe it will even allow users to choose the model, once there are more options.

thejazzman• 2 years ago

Remember that time Steve was excited to integrate Facebook and Twitter into iOS as first class citizens?

This reminds me so much of that moment

Then I remember how bad Siri is. OpenAI has basically shamed Apple out of no where. They desperately need this just to look competitive in the space .. which is weird and unfortunate

webninja• 2 years ago

They could use Mixtral, Minstral, XWin, Mythomax, or Claude 3 Haiku 196M. They’re all great and most/all are better than Siri.

I think the best thing Apple could do is let the user choose which AI to load in. It could be like iTunes or the App Store and called the AI store. Then AI providers can compete on privacy & effectiveness. The store would need lots of features to enable comparison though or users won’t be able to decide which ones might be worth picking.

secondary_op• 2 years ago

Snowden memoir not simply called "Permanent Record" if you like to read, it's in chapter: Part Three - Fourth Estate, or in audiobook at 08:30:24

He writes about CIA CTO, Gus Hunt talk at GigaOM's Structure:Data conference in 2013, still available to witness https://youtu.be/GUPd2uMiXXg?t=1258

TLDR: “At the CIA,” he said, “we fundamentally try to collect everything and hang on to it forever.”

The second event happened one year later, in March 2013—one week after Clapper lied to Congress and Congress gave him a pass. A few periodicals had covered that testimony, though they merely regurgitated Clapper’s denial that the NSA collected bulk data on Americans. But no so-called mainstream publication at all covered a rare public appearance by Ira “Gus” Hunt, the chief technology officer of the CIA. I’d known Gus slightly from my Dell stint with the CIA. He was one of our top customers, and every vendor loved his apparent inability to be discreet: he’d always tell you more than he was supposed to. For sales guys, he was like a bag of money with a mouth. Now he was appearing as a special guest speaker at a civilian tech event in New York called the GigaOM Structure: Data conference. Anyone with $40 could go to it. The major talks, such as Gus’s, were streamed for free live online.
I got insight, certainly, but of an unexpected kind. I had the opportunity of witnessing the highest-ranking technical officer at the CIA stand onstage in a rumpled suit and brief a crowd of uncleared normies—and, via the Internet, the uncleared world—about the agency’s ambitions and capacities. As his presentation unfolded, and he alternated bad jokes with an even worse command of PowerPoint, I grew more and more incredulous.
“At the CIA,” he said, “we fundamentally try to collect everything and hang on to it forever.” As if that wasn’t clear enough, he went on: “It is nearly within our grasp to compute on all human generated information.” The underline was Gus’s own. He was reading from his slide deck, ugly words in an ugly font illustrated with the government’s signature four-color clip art.

now, coupled with pentagon shenanigans https://www.reuters.com/investigates/special-report/usa-covi...

Nevertheless, the Pentagon’s clandestine propaganda efforts are set to continue. In an unclassified strategy document last year, top Pentagon generals wrote that the U.S. military could undermine adversaries such as China and Russia using “disinformation spread across social media, false narratives disguised as news, and similar subversive activities [to] weaken societal trust by undermining the foundations of government.”

Good luck trusting OpenAI's generative seductive female operative with Johansson voice.

egberts1• 2 years ago

Until AI provides citations, AI is inherently by design untrustworthy.

Snowden: "They've gone full mask-off: do not ever trust OpenAI or its products"

Related Stories